A resurgent HIV-1 epidemic among men who have sex with men in the era of potent antiretroviral therapy

Downloaded from UvA-DARE, the institutional repository of the University of Amsterdam (UvA)http://hdl.handle.net/11245/2.67805

File ID uvapub:67805Filename ThesisVersion unknown

SOURCE (OR PART OF THE FOLLOWING SOURCE):Type PhD thesisTitle Impact of antiretroviral therapy on HIV-1 transmission dynamicsAuthor(s) D.O. BezemerFaculty FNWI: Institute for Biodiversity and Ecosystem Dynamics (IBED), FNWI:

Institute for Biodiversity and Ecosystem Dynamics (IBED), AMC-UvAYear 2009

FULL BIBLIOGRAPHIC DETAILS: http://hdl.handle.net/11245/1.323167

Copyright It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/orcopyright holder(s), other than for strictly personal, individual use, unless the work is under an open content licence (likeCreative Commons). UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)(pagedate: 2015-07-27)

http://hdl.handle.net/11245/2.67805

http://hdl.handle.net/11245/1.323167

http://dare.uva.nl

DANIELA BEZEMER

DA

NIELA

BEZEM

ER

Impact of antiretroviral therapy on HIV-1 transmission dynamics

© 2009 Daniela Bezemer, Utrecht

ISBN 978-90-806415-9-4

http://dare.uva.nl/dissertaties

Reprints were made with permission from the publishersCover design by Ella Propeller http://ella.laLayout and printed by Gildeprint

The research described in this thesis was performed at the GGD and SHM, Amsterdam, the Nether-lands. This work was financially supported by grant 7014 from AIDS Fund Netherlands, and by IBED/FNWI.

Impact of antiretroviral therapy on HIV-1 transmission dynamics

AcAdemIscH proefscHrIftter verkrijging van de graad van doctor aan de Universiteit van Amsterdam

op gezag van de Rector Magnificus prof. dr. D.C. van den Boom ten overstaan van een door het college voor promoties ingestelde commissie,

in het openbaar te verdedigen in de Agnietenkapel op donderdag 3 september 2009, te 10:00 uur

door

daniela olga Bezemergeboren te Willemstad, Curaçao

promotiecommissie

promotores: Prof. dr. R.A. Coutinho Prof. dr. M.W. Sabelis

co-promotores: Dr. M. Prins Prof. dr. F. de Wolf

overige leden: Prof. dr. B. Berkhout Prof. dr. C.A.B. Boucher Prof. dr. A.C. Ghani Dr. M.E.E. Kretzschmar Prof. dr. J.M.A. Lange Prof. dr. H. Schuitemaker

Faculteit der Geneeskunde

What’s love got to do with it?

7

contents

I Introduction

II declining trend in transmission of drug-resistant HIV-1 in Amsterdam Daniela Bezemer, Suzanne Jurriaans, Maria Prins, Lia van der Hoek, Jan M. Prins, Frank de Wolf, Ben Berkhout, Roel A. Coutinho, Nicole K.T. Back, AIDS (2004) 18:1571-1577

III evolution of transmitted HIV-1 with drug-resistance mutations in the absence of therapy: effects on cd4+ t-cell count and HIV-1 rNA loadDaniela Bezemer, Anthony de Ronde, Maria Prins, Kholoud Porter, Robert Gifford, Deenan Pillay, Bernard Masquelier, Hervé Fleury, Francois Dabis, Nicole Back, Suzanne Jurriaans, Lia van der Hoek on behalf of the CASCADE collaboration, Antiviral Therapy (2006) 11(2):173-178

IV combination antiretroviral therapy failure and HIV super-infectionDaniela Bezemer, Ard van Sighem, Frank de Wolf, Marion Cornelissen, Antoinette C. van der Kuyl, Suzanne Jurriaans, Maria Prins, Roel A. Coutinho, and Vladimir V. Lukashov, AIDS (2008) 22(2):309-311

V A resurgent HIV-1 epidemic among men who have sex with men in the era of potent antiretroviral therapy Daniela Bezemer, Frank de Wolf, Maarten C. Boerlijst, Ard van Sighem, T. Deirdre Hollingsworth, Maria Prins, Ronald B. Geskus, Luuk Gras, Roel A. Coutinho, and Christophe Fraser, AIDS (2008) 22(9):1071-1077

VI 27 years of the HIV epidemic amongst men having sex with men in the Netherlands: an in depth mathematical model-based analysis Daniela Bezemer, Frank de Wolf, Maarten C. Boerlijst, Ard van Sighem, T. Deirdre Hollingsworth, and Christophe Fraser, submitted (2009)

VII transmission networks of HIV-1 among men having sex with men in the NetherlandsDaniela Bezemer, Ard van Sighem, Vladimir V. Lukashov, Lia van der Hoek, Nicole Back, Rob Schuurman, Charles A.B. Boucher, Eric C.J. Claas, Maarten C. Boerlijst, Roel A. Coutinho, Frank de Wolf for the ATHENA observational cohort, submitted (2009)

VIII discussion

Abstract samenvatting curriculum Vitae Acknowledgements

9

21

29

35

39

47

81

99

111113115119

9

I Introduction

Twenty-eight years ago the human immunodeficiency virus (HIV) pandemic took the world by surprise [1]. After the eradication of smallpox and the introduction of childhood vaccines, developed countries thought to have won from infectious diseases. But new infectious diseases emerged besides HIV, Hepatitis C virus, Severe Acute Respiratory Syndrome (SARS) virus, Ebola virus, and new variants of Influenza virus [2]. In protection measures against infectious diseases it is as the Red Queen explains to Alice in Wonderland: “it takes all the running you can do, to keep in the same place” [3].Virulence management is needed in order to understand the conditions for emergence and spread of infectious diseases [4, 5]. Molecular studies taught us that viruses are as old as life itself, shaped part of our DNA and have been an important drive of evolution [6, 7]. New infectious diseases can emerge from cross-species transmission, but might also hide in isolated persons and spread more easily when their contact networks change. Infectiousness is related to viral load [8], which often is related to virulence [9]. Evolution at the population level might select for more or less virulent strains through the process of inheritable diversity and selection thereon, by selecting the strains that are most successful to spread under certain conditions. Dependent on the contact network of the host the pathogen needs a certain amount of time to transmit in order to survive, resulting in a trade-off between virulence and transmission [10].

Global HIV pandemicIn 1981 the first healthy young gay men in the United States were diagnosed with Immuno Deficiency Syndrome, shortly after it was found to be Acquired (AIDS), the Netherlands fol-lowed in 1982 [11-14]. In 1983 infection with a new retrovirus was found to be responsible [15-17], later named Human Immunodeficiency Virus (HIV). Soon it became clear that HIV is transmitted not only via unprotected sexual contact among men having sex with men (MSM) but also via heterosexual intercourse, contaminated blood, and from mother to child. UNAIDS estimated that in 2008 globally 33 million people are infected with HIV, and yet another over-all 25 million people have died of HIV-related causes [18].

Based on molecular phylogenetic clustering HIV is classified in two variants HIV-1 and HIV-2, both distinctly related to simian immunodeficiency viruses (SIV) that infect other primates [19, 20]. HIV-1 is subsequently classified in three groups M (main), O (out-group), and N (new), presumed to represent separate cross-species introductions [20]. The main group M is again classified in several subtypes and recombinants thereof [21]. This diversification is likely to have originated early in the epidemic as several subtypes and recombinants are present in the Repub-

Impact of antIretrovIral therapy on hIv-1 transmIssIon dynamIcs

10

lic of Congo, regarded as the epicenter of HIV-1 group M [22]. By retrospective searching the earliest man found to be infected with HIV-1 is from a sample taken in 1959 in Kinshasa [22]. In North America subtype B is believed to have been introduced via Haiti in the seventies, and is till now the dominant circulating strain among MSM also in Western Europe [22-25].

Viral host dynamics and evolutionThe unique strategy of HIV is that it infects CD4+ receptor T helper cells which themselves are involved in antigen recognition and triggering of immune response in defense against infec-tious agents [26]. Upon infection, HIV uses and thereby destroys CD4 cells for replication of new HIV, leading to CD4 cell depletion causing immune deficiency. By error-prone replica-tion and high replication rates HIV forms a viral quasispecies that continuously changes upon selection as it escapes the host immune system. A dynamic process takes place between HIV and the immune system cells where, without treatment, some people die within months while others can live many years without disease [27]. This is influenced by both genetic variations in co-receptors, cytotoxic T cell (CTL) and human leukocyte antigen (HLA) polymorphisms and age amongst humans, and by inheritable and evolving properties of the transmitted virus [28-32]. These factors are also involved in determining if initial infection takes place. People homozygous for co-receptor CCR5 delta32 are resistant for HIV-1 infection, unless infected directly into the blood with HIV that attaches to co-receptor CXCR4 [33]. Humans also have innate immune defense mechanisms, one of which is the protein APOBEC3. It would cause HIV to hypermutate beyond its error-threshold [34] and thereby unable to maintain sufficient genetic information, if not that HIV evolved a protection by the gene vif [35].

Antiretroviral TherapyMonotherapy with AZT, was first available in 1987 [36, 37]. AZT inhibits the reverse tran-scription of the wildtype HIV RNA genome into viral DNA prior to integration into the host genome. But HIV resistance to AZT emerges within a few weeks by selection of strains with only 1- or 2 specific point mutations already present in the cloud of mutants formed upon infection [38-40]. These resistant strains were found to be capable of being transmitted [37]. A special viral mutant at amino acid position 215 of reverse transcriptase (RT) selected for dur-ing AZT mono therapy is two RNA mutations away from wildtype virus and is less fit than wildtype in absence of AZT. After transmission to a person not on antiretroviral therapy this resistant virus reverts to an AZT sensitive variant which is only a single RNA mutation removed from drug-resistant HIV, and which was found to be just as fit as wildtype in absence of therapy [41]. Here antiretroviral therapy helped HIV to pass low fitness gaps after which it can explore new routes in its fitness landscape [42, 43].

From 1996 in most developed countries combination antiretroviral therapy (cART) became available [40], consisting of three types of drugs attacking two processes of viral replication

I IntroductIon

11

simultaneously: nucleoside RT inhibitors (like AZT) that are chain terminators incorporated into the growing DNA; non-nucleoside RT inhibitors that bind to RT and stop it from func-tioning; and protease inhibitors that prevent cleavage of the gag precursor protein, which results in the production of non-infectious particles. cART proved effective, as now several mutations are needed for HIV to become resistant, and these combinations are not a priori there to be selected upon. cART can reduce virus load below the limit of ordinary detection methods of about 50 copies per ml blood. Death rates decreased drastically and cART turned HIV infec-tion from a terminal into a chronic disease [44]. cART also drastically reduced infectiousness as this is shown to be related to the viral load [8].

Once integrated HIV can remain in latent stable viral reservoirs. It is estimated that it might take a life-time of successful treatment to completely eradicate this reservoir that is built up during the first weeks of infection [45-48]. Under full suppression the virus is not able to evolve, but as the therapies do not have similar half-life times and do not all get to every body part in sufficient concentration, low-level replication might continue, and resistant mutations to the combination of therapies might accumulate and recombine. Especially when a patient has poor adherence to the therapy, and when a patient was yet infected with a partial resistant strain or was on mono-therapy previously. The worst case scenario is that a multi-drug-resistant virus evolves that in addition is more virulent and infectious then wildtype both on and of anti-retroviral therapy. At present multidrug resistant strains still result in a decreased viral load both with and without therapy and those strains are less efficiently transmitted compared to wild-type in a drug-free environment [49, 50].

Most antiretroviral therapies cause modest to severe toxic side-effects [51], in addition the human genome itself encodes for a related aspartic protease as HIV [6]. An effective vaccine that protects against HIV infection has not been realized and possibly by activating the immune response a recent trial rather did more harm then good [52]. Two new classes of antiretroviral medications for HIV treatment, fusion and integrase inhibitors, have recently been approved for use in patients in whom previous HIV treatment regimens have failed [53]. To the fusion inhibi-tor T20 emergence of drug-resistant HIV-1 variants have already emerged, including a curious case reported on a drug dependent strain [54]. Novel targets could make use of the intrinsic antiviral defence mechanisms against viral infection [55]. A few cases have been reported of people who were in need for bone marrow transplantation, because of leukemia, and by using stem cells with co-receptors resistant against HIV acquisition (homozygous for co-receptor CCR5 delta32) therewith also lost their HIV infection [56]. But this operation did not work for all patients, it is very risky and very expensive.


12

Behaviour – The NetherlandsIn the Netherlands AIDS was initially predominantly found among gay men and injecting drug users. At present, of the 10749 patients actively followed in one of the 24 national HIV treatment centers only 3% was infected by injecting drug use, 57% by homosexual, and 32% by heterosexual contact. Of those infected heterosexually 55% is from Sub-Sahara African origin [51]. A wide variety of HIV-1 subtypes (only 2% subtype B) is found among these immigrants reflecting the epidemic in Sub-Saharan Africa. HIV-1 subtype B is found amongst 96% of seropositive men having sex with men (MSM). This separation in subtypes indicates that there are different networks without substantial intermingling between risk-groups. Comprehensive programs including clean needle provision, and a shift in the nineties to non-injecting drug use, limited HIV transmission by injecting drug use [57]. Early in the epidemic MSM decreased their risk behaviour obviously in reaction to the increasing number of acquaintances dying of AIDS [58]. In response to beneficial effects of cART, increased risk behaviour among both posi-tive and negative MSM were observed in several countries, including the Netherlands [59-64]. This was confirmed in an increase of sexually transmitted infections and indications for increas-ing HIV incidence [59, 64, 65]. The increase in HIV diagnoses was initially interpreted as the effect of improved HIV testing stimulated by the availability of effective treatment. Diagnostic antibody tests were available since 1984, but have a window period of 3 months for HIV infec-tion to be identified. During these first months however HIV viral load and thus infectiousness is highest [8, 66-68]. In response to the knowledge of once HIV positive status MSM were found to reduce their risk-behaviour [69, 70], but when on successful cART a tendency of increased risk-behaviour was noticed, even though the last undetectable viral load test does not always reflect the actual viral load [60].New approaches to control HIV spread are pre-exposure prophylaxis of high risk HIV- negative MSM [71], and testing for acute primary infection, but also ordinary health care interventions such as earlier treatment and more frequent testing [72].

Impact of cArt on transmission dynamics?Mathematical modelling showed that widespread treatment with cART decreasing viral load and infectiousness has the potential to decrease the annual number of HIV infections, but that an increase in risk behaviour has the potential to counterbalance this beneficial effect [37, 73-80]. However these models did not take into account that the majority of people treated with cART successfully suppress their viral load, and that only for a small group their virus becomes resistant to all available therapies.

I IntroductIon

13

thesis outline

The aim of this thesis is to study the impact of cART on the transmission dynamics of HIV in the national well monitored population of MSM infected with HIV-1 subtype B. We used data from the prospective Amsterdam Cohort Studies and the nation wide ATHENA observational cohort. These cohorts allowed for in-depth analysis of the HIV epidemic in the Netherlands in the cART era. For that, we used a combination of molecular epidemiology methods and developed a mathematical model on HIV transmission and cART use. The main objectives were to study not only the proportion, but also the magnitude and the source of new infections with (and without) a resistant strain, and to quantify the impact of cART in intervention on transmission at the population level.

Amsterdam cohort studiesThe ‘prospective Amsterdam Cohort Studies (ACS) on HIV infection and AIDS’ initiated in 1984, followed HIV negative MSM and drug users over time. Participants are seen every 3-6 months to fill in questionnaires on risk behaviour and symptoms, and have blood samples taken for virological and immunological testing. For the participants who seroconverted dur-ing follow-up the distribution in time from HIV-1 infection to AIDS and death without the interference of cART was estimated. The ACS are also part of CASCADE, a collaboration between the investigators of 22 cohorts following persons with a well-estimated dates of HIV seroconversion.

AtHeNATo monitor HIV and the effect of cART the ATHENA observational cohort was initiated in 1996. The ATHENA cohort includes all patients treated and monitored longitudinally in one of the 24 treatment centers. Patients who died before 1996 are not included, neither are those who object (only 1.8%). Before ATHENA only the annual AIDS diagnoses were monitored by Statistics Netherlands. ATHENA collects anonymous information on patients’ first posi-tive HIV test, and when available the last negative test, CD4 cell count, HIV viral load, and initiation, content, success, failure or toxicity of antiviral therapy [51]. Since resistant strains are transmittable, as a standard of care in most hospitals patients have their HIV polymerase gene sequenced, coding for both protease and RT, to check for the presence of resistance associated mutations to antiretroviral drugs before choosing the therapy combination, and this is often also done when a person fails therapy [51, 81-84].

molecular epidemiology (chapters 2-4 and 7)Variation of the HIV polymerase gene is sufficient for reconstruction of HIV transmission net-works [85]. As HIV polymerase gene sequences are obtained for many patients and collected in ATHENA, this gave us the unique opportunity to use these sequences to study the percentage


14

transmission of resistance over calendar time using only sequences from persons with an identi-fied year of infection (Chapters 2 and 7), and to study the transmission networks these strains are in (Chapter 7). We studied the evolution of resistant strains after being transmitted to a person not on antiviral therapy (Chapter 3), and checked if transmission of resistant strains con-tributes to the failing of therapy in people who were initially successfully treated (Chapter 4). The sequences were also used to study the consistency of the MSM transmission network and in an attempt to estimate the time between infection and onward transmission (Chapter 7).

mathematical model (chapters 5 and 6)A mathematical model was developed as a tool to estimate the number of annual new infections and the changes in risk behaviour and time from infection to diagnosis needed to explain and so to fit the timing and magnitude of the longitudinal data on HIV diagnosis and first AIDS diag-nosis, incorporating information on survival distribution, cART initiation and failure. Here-with we could calculate the reproduction number, which summarizes the state of the epidemic in the number a newly infected MSM will on average infect over his life time: when larger then one the epidemic will increase, when smaller then one the epidemic will contract. It was also possible to estimate the number of undiagnosed infections and their impact on transmission, and to perform hypothetical scenarios on how many cases cART has actually prevented since it was used. The mathematical model was also applied to study the impact of: cART use and - efficacy, changes in risk-behaviour, and changes in time from infection to diagnosis.

reference List

1. Richman DD. Human Immunodeficiency Virus, International Medical Press 1-901769-09-7; 2003.

2. Woolhouse M, Gaunt E. Ecological origins of novel human pathogens. Crit Rev Microbiol 2007;

33(4):231-242.

3. Lewis Carol. Through the Looking Glass; 1865.

4. Dieckmann U, Metz JAJ, Sabelis MW, Sigmund K. Adaptive Dynamics of Infectious Diseases: In Pursuit of

Virulence Management: University Press Cambridge, UK, 532 pp.; 2002.

5. Anderson RM, May RM. Infectious Diseases of Humans.: Oxford: Oxford University Press; .0 19 854599-1;

1991.

6. Llorens C, Futami R, Bezemer D, Moya A. The Gypsy Database (GyDB) of mobile genetic elements. Nucleic

Acids Res 2008; 36(database issue):D38-D46.

7. Gifford RJ, Katzourakis A, Tristem M, Pybus OG, Winters M, Shafer RW. A transitional endogenous lenti-

virus from the genome of a basal primate and implications for lentivirus evolution. Proc Natl Acad Sci U S A

2008; 105(51):20362-20367.

8. Hollingsworth TD, Anderson RM, Fraser C. HIV-1 Transmission, by Stage of Infection. J Infect Dis 2008.

I IntroductIon

15

9. Geskus RB, Prins M, Hubert JB, Miedema F, Berkhout B, Rouzioux C, et al. The HIV RNA setpoint theory

revisited. Retrovirology 2007; 4:65.

10. van Ballegooijen WM, Boerlijst MC. Emergent trade-offs and selection for outbreak frequency in spatial

epidemics. Proc Natl Acad Sci U S A 2004; 101(52):18246-18250.

11. Gottlieb GJ, Ragaz A, Vogel JV, Friedman-Kien A, Rywlin AM, Weiner EA, et al. A preliminary communi-

cation on extensively disseminated Kaposi’s sarcoma in young homosexual men. Am J Dermatopathol 1981;

3(2):111-114.

12. Prummel MF, ten Berge RJ, Barrowclough H, Cejka V. Kaposi’s sarcoma and fatal opportunistic infections in

a homosexual man with immunodeficiency. Ned Tijdschr Geneeskd 1983; 127(19):820-824.

13. Postma MJ, Jager JC, Dijkgraaf MG, Borleffs JC, Tolley K, Leidl RM. AIDS scenarios for The Netherlands;

the economic impact on hospitals. Health Policy 1995; 31(2):127-150.

14. Reiss P, Razenberg PP, de Geus JP, Schellekens PT. A patient with a mild form of the acquired immunodefi-

ciency syndrome (AIDS). Ned Tijdschr Geneeskd 1983; 127(19):824-825.

15. Barre-Sinoussi F, Chermann JC, Rey F, Nugeyre MT, Chamaret S, Gruest J, et al. Isolation of a T-lympho-

tropic retrovirus from a patient at risk for acquired immune deficiency syndrome (AIDS). Science 1983;

220(4599):868-871.

16. Gallo RC, Sarin PS, Gelmann EP, Robert-Guroff M, Richardson E, Kalyanaraman VS, et al. Isola-

tion of human T-cell leukemia virus in acquired immune deficiency syndrome (AIDS). Science 1983;

220(4599):865-867.

17. Gallo RC, Salahuddin SZ, Popovic M, Shearer GM, Kaplan M, Haynes BF, et al. Frequent detection and

isolation of cytopathic retroviruses (HTLV-III) from patients with AIDS and at risk for AIDS. Science 1984;

224(4648):500-503.

18. UNAIDS. 2008 Report on the global AIDS epidemic. http://www.unaids.org/en/. In: 2008.

19. Hahn BH, Shaw GM, De Cock KM, Sharp PM. AIDS as a zoonosis: scientific and public health implica-

tions. Science 2000; 287(5453):607-614.

20. Sharp PM, Bailes E, Chaudhuri RR, Rodenburg CM, Santiago MO, Hahn BH. The origins of acquired

immune deficiency syndrome viruses: where and when? Philos Trans R Soc Lond B Biol Sci 2001;

356(1410):867-876.

21. Robertson DL, Anderson JP, Bradac JA, Carr JK, Foley B, Funkhouser RK, et al. HIV-1 nomenclature

proposal. Science 2000; 288(5463):55-56.

22. Worobey M, Gemmel M, Teuwen DE, Haselkorn T, Kunstman K, Bunce M, et al. Direct evidence of exten-

sive diversity of HIV-1 in Kinshasa by 1960. Nature 2008; 455(7213):661-664.

23. Lukashov VV, Goudsmit J. Recent evolutionary history of human immunodeficiency virus type 1 subtype

B: reconstruction of epidemic onset based on sequence distances to the common ancestor. J Mol Evol 2002;

54(5):680-691.

24. Hue S, Pillay D, Clewley JP, Pybus OG. Genetic analysis reveals the complex structure of HIV-1 transmission

within defined risk groups. Proc Natl Acad Sci U S A 2005; 102(12):4425-4429.


16

25. Lukashov VV, Kuiken CL, Vlahov D, Coutinho RA, Goudsmit J. Evidence for HIV type 1 strains of U.S.

intravenous drug users as founders of AIDS epidemic among intravenous drug users in northern Europe.

AIDS Res Hum Retroviruses 1996; 12(12):1179-1183.

26. Hazenberg MD, Hamann D, Schuitemaker H, Miedema F. T cell depletion in HIV-1 infection: how CD4+

T cells go out of stock. Nat Immunol 2000; 1(4):285-289.

27. Lipsitch M, Nowak MA. The Evolution of Virulence in Sexually-Transmitted Hiv Aids. Journal of Theoretical

Biology 1995; 174(4):427-440.

28. Piacentini L, Biasin M, Fenizia C, Clerici M. Genetic correlates of protection against HIV infection: the ally

within. J Intern Med 2009; 265(1):110-124.

29. Fraser C, Hollingsworth TD, Chapman R, de WF, Hanage WP. From the Cover: Variation in HIV-1 set-

point viral load: Epidemiological analysis and an evolutionary hypothesis. Proc Natl Acad Sci U S A 2007;

104(44):17441-17446.

30. van Opijnen T, Berkhout B. The host environment drives HIV-1 fitness. Reviews in Medical Virology 2005;

15(4):219-233.

31. Gali Y, Berkhout B, Vanham G, Bakker M, Back NK, Arien KK. Survey of the temporal changes in HIV-1

replicative fitness in the Amsterdam Cohort. Virology 2007; 364(1):140-146.

32. Geskus RB, Meyer L, Hubert JB, Schuitemaker H, Berkhout B, Rouzioux C, et al. Causal pathways of the

effects of age and the CCR5-Delta 32, CCR2-641, and SDF-1 3 ‘ A alleles on AIDS development. Jaids-

Journal of Acquired Immune Deficiency Syndromes 2005; 39(3):321-326.

33. O’Brien TR, Winkler C, Dean M, Nelson JA, Carrington M, Michael NL, et al. HIV-1 infection in a man

homozygous for CCR5 delta 32. Lancet 1997; 349(9060):1219.

34. John Maynard Smith, Eörs Szathmáry. The Origins of Life: Oxford University Press; 1999.

35. Goila-Gaur R, Strebel K. HIV-1 Vif, APOBEC, and intrinsic immunity. Retrovirology 2008; 5:51.

36. de Wolf F, Lange JM, Goudsmit J, Cload P, de Gans J, Schellekens PT, et al. Effect of zidovudine on serum

human immunodeficiency virus antigen levels in symptom-free subjects. Lancet 1988; 1(8582):373-376.

37. Goudsmit J, Weverling GJ, van der Hoek L, de Ronde A, Miedema F, Coutinho RA, et al. Carrier rate of

zidovudine-resistant HIV-1: the impact of failing therapy on transmission of resistant strains. AIDS 2001;

15(17):2293-2301.

38. Coffin JM. HIV population dynamics in vivo: implications for genetic variation, pathogenesis, and therapy.

Science 1995; 267(5197):483-489.

39. Malim MH, Emerman M. HIV-1 sequence variation: drift, shift, and attenuation. Cell 2001;

104(4):469-472.

40. Boucher CA, O’Sullivan E, Mulder JW, Ramautarsing C, Kellam P, Darby G, et al. Ordered appearance of

zidovudine resistance mutations during treatment of 18 human immunodeficiency virus-positive subjects. J

Infect Dis 1992; 165(1):105-110.

41. de Ronde A, van Dooren M, van der Hoek L, Bouwhuis D, de Rooij E, van Gemen B, et al. Establishment of

new transmissible and drug-sensitive human immunodeficiency virus type 1 wild types due to transmission of

nucleoside analogue-resistant virus. Journal of Virology 2001; 75(2):595-602.

I IntroductIon

17

42. Wright S. The roles of mutation, inbreeding, crossbreeding, and selection in evolution. In Proceedings of the

Sixth International Congress on Genetics, pp 355-366 1932.

43. Huynen MA. Exploring phenotype space through neutral evolution. J Mol Evol 1996; 43(3):165-169.

44. van Sighem AI, van de Wiel MA, Ghani AC, Jambroes M, Reiss P, Gyssens IC, et al. Mortality and progres-

sion to AIDS after starting highly active antiretroviral therapy. AIDS 2003; 17(15):2227-2236.

45. Perelson AS, Essunger P, Cao Y, Vesanen M, Hurley A, Saksela K, et al. Decay characteristics of HIV-1-

infected compartments during combination therapy. Nature 1997; 387(6629):188-191.

46. Siliciano JD, Siliciano RF. A long-term latent reservoir for HIV-1: discovery and clinical implications. J Anti-

microb Chemother 2004; 54(1):6-9.

47. Finzi D, Blankson J, Siliciano JD, Margolick JB, Chadwick K, Pierson T, et al. Latent infection of CD4+

T cells provides a mechanism for lifelong persistence of HIV-1, even in patients on effective combination

therapy. Nat Med 1999; 5(5):512-517.

48. Pierson T, McArthur J, Siliciano RF. Reservoirs for HIV-1: mechanisms for viral persistence in the presence

of antiviral immune responses and antiretroviral therapy. Annu Rev Immunol 2000; 18:665-708.

49. Leigh Brown AJ, Frost SD, Mathews WC, Dawson K, Hellmann NS, Daar ES, et al. Transmission fitness of

drug-resistant human immunodeficiency virus and the prevalence of resistance in the antiretroviral-treated

population. J Infect Dis 2003; 187(4):683-686.

50. Richman DD. Antiviral drug resistance. Antiviral Res 2006; 71(2-3):117-121.

51. Gras L, van Sighem A, Smit C, Zaheri S, de Wolf F, Schuitemaker H, et al. Monitoring of human immuno-

deficiency virus (HIV) infection in the Netherlands. In. Amsterdam: Stichting HIV Monitoring; 2008.

52. Cohen J. AIDS research. Did Merck’s failed HIV vaccine cause harm? Science 2007; 318(5853):1048-1049.

53. Dau B, Holodniy M. Novel targets for antiretroviral therapy: clinical progress to date. Drugs 2009;

69(1):31-50.

54. Baldwin CE, Sanders RW, Deng Y, Jurriaans S, Lange JM, Lu M, et al. Emergence of a drug-dependent

human immunodeficiency virus type 1 variant during therapy with the T20 fusion inhibitor. J Virol 2004;

78(22):12428-12437.

55. Greene WC, Debyser Z, Ikeda Y, Freed EO, Stephens E, Yonemoto W, et al. Novel targets for HIV therapy.

Antiviral Res 2008; 80(3):251-265.

56. Hutter G, Nowak D, Mossner M, Ganepola S, Mussig A, Allers K, et al. Long-term control of HIV by CCR5

Delta32/Delta32 stem-cell transplantation. N Engl J Med 2009; 360(7):692-698.

57. Lindenburg CE, Krol A, Smit C, Buster MC, Coutinho RA, Prins M. Decline in HIV incidence and

injecting, but not in sexual risk behaviour, seen in drug users in Amsterdam: a 19-year prospective cohort

study. AIDS 2006; 20(13):1771-1775.

58. Van Griensven GJP, de Vroome EMM, Goudsmit J, Coutinho RA. Changes in Sexual-Behavior and the Fall

in Incidence of Hiv Infection Among Homosexual Men. British Medical Journal 1989; 298(6668):218-221.

59. van der Bij AK, Stolte IG, Coutinho RA, Dukers NH. Increase of sexually transmitted infections, but not

HIV, among young homosexual men in Amsterdam: are STIs still reliable markers for HIV transmission?

Sex Transm Infect 2005; 81(1):34-37.


18

60. Stolte IG, De Wit JBF, van Eeden A, Coutinho RA, Dukers NHTM. Perceived viral load, but not actual

HIV-1-RNA load, is associated with sexual risk behaviour among HIV infected homosexual men. AIDS

2004; 18(14):1943-1949.

61. Stolte IG, Dukers NH. Response to ‘High-risk sexual behaviour increases among London gay men between

1998 and 2001: what is the role of HIV optimism?’. AIDS 2003; 17(13):2011-2012.

62. Stolte IG, Coutinho RA. Risk behaviour and sexually transmitted diseases are on the rise in gay men, but

what is happening with HIV? Current Opinion in Infectious Diseases 2002; 15(1):37-41.

63. Truong HHM, Kellogg T, Klausner JD, Katz MH, Dilley J, Knapper K, et al. Increases in sexually trans-

mitted infections and sexual risk behaviour without a concurrent increase in HIV incidence among men who

have sex with men in San Francisco: a suggestion of HIV serosorting? Sexually Transmitted Infections 2006;

82(6):461-466.

64. Scheer S, Kellogg T, Klausner JD, Schwarcz S, Colfax G, Bernstein K, et al. HIV is hyperendemic among

men who have sex with men in San Francisco: 10-year trends in HIV incidence, HIV prevalence, sexually

transmitted infections and sexual risk behaviour. Sexually Transmitted Infections 2008; 84(6):493-498.

65. Dukers NH, Spaargaren J, Geskus RB, Beijnen J, Coutinho RA, Fennema HS. HIV incidence on the

increase among homosexual men attending an Amsterdam sexually transmitted disease clinic: using a novel

approach for detecting recent infections. AIDS 2002; 16(10):F19-F24.

66. Wawer MJ, Gray RH, Sewankambo NK, Serwadda D, Li XB, Laeyendecker O, et al. Rates of HIV-1 trans-

mission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. Journal of Infectious Diseases 2005;

191(9):1403-1409.

67. Quinn TC, Wawer MJ, Sewankambo N, Serwadda D, Li CJ, Wabwire-Mangen F, et al. Viral load and

heterosexual transmission of human immunodeficiency virus type 1. New England Journal of Medicine 2000;

342(13):921-929.

68. Fideli US, Allen SA, Musonda R, Trask S, Hahn BH, Weiss H, et al. Virologic and immunologic deter-

minants of heterosexual transmission of human immunodeficiency virus type 1 in Africa. AIDS Res Hum

Retroviruses 2001; 17(10):901-910.

69. Marks G, Crepaz N, Senterfitt JW, Janssen RS. Meta-analysis of high-risk sexual behavior in persons aware

and unaware they are infected with HIV in the United States: implications for HIV prevention programs.

J Acquir Immune Defic Syndr 2005; 39(4):446-453.

70. Van Griensven GJP, Devroome EMM, Tielman RAP, Goudsmit J, Van der Noordaa J, de Wolf F, et al.

Impact of Hiv Antibody Testing on Changes in Sexual-Behavior Among Homosexual Men in the Nether-

lands. American Journal of Public Health 1988; 78(12):1575-1577.

71. Desai K, Sansom SL, Ackers ML, Stewart SR, Hall HI, Hu DJ, et al. Modeling the impact of HIV chemo-

prophylaxis strategies among men who have sex with men in the United States: HIV infections prevented and

cost-effectiveness. AIDS 2008; 22(14):1829-1839.

72. Granich RM, Gilks CF, Dye C, De Cock KM, Williams BG. Universal voluntary HIV testing with imme-

diate antiretroviral therapy as a strategy for elimination of HIV transmission: a mathematical model. Lancet

2009; 373(9657):48-57.

I IntroductIon

19

73. Velasco-Hernandez JX, Gershengorn HB, Blower SM. Could widespread use of combination antiretroviral

therapy eradicate HIV epidemics? Lancet Infectious Diseases 2002; 2(8):487-493.

74. Blower SM, Gershengorn HB, Grant RM. A tale of two futures: HIV and antiretroviral therapy in San Fran-

cisco. Science 2000; 287(5453):650-654.

75. Nagelkerke NJD, Jha P, de Vlas SJ, Korenromp EL, Moses S, Blanchard JF, et al. Modelling HIV/AIDS

epidemics in Botswana and India: impact of interventions to prevent transmission. Bulletin of the World

Health Organization 2002; 80(2):89-96.

76. Baggaley RF, Ferguson NM, Garnett GP. The epidemiological impact of antiretroviral use predicted by math-

ematical models: a review. Emerg Themes Epidemiol 2005; 2:9.

77. Law MG, Prestage G, Grulich A, Van de Ven P, Kippax S. Modelling HIV incidence in gay men: increased

treatment, unsafe sex and sexually transmissible infections. AIDS 2002; 16(3):499-501.

78. Xiridou M, Geskus R, de Wit J, Coutinho R, Kretzschmar M. The contribution of steady and casual

partnerships to the incidence of HIV infection among homosexual men in Amsterdam. AIDS 2003;

17(7):1029-1038.

79. Clements MS, Prestage G, Grulich A, Van de Ven P, Kippax S, Law MG. Modeling trends in HIV incidence

among homosexual men in Australia 1995-2006. J Acquir Immune Defic Syndr 2004; 35(4):401-406.

80. Salomon JA, Hogan DR. Evaluating the impact of antiretroviral therapy on HIV transmission. AIDS 2008;

22 suppl 1:S149-S159.

81. Johnson VA, Brun-Vezinet F, Clotet B, Gunthard HF, Kuritzkes DR, Pillay D, et al. Update of the Drug

Resistance Mutations in HIV-1. Top HIV Med 2008; 16(5):138-145.

82. Pozniak A, Gazzard B, Anderson J, Babiker A, Churchill D, Collins S, et al. British HIV Association

(BHIVA) guidelines for the treatment of HIV-infected adults with antiretroviral therapy. HIV Med 2003;

4 suppl 1:1-41.

83. Oette M, Kaiser R, Daumer M, Petch R, Fatkenheuer G, Carls H, et al. Primary HIV drug resistance and

efficacy of first-line antiretroviral therapy guided by resistance testing. J Acquir Immune Defic Syndr 2006;

41(5):573-581.

84. Shafer RW, Rhee SY, Pillay D, Miller V, Sandstrom P, Schapiro JM, et al. HIV-1 protease and reverse tran-

scriptase mutations for drug resistance surveillance. AIDS 2007; 21(2):215-223.

85. Hue S, Clewley JP, Cane PA, Pillay D. HIV-1 pol gene variation is sufficient for reconstruction of transmis-

sions in the era of antiretroviral therapy. AIDS 2004; 18(5):719-728.

21

Declining trend in transmission of drug-resistant HIV-1in Amsterdam

Daniela Bezemerab, Suzanne Jurriaansc, Maria Prinsa,

Lia van der Hoekc, Jan M. Prinsd, Frank de Wolfe, Ben Berkhoutc,

Roel Coutinhoa,c and Nicole K.T. Backc

Objective: Symptomatic primary HIV infections are over-represented in the mainlyhospital-based studies on transmission of resistant HIV-1. We examined a moregeneral population for the prevalence of resistant HIV-1 strains among primaryinfections.

Design: From 1994 to 2002 primary infections were identified within the AmsterdamCohort Studies (ACS) among homosexual men and drug users, and at the AcademicMedical Center (AMC). Whereas primary HIV-1-infected AMC patients, often pre-sented with symptoms of acute retroviral syndrome, ACS participants largely serocon-verted during follow-up and thus brought also asymptomatic primary infections to ourstudy.

Methods: Reverse transcriptase (RT) and protease sequences were obtained by popu-lation-based nucleotide sequence analysis of the first HIV RNA-positive sample avail-able. Subtypes were identified by phylogenetic analysis. Mutations were identifiedbased on the IAS–USA resistance table.

Results: A total of 100 primary HIV-1 infections were identified (32 AMC and 68ACS). Transmission of drug-resistant strains decreased over calendar time, with 20%[95% confidence interval (CI), 10–34%] of infections bearing drug-resistant mutationsbefore 1998 versus only 6% (95% CI, 1–17%) after 1998. No multi-drug resistancepattern was observed. The median plasma HIV-1 RNA level of the first RNA positivesample was significantly lower for the individuals infected with a resistant strain versusthose infected with wild-type, suggesting a fitness-cost to resistance. Four of sevennon-B subtypes corresponded with the prevalent subtype in the presumed country ofinfection, and none showed resistance mutations.

Conclusions: The transmission of drug-resistant HIV-1 strains in Amsterdam hasdecreased over time. Monitoring should be continued as this trend might change.

& 2004 Lippincott Williams & Wilkins

AIDS 2004, 18:1571–1577

Keywords: HIV-1, transmission, resistance associated mutations, therapy, cohort

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

From the aMunicipal Health Service Amsterdam, the bPopulation Biology Section, University of Amsterdam, the cDepartment ofHuman Retrovirology, Academic Medical Center, University of Amsterdam thedDepartment of Internal Medicine, Division ofInfectious Diseases, Tropical Medicine and AIDS, Academic Medical Center and the eHIV Monitoring Foundation, Amsterdam,The Netherlands and the Department of Infectious Disease Epidemiology, Faculty of Medicine, Imperial College, London, UK.

Correspondence to Daniela Bezemer, HIV & STI Research, Municipal Health Service, Nieuwe Achtergracht 100, 1018 WTAmsterdam, The Netherlands.

Tel: þ 31 (0)20 555 5231; fax þ 31 (0)20 555 5533; e-mail: [email protected]

Received: 1 March 2004; accepted: 7 May 2004.

DOI: 10.1097/01.aids.0000131357.52457.33

ISSN 0269-9370 & 2004 Lippincott Williams & Wilkins 1571


II


22

Introduction

Reports from several countries [1–7] show that in 10–30% of primary human immunodeficiency virus type 1(HIV-1) infections, the virus bears mutations conferringdrug-resistance, indicating quite frequent transmissionof such strains. The reported mutations conferredresistance to all three classes of antiretroviral therapy(ART): nucleoside reverse transcriptase inhibitors(NRTI), non-nucleoside reverse transcriptase inhibitors(nNRTI), and protease inhibitors (PI). As resistant HIVstrains limit therapy options [8], it is important tocontinue monitoring their frequency of transmission.However, investigation of longitudinal trends mustfocus on individuals shortly after infection. Newlydiagnosed patients without a known duration of infec-tion cannot indicate the calendar year of transmission.In the literature, primary HIV-1 infections are mostlyidentified among people that seek health care becausethey suffer from acute symptoms. Consequently, symp-tomatic primary HIV infections are over-represented inthese mainly hospital-based studies. Mutations confer-ring resistance tend to cause a lower viral fitness in theabsence of therapy, resulting in lower viral load [9,10]and possibly less severe primary HIV infection. Thushospital-based studies may be biased against HIV-1variants with resistance mutations, and it is importantto also study primary infections among persons notsuffering from acute symptoms.

The ‘prospective Amsterdam Cohort Studies (ACS) onHIV infection and AIDS’ enables study of HIVtransmission among participants who seroconvert dur-ing follow-up. Any selection bias with respect to asymptomatic acute HIV infection is therefore elimi-nated. Here we report the prevalence of resistant HIV-1 strains among primary infections within the ACS ofmen having sex with men (MSM) and drug users (DU)in the period 1994–2002. We extended our studypopulation with primary infections identified at theAcademic Medical Center (AMC) in Amsterdam.

Material and methods

Study populationIndividuals infected with HIV-1 between January 1994and January 2003 were identified within the ACS andat the AMC.

The ACS started in 1984 among MSM [11] and in1985 among DU [12]. Only young individuals, aged, 30 years have entered the MSM cohort since 1995and the DU cohort since 2000. The participants returnfor follow-up every 4 to 6 months. At entry andfollow-up, questionnaires on risk behaviour are filledin, and blood samples are taken for virological and

immunological testing. All serum samples are stored at�708C. When a sample from a previously HIV-nega-tive participant is found to be HIV-1 positive byenzyme immunoassay (EIA; using HIV-1/2 EIA testsfrom Abbott Laboratories, Abbott Park, IL, USA, andfrom bioMerieux, Boxtel, The Netherlands) and con-firmed by a positive Western blot (Genelabs Diagnos-tics, Singapore), the last antibody-negative sample istested for the presence of HIV-1 RNA. For our studyprimary infections were defined by a seroconversioninterval smaller than 18 months. For seroconverterswhose interval was larger than 18 months, a lesssensitive enzyme immunoassay (LS-EIA) [13] was usedto identify whether the person was recently infected.The LS-EIA was also used to identify primary HIV-1infections among ACS participants who entered theACS unaware of their HIV-positive status. The opticaldensity cut-off value of this assay was set at 0.5 to selectprimary infections at the individual level with highspecificity (B. Parekh, pers. comm.).

At the AMC primary HIV-1 infections were identifiedamong people attending the HIV clinic. Diagnosis ofprimary HIV-1 infection was defined as detectableHIV-1 RNA load and/or detectable serum p24 antigenin plasma combined with one of the following testresults: (1) HIV-1 specific antibody negative; (2) HIV-1 specific antibody-positive with a negative, incompleteor indeterminate western blot; (3) HIV-1 specific anti-body positive and positive western blot, but with anegative HIV-1 ELISA documented in the preceding180 days.

The date of infection was defined in the followingorder of importance: (1) the date of the last seronega-tive but RNA-positive sample; (2) the date of anindeterminate result on the western blot; (3) themidpoint between the last seronegative and the firstseropositive sample; (4) 100 days before testing negativeby the LS-EIA. Report of clinical symptoms wasobtained prospectively within the ACS and retrospec-tively for the AMC patients.

Sequence analysis of drug resistance andsubtypesPopulation-based nucleotide sequence analysis of theHIV-1 polymerase (pol) gene was performed using thefirst HIV-1 RNA-positive sample available. The 59 halfof the reverse transcriptase (RT) gene was sequencedfor all samples; and for those collected since 1996,when the protease inhibitors became commonly avail-able, codons 1–99 of the protease (PR) gene weresequenced as well. From 1996 till 1999, a home-brewassay was used, described earlier by de Jong et al. [14].From 1999 until mid 2002 a second home-brew assaywas used as described by Boom et al. [15]: viral RNAwas isolated from either 200 �l or 1 ml plasma. Theplasma was prepared from blood samples collected in


AIDS 2004, Vol 18 No 111572


II declInIng trend In transmIssIon of drug-resIstant hIv-1 In amsterdam

23

evacuated tubes containing ethylenediaminetetraaceticacid. The choice of volume was based on the HIV-1load: of samples with high loads (. 50 000 HIV-1RNA molecules/ml), 200 �l was used as input. PurifiedRNA was reverse-transcribed using MMLV-RT (InVi-trogen, Breda, The Netherlands) with 4 ng of reversetranscription primer (39-RT-OUT: 59-TCTACTTGTCCATGCATGGCTTC-39 pos HIV-1-BRU:3974). The entire RT mixture was added to the firstPCR mixture containing 100 ng of primer 59Prot-I(59-AGGCTAATTTTTTAGGGAAGATCTGGCCTTCC-39 pos HIV-1BRU: 1624) and 100 ng of primer39 ET21 (59-AGCTGGCTACTATTTCTTTTGCTACTACAGGTGG-39 pos HIV-1-BRU: 3930). A 35cycle PCR was performed. A nested PCR of 25 cyclescontained 5 �l of the first PCR, 100 ng of primer59Prot-II (59-TCAGAGCAGACCAGAGCCAACAG-39: pos HIV-1-BRU: 1718) and 39RT20 primer(59-CTGCCAGTTCTAGCTCTGCTTC-39: pos HIV-1-BRU: 3020). Ten microlitres of the PCR productswas analysed by agarose gel electrophoresis. The PCRproducts were subsequently sequenced with internalprimers and the BigDye terminator reagent. Electro-phoresis and data collection was performed on an ABI377 instrument. Protease and RT sequences wereassembled using the AutoAssembler DNA sequenceAssembly software version 2.0 (Applied Biosystems,Foster City, California, USA). With samples collectedsince mid 2002, the Viroseq HIV-1 genotyping kitversion 2 was used (Abbott Laboratories). Electrophor-esis and data collection have performed on an ABI3730 instrument since mid-2003.

Resistance-conferring mutations were screened for atthe amino acid sites described by the InternationalAIDS Society–USA [16]. Alternative substitutions atposition 215 (T215S/C/D/E/N/I/V), which representtransitional forms between wild-type and the resis-tance-conferring mutations Y and F, were included asmajor drug-resistance mutations [9,17–19]. Subtypeswere identified by phylogentic analysis of RT andprotease sequences, using reference sequences from theLos Alamos database [20] and our own database. Onethousand bootstrap replicates were performed underthe Kimura 2-parameter model [21] using the neigh-bour-joining method [22] in MEGA [23]. For boot-strap values under 85, SimPlot [24] software was usedto determine the subtype.

Plasma HIV-1 RNA and CD4 cell countHIV-1 RNA levels in the first HIV-1 RNA-positivesample available were measured using commerciallyavailable assays (NASBA HIV-1 RNA QT and Nucli-Sens; bioMerieux and Versant HIV-1 RNA 3.0, Bayer,New York, USA) according to the manufacturer’sinstructions. Absolute CD4 cell counts were measuredwith a FACS flow cytometer in all blood samples usedfor sequencing from the AMC patients, and in the

first-available HIV-seropositive sample from the ACSparticipants.

Statistical methodsThe AMC versus the ACS participants, the infectionswith resistance mutations versus those without, and thesubtype-B versus non-B infections, were comparedwith respect to baseline characteristics using theMann–Whitney U test, the Pearson chi-square testwith a continuity correction, the Fisher’s exact test anda linear regression model. A logistic regression modelwas used to test for trends in resistance over time andto perform multivariate analysis. P, 0.05 was consid-ered statistically significant.

Results

Study populationDuring the period 1994–2002, 74 ACS participantsacquired an HIV-1 infection during follow-up, ofwhom five had a seroconversion interval longer than18 months. Using the LS-EIA, one of these fiveparticipants (20%) was identified as a primary infection.In addition, 106 participants entered the ACS unawareof their HIV-1 positive status, of whom four (4%) wereidentified as primary infections by LS-EIA. Sequencingwas successfully performed for 68 (92%) of in total 74(69 þ 1 þ 4) primary infections. For three seroconver-ters the last HIV-1 antibody-negative but RNA-posi-tive sample was used for sequencing; whereas for theothers the first HIV-1 antibody-positive sample wasused. At the AMC 32 primary infections were identi-fied, and all had the HIV-1 polymerase gene se-quenced.

Table 1 shows characteristics of the 100 (68 þ 32)participants of our study. The 68 ACS participantswere mainly of Dutch origin, and the median age atseroconversion was 32 years; 38 were MSM and 30DU, of whom eight were female. Of the 30 serocon-verters from the DU cohort, three had never injecteddrugs. In the first RNA-positive sample, the medianHIV-1 RNA load was 4.7 log10 copies/ml, and themedian CD4 cell count was 550 3 106 cells/l. For the63 seroconverters in the ACS, the median intervalbetween the last HIV antibody-negative samples andthe samples from which an RT-sequence was obtainedwas 5.7 months; this interval was 6.0 months for theprotease-sequences. The 32 AMC participants weremainly of Dutch origin, and their median age atseroconversion was 39 years; 23 were MSM, and ninewere heterosexual, of whom three were female. Sixpatients had an RNA-positive but antibody-negativesample, 24 had an incomplete western blot, and twohad an antibody-negative test less than 180 days priorto the antibody-positive test. At the first RNA-positive


Transmission of drug-resistant HIV-1 Bezemer et al. 1573



24

sample, the median HIV-1 RNA load was 5.5 log10copies/ml, which was significantly higher than in theACS cases (P , 0.01). This most likely reflects theacute phase of infection of the AMC participants,which is characterized by a high viral load. The medianCD4 cell count was 520 3 106 cells/l, which did notdiffer significantly from the ACS participants. Fever hasbeen described as a symptom of primary infection thatpredicts HIV-1 disease progression [25]. Of the MSMparticipating in the ACS, 45% (17 of 38) reported feverin the 6 months prior to their first antibody-positivevisit, whereas among AMC patients, 82% (23 of 32)reported a recent fever episode. DU were left out ofthis analysis as drugs can both enhance and diminishsymptoms.

Genotypic analysis of drug resistance andsubtypesSequences of the RT and protease genes were obtainedfor, respectively, 100 and 83 participants. In 2002, aMSM and a male DU were found to be infected withstrains having identical pol genes. As later samples ofboth patients still showed a very high similarity,erroneous mixing of samples can be ruled out. TheDU was known to have sexual contacts with men.

Strains that were non-B for both RT and proteasegenes were found in only one MSM (subtypeCRF01_AE) from the ACS and five of the AMCpatients. The latter were Dutch heterosexuals of whom

four reported a non-Dutch partner from, respectively,Uganda (subtype A), Thailand (subtype CRF01_AE),Russia (subtype A), and Italy (subtype C). In the firstthree cases, the subtype found corresponds with theprevalent subtype in the respective country. Theindividual that did not report a non-Dutch partner wasinfected with subtype CRF02_AG. In addition, thestrain in one infection from the AMC was a recombi-nant having a subtype B RT sequence and subtypeCRF02_AG protease. This heterosexual patient’s originand country of infection appeared to be Ghana, wheresubtype CRF02_AG is the prevalent subtype andrecombinants are common.

Resistance to any ART was found in 13 of 100primary infections [13%; 95% confidence interval (CI),7–21%]. These included three of 32 patients from theAMC (9%; 95% CI, 2–25%), and eight of 63 serocon-verters from the ACS (13%; 95% CI, 6–24%); and twoof five patients identified by LS-EIA (40%; 95% CI, 5–85%). No significant difference was found in theprevalence of resistant infections between the ACS andthe AMC. No resistance mutations were detected inthe six non-B subtypes or the recombinant subtype.The respective resistance-conferring mutations areshown in Table 2. There were two infections withstrains bearing mutations associated with resistance tonNRTIs (2%; 95% CI, 0–7%), one with a strainresistant to PIs (1%; 95% CI, 0–5%), and ten with astrain resistant to NRTIs (10%; 95% CI, 5–18%).


Table 1. Characteristics of the study population with primary infections at the Amsterdam Cohort Studies(ACS) and the Academic Medical Center (AMC), 1994–2002.

ACS AMC

n 68 * 32Median age at estimated time of infection (years) 32 (IQR, 29–39) 39 (IQR, 32–45)Female 8 (12%) 3 (9%)Risk-groupMSM 38 (56%) 23 (72%)IDU 27 (40%)HS 3 (4%) 9 (28%)

Country of originThe Netherlands 50 (74%) 26 (81%)Other European 12 (18%) 3 (9%)Non European 6 (9%) 2 (6%)Unknown 1 (3%)

Median interval between the last antibody negative 5.7 (IQR, 3.8–6.3) , 6 monthsand the first RNA positive visit, in months. (n ¼ 2)

Median interval between the last antibody negative 5.7 (IQR, 3.9–6.5) n.a.and the sequenced RT sample, in months. (n ¼ 63)

Median interval between the last antibody negative 6.0 (IQR, 4.0–6.7) n.a.and the sequenced PR sample, in months. (n ¼ 51)

Median plasma HIV-1 RNA concentration at first 4.7 (IQR, 4.1–5.2) 5.5 (IQR, 4.7–6.2)RNA positive sample in log10 copies/ml. (n ¼ 65) (n ¼ 31)

Median CD4 cell count (3 106 cells/l) 550 (IQR, 360–850) 520 (IQR, 380–800)(n ¼ 48) ** (n ¼ 26) ***

*Sixty-three seroconverters and five selected by less sensitive enzyme immunoassay. **Median time since firstHIV-1 RNA positive visit ¼ 0.48 [interquartile range,(IQR), 0–1.6] months. ***At sequenced sample. MSM,men having sex with men; IDU, injecting drug user; HS, heterosexual; RT, reverse transcriptase; n.a., notavailable.

AIDS 2004, Vol 18 No 111574



25

There were no infections with multi-drug-resistantstrains.

The proportion of resistant infections per calendar yearare shown in Figure 1. In the years 1998, 2000 and2002, none of the primary infections showed mutationsconferring resistance to any ART. Leaving out thenon-B subtypes did not substantially change our results.The proportion of primary HIV-1 infections withdrug-resistant strains decreased over calendar time.Before 1998, 20% (95% CI, 10–34%) of 50 primaryinfections involved a virus bearing drug-resistant muta-tions versus only 6% (95% CI, 1–17%) of 50 primaryinfections after 1998 (P ¼ 0.074). A linear logisticregression model with calendar time as a continuousvariable showed a significantly declining trend in theproportion of transmitted resistance (odds ratio ¼ 0.75per year; 95% CI, 0.58–0.96). After adjustment for riskgroup, gender, age, or route of recruitment the oddsratio became 0.68 per year (95% CI, 0.50–0.94). Thesefactors did also not change significantly over time.However, the declining trend was not significant forthe ACS alone (adjusted odds ratio ¼ 0.78 per year;95% CI, 0.55–1.13). Addition of a quadratic term didnot significantly improve the fit of the models.

The median plasma HIV-1 RNA level in the firstRNA positive-sample was significantly (P ¼ 0.036)lower for the individuals infected with a resistant strain[4.4 (IQR, 3–5.1) log10 copies/ml, tested for 11 of 13infections] versus the individuals infected with a non-resistant strain [5.0 (IQR, 4.3–5.5) log10 copies/ml,tested for 85 of 87 infections]. Using linear regressionthis significant relation was robust to adjustment forpotential confounders such as the interval betweenestimated date of infection and the first HIV RNA-positive sample (within the ACS), the type of viral loadassay, age, sex, or risk group. For those participants

with a (pre-treatment) follow-up sample available(median time since estimated infection-date was 7.6(IQR, 6.4–9.8) months, the viral load was comparablefor both groups in univariate and multivariate analysis.Individuals bearing resistant versus non-resistant strainsdid not significantly differ by CD4 cell count, presenceof fever at the time of acute infection, sex, risk-group,or definition of primary infection.

Discussion

In this study we assessed the prevalence of drug-resistant strains among primary HIV-1 infections in


Table 2. Thirteen individuals with primary infections having mutations conferring resis-tance in Amsterdam in the period 1994–2002.

Patient Risk group Year Gene Mutations Drug class

1 MSM 1994 RT M41L, T215D NRTI2 MSM 1994 RT K70R NRTI3 DU 1994 RT L100I NNRTI4 MSM 1994 RT M41L, T215Y NRTI5 DU 1995 RT M41L, T215Y NRTI6 DU 1995 RT D67N, K70R, K219Q NRTI7 MSM 1995 PR M46I PR8 MSM 1996 RT D67N, K70R, T215F, K219Q NRTI9 MSM 1996 RT T215S NRTI10 MSM 1997 RT A62V, T215D NRTI11 MSM 1999 RT T215S NRTI12 MSM 2001 RT V108I NNRTI13 MSM 2001 RT T69N NRTI

MSM, men having sex with men; DU, drug users; Year, calendar year of seroconversion; RT,reverse transcriptase; PR, protease; NRTI, nucleoside reverse transcriptase inhibitors; NNRTI,non-nucleoside reverse transcriptase inhibitors.

Year

PI

NNRTI

NRTI

1994n = 13

1995n = 12

1996n = 13

1997n = 12

1998n = 7

1999n = 10

2000n = 7

2001n = 10

2002n = 16

% o

f res

ista

nt tr

ansm

issi

on

35

30

25

20

15

10

5

0

Fig. 1. Transmission of drug-resistant HIV-1 per drug classas the proportion of the total primary HIV-1 infectionsidentified per year. PI, protease inhibitors; NRTI, nucleosidereverse transcriptase inhibitors; NNRTI, non-nucleoside re-verse transcriptase inhibitors.




26

Amsterdam. During the observation period, from 1994to 2002, the overall prevalence was 13%. However, thelongitudinal pattern shows considerable variation. Thefirst strain that carried a mutation (L100I) conferringresistance to nNRTIs was transmitted to an ACS druguser in 1994. Since this mutation was never seenamong HIV-infected cohort participants before the firsttrials of nNRTIs, it most probably arose during mono-therapy trials with nevirapine or loviride in the early1990s. As drug users were not involved in those trials,a two-step transmission seems likely, although a naturalpolymorphism cannot be excluded. The only transmis-sion of a virus conferring resistance to PIs was found in1995; this mutation probably arose during PI-monotherapy trials in 1994–1995, prior to the introductionof highly active antiretroviral therapy (HAART).

Adherence to ART is likely to be relatively pooramong DU, but we did not find this risk group to berelated to the proportion of resistant HIV-1 transmis-sion. An explanation might be that HAART is morefrequently used among MSM than in DU [26].HAART is widely used in Amsterdam since it becamegenerally available in 1996. For the HIV-positiveparticipants of the ACS between 1996 and 2002:NRTIs were ever used in 37% of MSM and in 39% ofDU, PIs in 51% of MSM and in 22% of DU, andnNRTIs in 33% of MSM and in 11% of DU.

We found that the overall proportion of resistant HIVtransmission in Amsterdam decreased after the intro-duction of HAART in 1996. Furthermore, the numberand type of resistant mutations per resistant straindecreased, and could result from earlier transmissionsby people on zidovudine mono-therapy [27]. Ourresults differ from reports from the USA [1–3] and theUK [4], where transmission of drug-resistant HIVincreased in recent years, and transmission of multi-drug-resistant strains were reported. This geographicalvariation might be explained by differences in studydesign, such as sample size, study population and theresistance table used. Other important differences mayrelate to social–cultural factors, such as the physiciansreadiness to prescribe therapy, access to therapy, thehistory of mono-therapy among the treated HIV-1population, adherence and risk behavior. Our studywas not designed to evaluate such factors. However,the fact that in 1990 the Dutch government nominateda few specialist hospitals to concentrate HIV care andthe prescription of ART by specialists might haveplayed a role.

The conditions for transmission of drug-resistant strainsare present in Amsterdam, and may increase over time.Of patients on HAART in the Netherlands, 37%experienced virological failure during follow-up [28].In 64% of the patients from whom a sequence wasobtained after virological failure, drug-resistant muta-

tions were detected [29]. In addition, unsafe sex isincreasing among both the HIV-uninfected and -infected MSM population in Amsterdam [30], whichmay result in a rise in transmission of resistant strains.Fortunately, people on failing therapy probably switchto another therapy that keeps HIV RNA levels low.

Although numbers are small, none of the sevenparticipants infected with a non-B subtype harboured aresistant strain. This finding is consistent with anotherstudy in the Netherlands among 41 non-B mainlyrecently diagnosed patients [29]. The HIV epidemicamong MSM and DU still seems uninfluenced by non-B subtypes. In accordance with a previous report [31],most non-B subtypes in our study were introducedthrough heterosexual transmission by people fromcountries where these subtypes are prevalent. However,it is possible that non-B subtypes do spread in networksthat remained unidentified within our study amongprimary infections.

In six cases no sample was sequenced due to therebeing a viral load that was too-low, which could be aresult of resistance mutations that reduce the viralfitness [32]. Indeed, we found a significantly lower viralload in initial samples from patients infected with aresistant strain versus infections by wild-type virus. Nodifference in viral load was found in follow-up samples,which were taken after the viral peak at the beginningof an infection. In the short term, resistance may beassociated with a loss in viral fitness, but this loss maybe countered over time by subsequent evolution [33].We did not find a significant difference in symptomaticfever between infections with resistant and non-resistant HIV strains. We found no significant differ-ence in frequency of resistant infections between theACS and the AMC, and thus can not certify ifhospital-based studies are indeed biased against resistantstrains. One of the limitations of our study is that thenumber of individuals included in our study is rela-tively small, and more research is needed to gain insightinto this hypothesis. Nevertheless, our results areunique because within the ACS it was possible to studytransmission without any bias due to symptomaticinfections.

In conclusion, we found the transmission of drug-resistant HIV-1 in Amsterdam to be decreased sincethe introduction of HAART. Yet the conditions fortheir transmission are present, and argue strongly forcontinued monitoring.

Acknowledgements

We thank all the participants and the nurses andphysicians from the ACS for their co-operation, Lucy


AIDS 2004, Vol 18 No 111576



27

Phillips for editorial review, Marja Pospiech and JokeSpaargaren for the LS-EIA, Anneke Krol for assistancein data retrieval, Margreet Bakker for sample managing,Radjin Steingrover for AMC patient information,Remko van Leeuwen for trial information, MaartenBoerlijst and the reviewer for useful comments, andRonald Geskus for statistical advice. The AmsterdamCohort Studies on HIV infection and AIDS are acollaboration between the Municipal Health Service,the Academic Medical Center and the Central Labora-tory of the Netherlands Red Cross Blood TransfusionService, Sanquin Division, Amsterdam, the Nether-lands.

Sponsorship: This research was funded by grant number7014 from AIDS Fonds Netherlands.

References

1. Little SJ, Holte S, Routy JP, Daar ES, Markowitz M, Collier AC,et al. Antiretroviral-drug resistance among patients recentlyinfected with HIV. N Engl J Med 2002; 347:385–394.

2. Grant RM, Hecht FM, Warmerdam M, Liu L, Liegler T, Petropou-los CJ, et al. Time trends in primary HIV-1 drug resistanceamong recently infected persons. JAMA 2002; 288:181–188.

3. Simon V, Vanderhoeven J, Hurley A, Ramratnam B, Louie M,Dawson K, et al. Evolving patterns of HIV-1 resistance toantiretroviral agents in newly infected individuals. AIDS 2002;16:1511–1519.

4. UK Collaborative Group on Monitoring the Transmission of HIVDrug Resistance. Analysis of prevalence of HIV-1 drug resistancein primary infections in the United Kingdom. BMJ 2001;322:1087–1088.

5. Ammaranond P, Cunningham P, Oelrichs R, Suzuki K, Harris C,Leas L, et al. No increase in protease resistance and a decreasein reverse transcriptase resistance mutations in primary HIV-1infection: 1992–2001. AIDS 2003; 17:264–267.

6. Harzic M, Pellegrin I, Deveau C, Chaix ML, Dubeaux B, GarrigueI, et al. Genotypic drug resistance during HIV-1 primaryinfection in France (1996-1999): frequency and response totreatment. AIDS 2002; 16:793–796.

7. Yerly S, Vora S, Rizzardi P, Chave JP, Vernazza PL, Flepp M, et al.Acute HIV infection: impact on the spread of HIV and transmis-sion of drug resistance. AIDS 2001; 15:2287–2292.

8. Hirsch MS, Brun-Vezinet F, Clotet B, Conway B, Kuritzkes DR,D’Aquila RT, et al. Antiretroviral drug resistance testing in adultsinfected with human immunodeficiency virus type 1: 2003recommendations of an International AIDS Society-USA Panel.Clin Infect Dis 2003; 37:113–128.

9. Kuritzkes DR. A fossil record of zidovudine resistance intransmitted isolates of HIV-1. Proc Natl Acad Sci USA 2001;98:13485–13487.

10. Brenner B, Wainberg MA, Salomon H, Rouleau D, Dascal A,Spira B, et al. Resistance to antiretroviral drugs in patients withprimary HIV-1 infection. Investigators of the Quebec PrimaryInfection Study. Int J Antimicrob Agents 2000; 16:429–434.

11. van Griensven GJ, Tielman RA, Goudsmit J, van der Noordaa J,de Wolf F, de Vroome EM, et al. Risk factors and prevalence ofHIV antibodies in homosexual men in the Netherlands. Am JEpidemiol 1987; 125:1048–1057.

12. van den Hoek JA, Coutinho RA, van Haastrecht HJ, van ZadelhoffAW, Goudsmit J. Prevalence and risk factors of HIV infectionsamong drug users and drug-using prostitutes in Amsterdam.AIDS 1988; 2:55–60.

13. Parekh BS, Pau CP, Kennedy MS, Dobbs TL, McDougal JS.Assessment of antibody assays for identifying and distinguishingrecent from long-term HIV type 1 infection. AIDS Res HumRetroviruses 2001; 17:137–146.

14. de Jong JJ, Goudsmit J, Lukashov VV, Hillebrand ME, Baan E,Huismans R, et al. Insertion of two amino acids combined withchanges in reverse transcriptase containing tyrosine-215 of HIV-1 resistant to multiple nucleoside analogs. AIDS 1999; 13:75–80.

15. Boom R, Sol CJ, Salimans MM, Jansen CL, Wertheim-van DillenPM, van der Noordaa J. Rapid and simple method for purifica-tion of nucleic acids. J Clin Microbiol 1990; 28:495–503.

16. D’Aquila RT, Schapiro JM, Brun-Vezinet F, Clotet B, Conway B,Demeter LM, et al. Drug resistance mutations in HIV-1. Top HIVMed 2003; 11:92–96.

17. Garcia-Lerma JG, Nidtha S, Blumoff K, Weinstock H, Heneine W.Increased ability for selection of zidovudine resistance in adistinct class of wild-type HIV-1 from drug-naive persons. ProcNatl Acad Sci USA 2001; 98:13907–13912.

18. de Ronde A, van Dooren M, van der Hoek L, Bouwhuis D, deRooij E, van Gemen B, et al. Establishment of new transmissibleand drug-sensitive human immunodeficiency virus type 1 wildtypes due to transmission of nucleoside analogue-resistant virus.J Virol 2001; 75:595–602.

19. Yerly S, Rakik A, De Loes SK, Hirschel B, Descamps D, Brun-Vezinet F, et al. Switch to unusual amino acids at codon 215 ofthe human immunodeficiency virus type 1 reverse transcriptasegene in seroconvertors infected with zidovudine-resistant var-iants. J Virol 1998; 72:3520–3523.

20. Kuiken C, Korber B, Shafer RW. HIV sequence databases. AIDSRev 2003; 5:52–61.

21. Kimura M. A simple method for estimating evolutionary rates ofbase substitutions through comparative studies of nucleotidesequences. J Mol Evol 1980; 16:111–120.

22. Saitou N, Nei M. The neighbor-joining method: a new methodfor reconstructing phylogenetic trees. Mol Biol Evol 1987;4:406-425.

23. Kumar S, Tamura K, Jakobsen IB, Nei M. MEGA2: molecularevolutionary genetics analysis software. Bioinformatics 2001;17:1244–1245.

24. Ray SC. SimPlot. Baltimore, MD, USA: Division of InfectiousDiseases, Johns Hopkins University School of Medicine; 1998.http://sray.med.som.jhmi.edu/RaySoft/SimPlot/.

25. Veugelers PJ, Kaldor JM, Strathdee SA, Page-Shafer KA, SchechterMT, Coutinho RA, et al. Incidence and prognostic significance ofsymptomatic primary human immunodeficiency virus type 1infection in homosexual men. J Infect Dis 1997; 176:112–117.

26. Bangsberg DR, Charlebois ED, Grant RM, Holodniy M, DeeksSG, Perry S, et al. High levels of adherence do not preventaccumulation of HIV drug resistance mutations. AIDS 2003;17:1925–1932.

27. Goudsmit J, Weverling GJ, van der Hoek L, de Ronde A,Miedema F, Coutinho RA, et al. Carrier rate of zidovudine-resistant HIV-1: the impact of failing therapy on transmission ofresistance strains. AIDS 2001; 15:2293–2301.

28. Gras L, van Sighem A, van Valkengoed I, de Wolf F. for theDutch Collaborative HIV treatment Centers. Monitoring of Hu-man Immunodeficiency Virus Type 1 (HIV-1) in the Netherlands.Amsterdam: Stichting HIV Monitoring; 2002.

29. Gras L, van Sighem A, Zaheri S, van Valkengoed I, de Wolf F. forthe Dutch Collaborative HIV treatment Centers. Monitoring ofHuman Immunodeficiency Virus Type 1 (HIV-1) in the Nether-lands. Amsterdam: Stichting HIV Monitoring; 2003.

30. Stolte IG, Coutinho RA. Risk behaviour and sexually transmitteddiseases are on the rise in gay men, but what is happening withHIV? Curr Opin Infect Dis 2002; 15:37–41.

31. Op de Coul EL, Lukashov VV, van Doornum GJ, Goudsmit J,Coutinho RA. Multiple HIV-1 subtypes present amongst hetero-sexuals in Amsterdam 1988-1996: no evidence for spread ofnon-B subtypes. AIDS 1998; 12:1253–1255.

32. Back NK, Nijhuis M, Keulen W, Boucher CA, Oude EB, vanKuilenburg AB, et al. Reduced replication of 3TC-resistant HIV-1variants in primary cells due to a processivity defect of thereverse transcriptase enzyme. EMBO J 1996; 15:4040–4049.

33. de Ronde A, van Dooren M, van Der Hoek L, Bouwhuis D, deRooij E, van Gemen B, Goudsmit J. Establishment of newtransmissable and drug-sensitive human immunodeficiency virustype 1 wild types due to transmission of nucleoside analogue-resistant virus. J Virol 2001; 75: 595–602.




29

Sequence analysis of HIV-1 from 440 therapy-naive indi-viduals included within the CASCADE study, whoseroconverted within 18 months of the last negative test,identified 65 persons infected with a strain carryingresistance-associated mutations. Population-basedsequencing was performed for 20 of these individualsduring the therapy-free follow-up period. The mediantime of follow-up was 15 months (interquartile rangefrom 10 to 23 months). Of these individuals, 12 showedsubsequent evolution at the resistance positions, whereasthe virus of 8 people was stable during this period. In thereverse transcriptase (RT) gene, the drug-resistant 215Yor 215F codons evolved to alternative codons in all sixcases, 70R reverted to the wild-type 70K in 3 of the 4individuals, 67N evolved only in 1 of 4 patients to a wild-type 67D, 215S evolved to wild-type 215T in 1 of 3patients, 219N evolved to 219K in 1 of 2 patients, and

one patient with 184V reversed to the wild-type 184M.The 181C variant evolved to the wild-type 181Y in 1 of 2individuals. These codon changes were caused by singlenucleotide mutations. No evolution was observed forother RT mutations: 41L, 69D, 69N, 190S, 210W, 215L,215C, 215E and 219Q. In the protease gene, resistancemutations 84V and 90M were stable in 2 individuals.Comparing the CD4+ T-cell count of the 12 evolvingversus the 8 stable cases revealed no statistically signif-icant difference at the date of the first sequencefollowing seroconversion. Interestingly, a lower CD4+ T-cell count was observed in the group without evolutionat the second sequence time point (P=0.043). No differ-ence in HIV-1 RNA load was observed. These results,together with the apparent pressure to mutate at theresistance-associated positions exemplify the decreasedfitness of viruses carrying 215Y/F, 70R or 184V.

Evolution of transmitted HIV-1 with drug-resistancemutations in the absence of therapy: effects onCD4+ T-cell count and HIV-1 RNA loadDaniela Bezemer1,2, Anthony de Ronde3, Maria Prins1,3, Kholoud Porter4, Robert Gifford 5, Deenan Pillay5,Bernard Masquelier 6, Hervé Fleury6, Francois Dabis7, Nicole Back3, Suzanne Jurriaans3 and Lia van derHoek3* on behalf of the CASCADE collaboration†

1Municipal Health Service, Amsterdam, the Netherlands2Population Biology Section, University of Amsterdam, the Netherlands3Department of Human Retrovirology, Academic Medical Centre, University of Amsterdam, the Netherlands4MRC Clinical Trials Unit, London UK5University College London, UK6Département de Virologie et Immunologie biologique, CHU Bordeaux, France7INSERM U 593, Bordeaux, France†See appendix

*Corresponding author: Tel: +31 20 5667510; Fax: +31 20 6916531; E-mail: [email protected]

Antiviral Therapy 11:173–178

HIV-1 strains containing mutations conferring resistanceto one or more drug classes constitute a considerableproportion of new infections [1–7]. Infection with drug-resistant HIV-1 may affect the success of antiretroviraltherapy (ART) [8], and persistence of such strains canresult in transmission. Long-term persistence of single-drug- and multidrug-resistant strains has been reportedfollowing transmission [1,3,9,10]. In a therapy-freesetting such viral strains can evolve differently fromdrug-sensitive and -resistant strains in a therapysurrounding [11]. It is therefore important to study the

evolution of resistant strains in vivo in order to recog-nize mutations that evolve during the course of infec-tion and mutations that are stable. It is known thatseveral resistance mutations confer impaired fitness onthe virus [12–14]. Mutations reverting back to wild-type or to alternative amino acid residues reflect theimpaired fitness of the corresponding HIV-1 variants,as substitutions that enhance viral fitness are positivelyselected for. We studied the evolution of drug-resis-tance-associated mutations of reverse transcriptase(RT) and protease (PR) genes after transmission in 20

Introduction

© 2006 International Medical Press 1359-6535 173

AVT-857 Bezemer.qxp 15/3/06 12:23 pm Page 173

III


30

seroconverters not receiving therapy, and identifiedresistance mutations that evolve in time and those thatremained stable during follow-up. Based on these find-ings, the patients infected with a relatively less fit viruscould be identified. We also investigated the impact ofsubsequent evolution of the resistance mutations bycomparing the CD4+ T-cell counts and HIV-1 RNAlevels of the evolving and the stable groups.

Material and methods

Study populationPatients included in the present study belong to 11cohorts of individuals with HIV-1 infection from 7countries in Western Europe and Canada. Thesecohorts participate in the CASCADE Collaboration(Concerted Action on Seroconversion to AIDS andDeath in Europe) and contains collected data on sero-converters from 1984 to 2005 [15]. Patients included inCASCADE are older than 15 years and have a reliablyestimated date of HIV-1 seroconversion, either throughlaboratory evidence of acute infection, or with a previ-ously seronegative test within 3 years of the firstseropositive test. The date of seroconversion wasdefined as the midpoint between the last seronegativeand first seropositive sample, or, the date of laboratoryevidence of acute infection. We focused only on thosepatients with a resistance test within 18 months of thelast negative test and followed all individuals who wereinfected with an HIV-1 isolate carrying resistancemutations, and who were genotyped at several timepoints while naive to therapy.

Genotypic resistance analysisSequences of RT and PR were obtained by population-based nucleotide sequence analysis of the HIV pol geneat a minimum of two time points: (i) the first HIV RNAsample available within 18 months of the last seroneg-ative test; and (ii) at a subsequent sample before inita-tion of ART. Laboratory partners in the country oforigin performed sequencing by Viroseq HIVGenotyping System v2.0 (Abbott Diagnostics, TruGeneHIV-1 Genotyping Test, Bayer) or homemade proce-dures. Participating centres were asked to providenucleotide sequence data spanning the entire PR geneand at least codons 41–236 of RT. Major drug-resis-tance mutations were identified based on the IAS-USAresistance table [16]. Alternative substitutions at posi-tion 215 (T215S/C/D/E/N/I/V/H/L), which representtransitional forms between wild-type and the resis-tance-conferring mutations Y and F, were included asmajor drug-resistance mutations [17–19]. Furthermore219N was regarded as associated with resistance tothymidine analogues (Ben Berkhout et al. unpublishedresults; and the HIV Drug Resistance Database,

Stanford University [http://hivdb.stanford.edu/cgi-bin/SurveillanceTable.cgi?class=NRTI]).

Statistical methodsUsing Kaplan–Meier methods, we estimated the cumu-lative incidence of evolution to wild-type or alternativeresidues. The midpoint between the date of the lastsequence with the initial resistance mutations and thedate of the sequence with a change was considered asthe moment of onset of evolution. Non-evolving muta-tions were censored at the last therapy-naive sequencedate. Using the Mann–Whitney U test, we comparedCD4+ T-cell counts and HIV RNA levels cross-sectionalat baseline and follow-up visits of individuals with orwithout evolutions at the drug-resistance positions.

Results

Therapy-naive seroconverters with drug-resistancemutations in the HIV-1 genomeFor 440 seroconverters enrolled in 11 participatingcohorts a genotypic resistance analysis was availableprior to ART within 18 months of the last negative test.A total of 65 strains showed major resistance-confer-ring mutations. Of 20 individuals at least one follow-up sample was sequenced before the initiation of ART(Table 1). Of these, 17 individuals had only nucleosidereverse transcriptase inhibitor (NRTI)-associated muta-tions, 1 only non-NRTI (NNRTI) mutations, 1 onlyprotease inhibitor (PI), and 1 person had a combina-tion of all 3 classes of drug-resistance mutations. All 20strains were HIV-1 subtype B.

Evolution of resistance mutationsCertain resistance mutations in RT were very stable intime, whereas others reverted to wild-type or to alter-native codons. No evolution of the PI-resistancemutations was observed: 84V (n=1), 90M (n=2).Within the RT gene the following positions evolved:67N to D (1 of 4); 70R to K (3 of 4); 181C to Y (1 of2); 184V to M (1 of 1); 215Y/F to H/D/C/L (6 of 6),215S to T (1 of 3) and 219N to K (1 of 2). In contrast,there are several mutations that did not evolve duringour study period: 41L (n=9), 69D (n=1), 69N (n=1),210W (n=3), 215L (n=2), 215C (n=1), 215E (n=1),190S (n=1). The cumulative incidence(Kaplan–Meier) of evolution of some of the mutationsin the RT gene is shown in Figure 1.

A total of 12 cases, with different combinations ofmutations, evolved at resistance sites (called ‘evolvinggroup’) and in 8 cases all resistance mutationsremained stable during the observation periodwithout ART (called ‘stable group’). We compared theHIV-1 viral load and the CD4+ T-cell counts in theevolving group of 12 individuals to that of the stable

D Bezemer et al.

© 2006 International Medical Press174


III evolutIon of transmItted hIv-1 wIth drug-resIstance mutatIons In the absence of therapy

31

Antiviral Therapy 11:2 175

Evolution of transmitted HIV-1 with drug-resistance mutations

Table 1. Evolution of transmitted resistance mutations

Months PI-resistance-since NNRTI-resistance- associated

ID seroconversion NRTI-resistance-associated mutations associated mutations mutations

Evolving groupWT M41 D67 T69 K70 V75 M184 L210 T215 K219 L100 K103 Y181 G190 I84 L90

A1 8.3 - - - - R* - - - - - - - - - - -A1 20.0 - - - - K - - - - - - - - - - -

A2 9.0 - - - - - - - - - N* - - - - - -A2 22.5 - - - - - - - - - K I/V - - - - -

A3 7.9 - L N D R* - - - L Q - - - - - -A3 12.2 - L N D K - - - L Q - - - - - -

A4 12.7 - - - - - - V* - - - - - - - - -A4 21.4 - - - - - - M - - - - - - - - -A4 74.6 - - - - - - M - - - - - - - - -

A5 6.5 - L N - - T - W Y* N - - C S V MA5 9.1 - L N - - T - W Y* N - - C S V MA5 23.2 - L N - - T - W H N - - C S V MA6 1.3 - - - - - - - - S* - - - - - - -A6 15.3 - - - - - - - - T - - - - - - -

B1 3.0 - - N* - R* - - - F* Q - - - - - -B1 8.9 - - D - K - - - F/L Q - - - - - -

B2 3.3 - L - - - - - - Y* - - - - - - -B2 4.8 - L - - - - - - Y* - - - - - - -B2 13.1 - L - - - - - - Y* - - - - - - -B2 46.7 - L - - - - - F D - - - - - - -B2 92.7 - L - - - - - F D - - - - - - -

C1 2.4 - L - - - - - - Y* - - - - - - -C1 10.0 - L - - - - - - D - - - - - - -

C2 1.8 - L - - - - - W Y* - - - - - - -C2 2.5 - L - - - - - W Y* - - - - - - -C2 6.4 - L - - - - - W D - - - - - - -C2 37.2 - L - - - - - W D - - - - - - -

D1 1.8 - - - - - - - - - - - - C* - - -D1 27.0 - - - - - - - - - - - - Y - - -

D2 3.4 - L - - - - - - Y* - - - - - - -D2 36.2 - L - - - - - - C - - - - - - -

Stable groupWT M41 D67 T69 K70 V75 M184 L210 T215 K219 L100 K103 Y181 G190 I84 L90

A7 6.5 - L - - - - - W C - - - - - - -A7 10.6 - L - - - - - W C - - - - - - -

B3 10.3 - - - - - - - - S - - R - - - -B3 13.0 - - - - - - - - S - - R - - - -

B4 3.3 - - - - - - - E S - - - - - - -B4 56.4 - - - - - - - E S - - - - - - -

C3 2.4 - - N - R - - - - Q - - - - - -C3 9.9 - - N - R - - - - Q - - - - - -

D3 6.0 - L - - - - - - E - - - - - - -D3 8.5 - L - - - - - - E - - - - - - -

D4 1.2 - L - - - - - - L - - - - - - -D4 7.7 - L - - - - - - L - - - - - - -

D5 0.5 - - - - - - - - - - - - - - - MD5 15.7 - - - - - - - - - - - - - - - M

D6 0.0 - - - N - - - - - - - - - - - -D6 15.0 - - - N - - - - - - - - - - - -

*Transmitted resistance mutations that evolved during follow-up. NRTI, nucleoside revserse transcriptase inhibitor; PI, protease inhibitor.



32

D Bezemer et al.


group. At the time of the first sequence, there was nostatistically significant difference in CD4+ T-cellcounts between both groups (CD4+ T-cell count540/mm3 for the stable group [interquartile range(IQR): 400–648] and 605 for the evolving group[IQR: 431-906]). However, at the moment of thesubsequent sequenced sample (which in the case ofevolving virus was after the time that the mutationreverted back to wild type or an alternative mutationappeared) a borderline statistically significant differ-ence between the two groups was seen (P=0.043).The median CD4+ T-cell count in the evolving caseswas 400 (IQR: 294–620), whereas the median CD4+

T-cell count in the stable-group was 290 (IQR:220–393) (Figure 2A). At both sampling moments,the HIV-1 viral load did not differ between theevolving and the stable group (Figure 2B).

To verify that the lower CD4+ T-cell counts in theevolving group at the second sampling moment wasnot caused by a larger time interval between thesequencing dates, we calculated the time interval fromseroconversion to the first and second samplingmoment. For the evolving group the median date offirst sequence was 100 days after seroconversion(IQR: 60–250 days) and for the stable group 87 days(IQR: 20–195 days; P=0.32). At the second momentthe median was 631 days since seroconversion (IQR:320–793 days) for the evolving group and 359 days(IQR: 280–467 days, P=0.22) for the stable group.Since the time interval after seroconversion andbetween sequencing dates of the evolving group waseven longer than that of the stable group, the

difference in CD4+ T-cell counts between the twogroups was not likely explained by shorter observationperiods in the evolving group.

Inclusion in a random effects model of all therapy-free CD4+ T-cell count measurements in the first yearsfollowing serocoversion also found a slower decline ofCD4+ T-cell counts in the evolving group than in thestable group.

Discussion

We found that, in the absence of drug pressure, thestability of transmitted resistance mutations variesmarkedly. For instance, in multiple individuals,including one that was not treated for 7 years (B2), the41L mutation in RT was stably maintained. Also the215D and 215S variants are very stable and observedfor >3 years during the therapy-free period. It has beenreported that these mutations (41L-215D/S in combi-nation) have a fitness comparable to wild-type viruses[17,20]. On the other hand, the mutations 70R, 184V,215Y and 215F did change. Viruses carrying thesemutations are known to have an impaired fitness[12,13], and improving fitness may thus be the drivingforce to evolve. The observed transmitted mutations inthe PR gene remained stable during the observationperiod. Although the observation period did not exceed2 years following seroconversion, this suggests thatthese mutations do not grossly influence viral fitness.

The instability of the 215Y mutation was not influ-enced by the presence of other resistance mutations inRT or PR. 215Y in combination with several other PI,NRTI and NNRTI mutations evolved to an alterna-tive 215 codon similar to 215Y combined with only41L or 41L/210W. For 67N, 70R and 181C, weobserved that some but not all viruses evolved duringfollow-up. Stability of resistance mutation variantscould be the result of a neutral effect on fitness or dueto compensatory mutations elsewhere in the RT gene,and, as a result, a good fitness [21]. All mutations thatwere stable would have needed only one-nucleotidesubstitutions to obtain a wild-type codon.

A relatively large fraction (7 of 20) of the samplesshowed the 215L/S/C/E mutation at seroconversion.This might represent transmission of a virus withthese alternative mutations, or it may represent trans-mission of a 215Y/F variant and subsequent fastevolution to the alternative mutations within the firstmonths of infection. Indeed, the median sequencingdate after seroconversion was 94 days (IQR: 70–125days) for 215Y/F and 182 days (IQR: 39–240) for the215L/S/C/E group (P=0.39). Since we observed forpatient C2 that evolution from 215Y to 215Doccurred within 7 months following seroconversion,

1.0

0.8

0.6

0.4

0.2

0.0

70R

70R

67N

67N

41L

41L

Time since first genotypic test, months

0 10 20 30 40

Evol

utio

n-fr

ee s

urvi

val

184V

184V

215Y/F

215Y/F

215S/C/L/E

215S/C/L/E

Figure 1. Kaplan–Meier of the evolution of resistance-associated mutations

The evolution-free survival of resistance associated mutations 41L, 67N, 70R,184V, 215Y/F and 215S/C/L/E in the reverse transcriptase (RT) gene is shown.


III evolutIon of transmItted hIv-1 wIth drug-resIstance mutatIons In the absence of therapy

33

it could explain some of the transmissions with thealternative 215 mutations.

Individuals infected by viruses with evolving resis-tance mutations have relatively slow CD4+ T-celldecline in the first years following seroconversioncompared to individuals infected by viruses that do

not evolve. This observation is in agreement withimpaired fitness of transmitted virus, which isrepaired by subsequent virus evolution. Despite thefact that these viruses over time return to wild-typefitness and pathogenicity albeit with alternativeamino acids in RT, the primary infection with a lowfitness virus seems to have a protective effect on CD4+

T-cell count decline. The absence of any effect on theviral load level is more difficult to explain. Weobserved that patients with an evolving virus did nothave decreased viral load. This is however notuncommon and has also been described duringtherapy, where replication of a resistant but less fitvirus does not result in a decreased viral load. Insteada stable viral load is observed combined with anincrease in CD4+ T-cell count (the so-called ‘discor-dant’ response) [22]. From a public health perspectivemonitoring transmitted resistance mutations is impor-tant as they can evolve and transmit. At the individuallevel, in most cases it remains possible to design aneffective drug regimen. The results of our study areessential for understanding the spread of resistantstrains, especially because mathematical models cannow be adapted with up to date information on thepersistence of mutations in the ART era [23,24]. Sinceevolution of resistant strains can influence viralfitness it remains very important to monitor thisevolution, and the immunological and virologicalcharacteristics.

Acknowledgements

We thank Margreet Bakker and Anneke Krol for assis-tance in data analysis, Maarten Boerlijst and BenBerkhout for critically reading the manuscript, andRonald Geskus for statistical help. The study wasfunded by grant QLK2-2000-01431 and QLRT-2001-01708 from the EU. Daniela Bezemer was supported bygrant 7014 from AIDS Fund Netherlands.

Appendix

Coordinating CentresMRC Clinical Trials Unit (Abdel Babiker, KrishnanBhaskaran, Janet Darbyshire, Kholoud Porter, ASarah Walker) and Royal Free & University CollegeMedical School Windeyer Institute (Rob Gifford andDeenan Pillay).

Collaborators contributing data to Virology studyAquitaine cohort, France (Eric Balestre, SophieCapdepont, Geneviève Chêne, Francois Dabis, HervéFleury, Bernard Masquelier, Rodolphe Thiébaut);German cohort, Germany (Osamah Hamouda,Claudia Kücherer, Gabriele Poggensee); Italian

Antiviral Therapy 11:2 177

Evolution of transmitted HIV-1 with drug-resistance mutations

1600

P=0.27

1200

1400

1000

600

800

400

200

0

Moment of sequence analysis

1st momentstable group

Last momentevolving group

1st momentevolving group

Last momentstable group

n 7 12 7 12

CD4+ T

-cel

l cou

nt/m

m3

P=0.043

Figure 2. CD4+ T-cell counts and viral load in individualsafter transmission of stable and evolving viruses

Box plots with median, quartile, and 95% confidence interval of CD4+ T-cellcounts (A) and viral load (B) of two sample moments of stable and evolvingviruses. The P-value represents the statistical significance of the differencebetween the two groups at the same moment.

7

P=0.91

6

5

4

3

2

Moment of sequence analysis

1st momentstable group

Last momentevolving group

1st momentevolving group

Last momentstable group

n 8 12 8 11

Log

vira

l loa

d, c

opie

s/m

l

P=0.87

A

B



34

Seroconversion Study, Italy (Claudia Balotta, BenedettaLongo, Giovanni Rezza, Lorenzo Deho); GEMES,Spain (Carmen Rodriguez, Vicente Soriano, AlfredoGarcía-Saiz, Julia del Amo, Jorge del Romero, MartaOrtiz, Carmen de Mendoza); Amsterdam CohortStudies among homosexual men and drug users, theNetherlands (Nicole Back, Roel Coutinho, MariaPrins, Lia van der Hoek); Copenhagen cohort,Denmark (Louise Bruun Jørgensen, Claus Nielsen,Court Pedersen); UK Register of HIV Seroconverters,United Kingdom (Abdel Babiker, Janet H Darbyshire,Noël Gill, Anne M Johnson, Andrew N Phillips,Kholoud Porter); South Alberta clinic, Canada: (MJohn Gill, Sonia Gingues).

References1. Little SJ, Holte S, Routy JP, et al. Antiretroviral-drug resis-

tance among patients recently infected with HIV. N Engl JMed 2002; 347:385–394.

2. Bezemer D, Jurriaans S, Prins M, et al. Declining trend intransmission of drug-resistant HIV-1 in Amsterdam. AIDS2004; 18:1571–1577.

3. Grant RM, Hecht FM, Warmerdam M, et al. Time trendsin primary HIV-1 drug resistance among recently infectedpersons. JAMA 2002; 288:181–188.

4. Simon V, Vanderhoeven J, Hurley A, et al. Evolvingpatterns of HIV-1 resistance to antiretroviral agents innewly infected individuals. AIDS 2002; 16:1511–1519.

5. Ammaranond P, Cunningham P, Oelrichs R, et al. Noincrease in protease resistance and a decrease in reversetranscriptase resistance mutations in primary HIV-1 infec-tion: 1992-2001. AIDS 2003; 17:264–267.

6. Yerly S, Vora S, Rizzardi P, et al. Acute HIV infection:impact on the spread of HIV and transmission of drugresistance. AIDS 2001; 15:2287–2292.

7. Harzic M, Pellegrin I, Deveau C, et al. Genotypic drugresistance during HIV-1 primary infection in France(1996–1999): frequency and response to treatment. AIDS2002; 16:793–796.

8. Hirsch MS, Brun-Vezinet F, Clotet B, et al. Antiretroviraldrug resistance testing in adults infected with humanimmunodeficiency virus type 1: 2003 recommendations ofan International AIDS Society-USA Panel. Clin Infect Dis2003; 37:113–128.

9. Brenner BG, Routy JP, Petrella M, et al. Persistence andfitness of multidrug-resistant human immunodeficiencyvirus type 1 acquired in primary infection. J Virol 2002;76:1753–1761.

10. Barbour JD, Wrin T, Grant RM, et al. Evolution ofphenotypic drug susceptibility and viral replication

capacity during long-term virologic failure of proteaseinhibitor therapy in human immunodeficiency virus-infected adults. J Virol 2002; 76:11104–11112.

11. de Ronde A, van Dooren M, de Rooij E, et al. Infection byzidovudine-resistant HIV-1 compromises the virologicalresponse to stavudine in a drug-naive patient. AIDS 2000;14:2632–2633.

12. de Ronde A, van Dooren M, van der Hoek L, et al.Establishment of new transmissible and drug-sensitivehuman immunodeficiency virus type 1 wild types due totransmission of nucleoside analogue-resistant virus. J Virol2001; 75:595–602.

13. Back NKT, Nijhuis M, Keulen W, et al. Reduced replicationof 3TC-resistant HIV-1 variants in primary cells due to aprocessivity defect of the reverse transcriptase enzyme.EMBO J 1996; 15:4040–4049.

14. Weber J, Rangel HR, Chakraborty B, et al. Role of baselinepol genotype in HIV-1 fitness evolution. J Acquir ImmuneDefic Syndr 2003; 33:448–460.

15. Changes in the uptake of antiretroviral therapy andsurvival in people with known duration of HIV infection inEurope: results from CASCADE. HIV Med 2000;1:224–231.

16. Johnson VA, Brun-Vezinet F, Clotet B, et al. Update of thedrug resistance mutations in HIV-1: 2005. Top. HIV Med2005; 13:51–57.

17. Kuritzkes DR. A fossil record of zidovudine resistance intransmitted isolates of HIV-1. Proc Natl Acad Sci U S A2001; 98:13485–13487.

18. Garcia-Lerma JG, Gerrish PJ, Wright AC, et al. Evidence ofa role for the Q151L mutation and the viral background indevelopment of multiple dideoxynucleoside-resistanthuman immunodeficiency virus type 1. J Virol; 2000,74:9339–9346.

19. Yerly S, Rakik A, Kinloch de Loes S, et al. Switch tounusual amino acids at codon 215 of the human immun-odeficiency virus type 1 reverse transcriptase gene inseroconverters infected with zidovudine-resistant variants.J Virol 1998; 72:3520–3523.

20. Goudsmit J, de Ronde A, de Rooij E, de Boer RJ. Broadspectrum of in vivo fitness of human immunodeficiencyvirus type 1 subpopulations differing at reverse transcrip-tase codons 41 and 215. J Virol 1998; 71:4479–4484.

21. de Boer RJ, Boerlijst MC. Diversity and virulence thresh-olds in AIDS. Proc Natl Acad Sci U S A 1994; 91:544–548.

22. Solomon A, Lane N, Wightman F, Gorry PR, Lewin SR.Enhanced replicative capacity and pathogenicity of HIV-1isolated from individuals infected with drug-resistant virusand declining CD4+ T-cell counts. J Acquir Immune DeficSyndr 2005; 40:140–148.

23. Blower SM, Gershengorn HB, Grant RM. A tale of twofutures: HIV and antiretroviral therapy in San Francisco.Science 2000; 287:650–654.

24. Goudsmit J, Weverling GJ, van der Hoek L, et al. Carrierrate of zidovudine-resistant HIV–1: the impact of failingtherapy on transmission of resistant strains. AIDS 2001;15:2293–2301.

D Bezemer et al.


Received 26 July 2005, accepted 16 December 2005


35

IV combination antiretroviral therapy

failure and HIV super-infection



36Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Kimura two-parameter distances and bootstrap analysis(1000 replications), ignoring the ambiguous sites. Boththe protease (PR) and reverse transcriptase (RT) regioncould be included in the analysis for 98 patients, and,either PR or RT was available for three patients.Therefore, two phylogenetic analyses were performed:one for the sequence set sharing at least RT, the other forsequences sharing at least PR. Bootstrap values higherthan 80 were considered to be significant. Sequences werescreened for resistance-conferring mutations at the aminoacid positions described by the International AIDSSociety (USA) [22], these sites were not ignored in theanalysis.

The study group included 85 men and 16 women.Transmission risk groups were MSM (n¼ 68), heterosex-ual transmission (n¼ 21), injecting drug use (IDUs, n¼ 6),and blood transfusion (n¼ 2). For four patients the route oftransmission was unknown. Median CD4 cell count atcART initiation was 200� 106 cells/ml (IQR¼ 70–310 cells/ml). Median time between the start of cARTand viral suppression was 4.1months (IQR¼ 1.5–7.1months). Initial viral suppression lasted 4.9months(IQR¼ 2.5–9.6months). Median time between the startof failure while still on cART and the first blood sampleused for HIV isolation and sequencing was 3.3months(IQR¼ 0.2–20.9months). Half of the patients (n¼ 51)were antiretroviral treatment naıve at the start of cART, ofwhom 23% (n¼ 11) presented with resistance-conferringmutations before cART. The other half of the patients(n¼ 50) had experienced antiretroviral treatmentbefore, and 72% (n¼ 36) presented with drug-resistantmutations before start of cART. Sequences obtained afterinitial cART failure showed in 83 (81%) of 101 patientsdrug-resistant mutations and in all sequences obtainedthereafter.

Phylogenetic analysis showed 85 patients being infectedwith HIV-1 subtype B viruses. For 101 patients, thesequences obtained before start and during failure of

cART clustered together with bootstrap values above 90.Two pairs of patients had sequences that clusteredtogether with bootstrap values of 99 but, within thoseclusters, the sequence clusters from the respective patientsdid not intermix.

Table 1 shows median pairwise nucleotide sequencedistances, which were significantly smaller at intra- than atinter-patient level (P< 0.001). The highest absolutedistances between the last sample taken before cARTandthe first sample taken during failure become smaller aftercorrection for time between sequences, whereas thedistance corresponding to the shortest time interval(0.4 years) became highest when extrapolated to a yearlyrate. Positive selection between the last sample takenbefore cART and the first sample taken during failureat PR (nonsynonymous mutations per nonsynonymoussite/synonymous mutations per synonymous site > 1)was found in 14 patients, 12 at PR and two at RT.Those intra-patient sequence pairs with a nucleotidedistance > 4.5% in PR (seven patients) or > 3.0% in RT(15 patients) revealed no signs of recombination with adifferent strain at the amino acid level because thedistance was only due to single (several ambiguous)substitutions, many at known resistance conferring sites.

In conclusion, in this selected subgroup of patients whoexperienced virological failure while still on initiallysuccessful cART, no evidence for super-infection withresistant HIV-1 was observed. Transmission risk beha-viour around cART failure was reported in this smallstudy group. Three IDUs reported risk behaviourbetween the cART start date and the date of cARTfailure: injecting drugs in two, one including needlesharing, and unprotected sex with a steady HIV-1 positivepartner in the third. Four MSM reported unprotectedanal sex between the date of starting cART and thatof virological failure. When HIV is transmitted from adonor in a tight and limited transmission network(e.g. the originally infecting or infected partner),

310 AIDS 2008, Vol 22 No 2

Table 1. Median pairwise nucleotide sequence distances, according to the mixed weighted distance method as described in Gonzales et al. [12].

Median percentage pairwise nucleotidesequence distance between At synonymous sites At nonsynonymous sites Total distance

All subtype B sequences obtained from the firstsample taken during cART failure at PR

13.2 (IQR¼10.1–18.6) 4.7 (IQR¼3.5–6.0) 6.7 (IQR¼5.4–8.6)

All subtype B sequences obtained from the firstsample taken during cART failure at RT

17.4 (IQR¼14.7–20.8) 3.2 (IQR¼2.5–3.9) 6.0 (IQR¼5.2–7.1)

The last sample taken before cART and the firstsample taken during failure at PR

2.2 (range¼0–13.7) 0.9 (range¼0.0–5.7) 1.2 (range¼0.0–6.1)

The last sample taken before cART and the firstsample taken during failure at RT

3.6 (range¼0.0–11.0) 0.8 (range¼0.0–3.2) 1.4 (range¼0.0–4.7)

The last sample taken before cART and the firstsample taken during failure in rates per year at PR

0.8 (range¼0–18.4) 0.3 (range¼0.0–9.2) 0.4 (range¼0.0–11.3)

The last sample taken before cART and the firstsample taken during failure in rates per year at RT

1.3 (range¼0.0–15.3) 0.3 (range¼0.0–1.8) 0.5 (range¼0.0–4.6)

Intra-patient sequence pairs PR (n¼1094) 2.1 (range¼0.0–13.7) 1.1 (range¼0.0–6.6) 1.5 (range¼0.0–6.6)Intra-patient sequence pairs RT (n¼1040) 2.9 (range¼0.0–11.3) 0.8 (range¼0.0–3.4) 1.3 (range¼0.0–4.7)

cART, Combination antiretroviral therapy; IQR, interquartile range; PR, protease; RT, reverse transcriptase.


Iv combInatIon antIretrovIral therapy faIlure and hIv super-InfectIon


detecting super-infection is almost impossible. Differenttreatment regimens for sero-concordant couples might beprotective, but substantial cross-resistance between drugsshould be considered [22].

aHIV Monitoring Foundation, Academic MedicalCentre, University of Amsterdam, the Netherlands;bDepartment of Infectious Disease Epidemiology,Imperial College London, UK; cLaboratory of Experi-mental Virology, Department of Medical Microbiology,Center for Infection and Immunity Amsterdam(CINIMA), Academic Medical Centre, University ofAmsterdam, Amsterdam, the Netherlands; dDepart-ment of Internal Medicine, Academic Medical Center,University of Amsterdam, the Netherlands; eDepart-ment of Infectious Diseases, Health Service Amsterdam,the Netherlands; and fCenter for Infectious DiseaseControl, National Institute of Public Health and theEnvironment, Bilthoven, the Netherlands.

Sponsorship: DB was supported by grant 7014 fromAIDS Fund Netherlands.

Received: 11 October 2006; accepted: 17 October2007.

References

1. de Ronde A, van Dooren M, van der Hoek L, Bouwhuis D, deRooij E, van Gemen B, et al. Establishment of new transmissibleand drug-sensitive human immunodeficiency virus type 1 wildtypes due to transmission of nucleoside analogue-resistantvirus. J Virol 2001; 75:595–602.

2. Lukashov VV, de Ronde A, de Jong JJ, Goudsmit J. Epidemiologyof HIV-1 and emerging problems. Int J Antimicrob Agents 2000;16:463–466.

3. Vella S, Palmisano L. The global status of resistance to anti-retroviral drugs. Clin Infect Dis 2005; 41 (Suppl 4):S239–S246.

4. Bezemer D, Jurriaans S, Prins M, van der Hoek L, Prins JM, deWolf F, et al. Declining trend in transmission of drug-resistantHIV-1 in Amsterdam. AIDS 2004; 18:1571–1577.

5. van der Kuyl AC, Kozaczynska K, van den Burg R, Zorgdrager F,Back N, Jurriaans S, et al. Triple HIV-1 infection. N Engl J Med2005; 352:2557–2559.

6. Cornelissen M, Jurriaans S, Kozaczynska K, Prins JM, HamidjajaRA, Zorgdrager F, et al. Routine HIV-1 genotyping as a tool toidentify dual infections. AIDS 2007; 21:807–811.

7. Allen TM, Altfeld M. HIV-1 superinfection. J Allergy ClinImmunol 2003; 112:829–835.

8. Pernas M, Casado C, Fuentes R, Perez-Elias MJ, Lopez-GalindezC. A dual superinfection and recombination within HIV-1subtype B 12 years after primoinfection. J Acquir ImmuneDeficSyndr 2006; 42:12–18.

9. Smith DM, Wong JK, Hightower GK, Ignacio CC, Koelsch KK,Petropoulos CJ, et al. HIV drug resistance acquired throughsuperinfection. AIDS 2005; 19:1251–1256.

10. Brenner B, Routy J, Quan YD, Moisi D, Oliveira M, Turner D,et al. Persistence of mutidrug-resistant HIV-1 in primaryinfection leading to superinfection. AIDS 2004; 18:1653–1660.

11. Diaz RS, Pardini R, Catroxo M, Operskalski EA, Mosley JW,Busch MP. HIV-1 superinfection is not a common event. J ClinVirol 2005; 33:328–330.

12. Gonzales MJ, Delwart E, Rhee SY, Tsui R, Zolopa AR, Taylor J,et al. Lack of detectable human immunodeficiency virustype 1 superinfection during 1072 person-years of observation.J Infect Dis 2003; 188:397–405.

13. Fultz PN. HIV-1 superinfections: omens for vaccine efficacy?AIDS 2004; 18:115–119.

14. Truong HHM, Kellogg T, Klausner JD, Katz MH, Dilley J,Knapper K, et al. Increases in sexually transmitted infectionsand sexual risk behaviour without a concurrent increase in HIVincidence amongmenwho have sex withmen in San Francisco:a suggestion of HIV serosorting? Sex Transm Infect 2006;82:461–466.

15. Stolte IG, Dukers NHTM, Geskus RB, Coutinho RA, DeWit JBR.Homosexual men change to risky sex when perceiving lessthreat of HIV/AIDS since availability of highly active antire-troviral therapy: a longitudinal study. AIDS 2004; 18:303–309.

16. Stolte IG, Dukers NH, de Wit JB, Fennema JS, Coutinho RA.Increase in sexually transmitted infections among homosexualmen in Amsterdam in relation to HAART. Sex Transm Infect2001; 77:184–186.

17. Cox J, Beauchemin J, Allard R. HIV status of sexual partners ismore important than antiretroviral treatment related percep-tions for risk taking by HIV positive MSM in Montreal, Canada.Sex Transm Infect 2004; 80:518–523.

18. Elford J. Changing patterns of sexual behaviour in the era ofhighly active antiretroviral therapy. Curr Opin Infect Dis 2006;19:26–32.

19. Stolte IG, De Wit JBF, van Eeden A, Coutinho RA, DukersNHTM. Perceived viral load, but not actual HIV-1-RNA load,is associated with sexual risk behaviour among HIV infectedhomosexual men. AIDS 2004; 18:1943–1949.

20. van der Bij AK, Kolader ME, de Vries HJ, Prins M, Coutinho RA,Dukers NH. Condom use rather than serosorting explainsdifferences in HIV incidence among men who have sex withmen. J Acquir Immune Defic Syndr 2007; 45:574–580.

21. van Sighem AI, van de Wiel MA, Ghani AC, Jambroes M, ReissP, Gyssens IC, et al. Mortality and progression to AIDS afterstarting highly active antiretroviral therapy. AIDS 2003;17:2227–2236.

22. Johnson VA, Brun-Vezinet F, Clotet B, Kuritzkes DR, Pillay D,Schapiro JM, et al. Update of the drug resistance mutations inHIV-1: Fall 2006. Top HIV Med 2006; 14:125–130.

Prevalence and impact of HIV-1 protease mutationL76V on lopinavir resistance

Carmen de Mendoza, Carolina Garrido, AngelicaCorral, Natalia Zahonero and Vincent Soriano

Besides I47A, mutation L76V at the HIV proteasegene has recently been proposed to cause lopinavirresistance. This change was present in 37 (2.7%)out of 1376 patients failing protease inhibitorcontaining regimens. Although 26 (70%) wereon lopinavir, most had previously failed otherprotease inhibitors and carried multiple proteaseinhibitor resistance mutations. Therefore, L76Vdoes not appear to be a primary lopinavir resistancechange when the drug is used in combinationtherapy.

Virological failure in drug-naıve HIV-1-infectedpatients treated with lopinavir/ritonavir-based regimenshas rarely been associated with selection of resistancemutations at the protease gene, and mainly in individualswith low drug compliance [1]. Selection of the proteasechange I47A has been the most frequent mutationfound in this situation [1,2], although other substi-tutions, including V32I, have also been associated withhigh-level lopinavir resistance in prior drug-naıveindividuals [1,2]. Mutation I47A may cause a morethan 100-fold loss of susceptibility to lopinavir [2,3].

Research Letters 311



A resurgent HIV-1 epidemic among men who have sexwith men in the era of potent antiretroviral therapy

Daniela Bezemera, Frank de Wolfa,b, Maarten C. Boerlijstc,

Ard van Sighema, T. Deirdre Hollingsworthb, Maria Prinsd,e,

Ronald B. Geskusd,f, Luuk Grasa, Roel A. Coutinhog,h

and Christophe Fraserb

Objective: Reducing viral load, highly active antiretroviral therapy has the potential tolimit onwards transmission of HIV-1 and thus help contain epidemic spread. However,increases in risk behaviour and resurgent epidemics have been widely reported post-highly active antiretroviral therapy. The aim of this study was to quantify the impact thathighly active antiretroviral therapy had on the epidemic.

Design: We focus on the HIV-1 epidemic among men who have sex with men in theNetherlands, which has been well documented over the past 20 years within severallong-standing national surveillance programs.

Methods: Weused amathematical model including highly active antiretroviral therapyuse and estimated the changes in risk behaviour and diagnosis rate needed to explainannual data on HIV and AIDS diagnoses.

Results: We show that the reproduction number R(t), a measure of the state of theepidemic, declined early on from initial values above two and was maintained belowone from 1985 to 2000. Since 1996, when highly active antiretroviral therapy becamewidely used, the risk behaviour rate has increased 66%, resulting in an increase of R(t) to1.04 in the latest period 2000–2004 (95% confidence interval 0.98–1.09) near or justabove the threshold for a self-sustaining epidemic. Hypothetical scenario analysisshows that the epidemiological benefits of highly active antiretroviral therapy andearlier diagnosis on incidence have been entirely offset by increases in the riskbehaviour rate.

Conclusion: We provide the first detailed quantitative analysis of the HIV epidemic in awell defined population and find a resurgent epidemic in the era of highly activeantiretroviral therapy, most likely predominantly caused by increasing sexual riskbehaviour. � 2008 Wolters Kluwer Health | Lippincott Williams & Wilkins

AIDS 2008, 22:1071–1077

Keywords: antiretroviral therapy, homosexual men, infectious diseases,mathematical models, models/projections, sexual behaviour, surveillance

From the aHIV Monitoring Foundation, Amsterdam, The Netherlands, the bDepartment of Infectious Disease Epidemiology,Imperial College London, UK, the cInstitute for Biodiversity and Ecosystem Dynamics, University of Amsterdam, the dHealthService Amsterdam, the eCenter for Infection and Immunity Amsterdam (CINIMA), Academic Medical Center, University ofAmsterdam, the fDepartment of Clinical Epidemiology Biostatistics and Bioinformatics, Academic Medical Center, thegDepartment of Human Retrovirology, AcademicMedical Center, University of Amsterdam, and the hCenter for Infectious DiseaseControl, National Institute of Public Health and the Environment, Bilthoven, The Netherlands.

Correspondence to Christophe Fraser, Department of Infectious Disease Epidemiology, Faculty of Medicine at St Mary’s Campus,Imperial College London, London, W2 1PG. UK.

E-mail: [email protected]: 9 May 2007; revised: 7 February 2008; accepted: 15 February 2008.

ISSN 0269-9370 Q 2008 Wolters Kluwer Health | Lippincott Williams & Wilkins 1071

V


A resurgent HIV-1 epidemic among men who have sexwith men in the era of potent antiretroviral therapy

Daniela Bezemera, Frank de Wolfa,b, Maarten C. Boerlijstc,

Ard van Sighema, T. Deirdre Hollingsworthb, Maria Prinsd,e,

Ronald B. Geskusd,f, Luuk Grasa, Roel A. Coutinhog,h

and Christophe Fraserb

Objective: Reducing viral load, highly active antiretroviral therapy has the potential tolimit onwards transmission of HIV-1 and thus help contain epidemic spread. However,increases in risk behaviour and resurgent epidemics have been widely reported post-highly active antiretroviral therapy. The aim of this study was to quantify the impact thathighly active antiretroviral therapy had on the epidemic.

Design: We focus on the HIV-1 epidemic among men who have sex with men in theNetherlands, which has been well documented over the past 20 years within severallong-standing national surveillance programs.

Methods: Weused amathematical model including highly active antiretroviral therapyuse and estimated the changes in risk behaviour and diagnosis rate needed to explainannual data on HIV and AIDS diagnoses.

Results: We show that the reproduction number R(t), a measure of the state of theepidemic, declined early on from initial values above two and was maintained belowone from 1985 to 2000. Since 1996, when highly active antiretroviral therapy becamewidely used, the risk behaviour rate has increased 66%, resulting in an increase of R(t) to1.04 in the latest period 2000–2004 (95% confidence interval 0.98–1.09) near or justabove the threshold for a self-sustaining epidemic. Hypothetical scenario analysisshows that the epidemiological benefits of highly active antiretroviral therapy andearlier diagnosis on incidence have been entirely offset by increases in the riskbehaviour rate.

Conclusion: We provide the first detailed quantitative analysis of the HIV epidemic in awell defined population and find a resurgent epidemic in the era of highly activeantiretroviral therapy, most likely predominantly caused by increasing sexual riskbehaviour. � 2008 Wolters Kluwer Health | Lippincott Williams & Wilkins

AIDS 2008, 22:1071–1077

Keywords: antiretroviral therapy, homosexual men, infectious diseases,mathematical models, models/projections, sexual behaviour, surveillance

From the aHIV Monitoring Foundation, Amsterdam, The Netherlands, the bDepartment of Infectious Disease Epidemiology,Imperial College London, UK, the cInstitute for Biodiversity and Ecosystem Dynamics, University of Amsterdam, the dHealthService Amsterdam, the eCenter for Infection and Immunity Amsterdam (CINIMA), Academic Medical Center, University ofAmsterdam, the fDepartment of Clinical Epidemiology Biostatistics and Bioinformatics, Academic Medical Center, thegDepartment of Human Retrovirology, AcademicMedical Center, University of Amsterdam, and the hCenter for Infectious DiseaseControl, National Institute of Public Health and the Environment, Bilthoven, The Netherlands.

Correspondence to Christophe Fraser, Department of Infectious Disease Epidemiology, Faculty of Medicine at St Mary’s Campus,Imperial College London, London, W2 1PG. UK.

E-mail: [email protected]: 9 May 2007; revised: 7 February 2008; accepted: 15 February 2008.

ISSN 0269-9370 Q 2008 Wolters Kluwer Health | Lippincott Williams & Wilkins 1071




Introduction

To determine the success of a decade of widespreadintervention with highly active antiretroviral therapy(HAART) on controlling the human immunodeficiencyvirus type1 (HIV-1)epidemic,weanalysedthe transmissiondynamics of HIV-1 over the past 25 years among menhaving sex with men (MSM) in the Netherlands.

The first AIDS cases in the Netherlands among MSMwere identified in 1981 [1]. HAART, consisting of acombination of drugs from three, and later four, differentdrug classes, became widely available from 1996 onwards.HAART dramatically reduced plasma and seminal viralload [2,3], resistance [4] and mortality rates [5,6]. Asinfectivity is shown to be strongly correlated to viral load[7,8], the widespread use of HAART might thus beexpected to have reduced the incidence of HIVinfections. Paradoxically, resurgent epidemics have beenwidely reported post-HAART [9,10]. Increases in riskbehaviour [11–13] and syphilis and gonorrhoea diagnoseshave also been documented in populations of MSM inseveral developed countries [13–15]. Earlier mathema-tical modelling studies have demonstrated that an increasein risk behaviour has the potential to counterbalance thebeneficial effect of HAART [16–23].

In the present study, we aim to evaluate the separateimpact of risk behaviour, HIV testing behaviour andHAARTon the HIVepidemic in Dutch MSM by meansof a mathematical model fitted to data recorded withinseveral national databases, which provided extensiveinformation on epidemic trends.

Methods

ModelA mathematical model describing HIV transmission andHAART use among MSM in the Netherlands wasconstructed. The model we used described natural(untreated) disease progression, diagnosis and subsequentuse of HAART. The basic structure of the model isillustrated in Fig. 1. The modelling strategy was tailored tothe task of analyzing annual HIVand AIDS diagnosis timeseries, and specifically to tracking changes in per capitatransmission rates. The most important factor in thisrespect is a simultaneous estimation of the prevalence ofinfectious individuals, weighted by their relative infec-tiousness, which depends on stage of infection andtreatment status, and the incidence of new infections.Mathematical details and analyses of the model, includingsensitivity analyses, hypothetical scenarios and predictions,and further data are available on request from the authors.

Survival distributionA method to increase realism in compartmental models isto include a unidirectional flow through several compart-

ments, corresponding to an Erlang survival distributionBy fitting such a distribution to data from 130 MSMseroconverters before the HAARTera in the AmsterdamCohort Studies [24], the maximum likelihood estimatecorresponded to five compartments with mean stay ineach of 1.89 years. Patients starting their diseaseprogression first spend on average 0.24 year in an extrainitial compartment that represents primary infection,and we equated the last stage of infection with AIDS, anapproximation that seemed reasonable given the matchwith the estimated duration of high transmissibility that

1072 AIDS 2008, Vol 22 No 9

Fig. 1. Model structure. (a) Flow diagram of model of HIV-1transmission among MSM. The model describes progressionthrough different stages of natural history and treatment.Arrows depict the different flow rates between compartments.New infections start with primary HIV infection and thenprogress to death through stages of infection. Undiagnosedinfections get diagnosed, after which risk behaviour can bereduced. The diagnosis rate varies over time. From 1996,antiretroviral treatment is available that can be long-term ortransiently successful. Disease progression is represented bystacks: white stacks represent stages that are infectious and inwhich disease progression occurs. The nature of diseaseprogression within a stack is shown in detail in (b). Greystacks represent stages that are not infectious and in whichdisease progression does not occur as viral replication issuppressed by treatment. Infectiousness is highest duringprimary infection and AIDS, and lower during stages 1–4.All stages, weighted by their relative infectiousness and fittedby the risk behaviour rate parameter, contribute to the esti-mated annual new infections such that the annual data onHIVand AIDS diagnosis can be described. Imported infectionsflow into primary infection and undiagnosed compartments.(b) To enhance realism on survival distribution in the model,disease progression is represented by a unidirectional flowthrough five compartments with mean stay in each of 1.89years.


v resurgent hIv-1 epIdemIc among men havIng sex wIth men


has been observed prior to death [25]. The relativeinfectiousness of each stage is calculated fromWawer et al.[25], with primary and last stages of infection beingmore infectious than the other four stages. As 97% of theMSM population is infected with subtype B, importedinfections are all assumed to be in the primary infectionstage entering the Netherlands after short holidays, butthis assumption was not critical (unpublished analysisavailable on request).

Highly active antiretroviral therapyIn the HAART era, patients start HAART after beingdiagnosed, and they will be either on successful treatment(no detectable viral load) or experience therapy failure,with a viral rebound and infectiousness [26]. Duringsuccessful treatment, HAART is assumed to block bothHIV transmission and disease progression [2–8,27,28],and viral blips are not taken into account [29]. Patientsexperiencing treatment failure are assumed to haveperiods of apparent successful treatment before failure[26]. After failure, patients go through the unidirectionalstages of natural disease progression. We assume that thereare three HAART treatment opportunities beforepatients fail completely and progress to death (represent-ing the diversity of treatments available) [26]. TheHAART era started in 1995 with clinical trials andcompassionate use, and the mass treatment programmestarted in 1996 and was fully implemented by 1998. Theinfluence of pre-HAART therapy on HIV viral load andtransmission before 1995 is neglected. People startHAARTwith a rate irrespective of their stage of infection(as multistage disease progression is included in themodel, this assumption approximately reproduces theobserved pattern of HAART initiation), but at the AIDSstage people are set on HAART immediately. Diseaseprogression is unidirectional. Parameters on HAARTuseand failure were obtained from the ATHENA nationalobservational cohort [26].

Transmission rate and risk behaviourThe standardized per infectious capita transmission rateb(t) is a time-varying function that measures the relativerate at which an HIV-positive infectious individual in-fects new individuals. It is standardized by setting itequal to 1.0 for untreated, undiagnosed individuals in theasymptomatic stage of infection during the first phase ofthe epidemic (1980–1983), so that all other values aremeasured relative to this. It is primarily intended as ameasure of changes in risk behaviour that can beestimated in our study, and for convenience b(t) will bereferred to as risk behaviour rate. b(t) is in fact acompound measure that is affected by changes in thepartner change rate, by the rate and nature of risky sexacts within partnerships, by the effect of ’saturation’ ofthe susceptible population (when new sexual partners arealready previously infected) and by the effect of thechanging prevalence of other sexually transmittedinfections (STIs) in modulating HIV transmission.

Parameters in the model explicitly adjust for the effect ofHAART in reducing infectiousness, for the increasedinfectiousness during primary and late (AIDS) stages ofdisease and for the effect of diagnosis in reducing riskbehaviour. We assumed that MSM have a 50% reductionin risk behaviour after becoming aware of theirseropositive status and implemented this into ourmodel [30]. These assumptions were all encoded asdisease-stage-specific scaling parameters of risk beha-viour rate b(t).

Reproduction numberWe define the reproduction number R(t) as the averagenumber of people an infected person at time t wouldinfect over his whole infectious lifespan if conditionsremained the same as at time t [31,32]. It incorporates allfactors including risk behaviour, effect of diagnosis andthe effects of treatment with HAART in preventinginfection. If the within-country R(t) is greater than 1,then the epidemic will grow exponentially driven by localtransmission, and conversely if this number is less than 1,the epidemic will contract down to a number pro-portional to the number of imported cases [31]. It is a keyaim of public health interventions to avoid a locally drivenepidemic and maintain R(t) below one. The state ofthe epidemic can be characterized by R(t) that can becalculated from the best fit parameters in the model.

Model fitWe fitted our model simultaneously to the observed timeseries of annual new diagnoses [32] and annual new AIDScases (see below) [14,32,33], which are constrained by thediagnosis rate and the risk behaviour rate b(t), and thismade it feasible to estimate both these unknownparameters. Changing these independent parametershas different effects, which differently affect the goodnessof fit of the model to the time series. Increasing riskbehaviour increases both the number of diagnoses andAIDS cases, whereas increasing the diagnosis rateincreases the number of diagnoses in the short termbut leads to sustained long-term reductions in the numberof diagnoses and AIDS cases.

The analysis was stratified into four distinct historicalintervals: 1980–1983, the first AIDS cases were diagnosed[1]; 1984–1995, serological testing became available,increasing HIV awareness, introduction of first mono-antiretroviral and dual-antiretroviral therapies [3,6]; 1996–1999, earlyHAARTera; and2000–2004, currentHAARTera. The diagnosis rate during asymptomatic stages wasestimated but was assumed to be zero during the firstperiod (1980–1983). Diagnosis was assumed to be rapid(within 1 month) after AIDS, whereas zero duringprimary infection. The mean time to diagnosis, definedunder conditions at time t, was calculated from the esti-mated diagnosis rate. The epidemic is assumed to havestarted with an import of cases in 1980. The model wassolved numerically using Runge-Kutta 4 algorithm and

Resurgent HIV epidemic among men having sex with men Bezemer et al. 1073




was fitted to data by a custom maximum likelihoodmethod. All analyses were performed in BerkeleyMadonna, version 8.0.1 (http://www.berkeleymadonna.com).

Annual new AIDS casesTo account for the effect of HAART in preventingprogression to AIDS, we used different data sets tosimultaneously fit to the following: before 1997, themodel is fitted to annual data on AIDS diagnosis amongMSM and collected by Statistics Netherlands from thebeginning of the HIV epidemic [14,34]; from 1996, themodel is fitted to annual data of number of MSM gettingHIV diagnosed while having AIDS in the ATHENAnational observational cohort [33].

Annual new diagnosesFrom1984 themodelwas fitted to data on annual diagnosesper year among MSM in the ATHENA national obser-vational cohort [33]. Since 1998 all HIV patients in theNetherlands have been registered and monitored as part ofthe ATHENA national observational cohort. The year offirst HIV diagnosis is recorded retrospectively at the pointof registration into ATHENA. Patients who receivedHAARTand died in the period 1996–1997were includedin the ATHENA database retrospectively. Although thereis some uncertainty on the completeness of the retrospec-tive inclusion, it is expected to have only minor bearing on

our results.MSMwhodied before 1996 are not included inthe ATHENA database. We explicitly accounted for thisdata truncation process in our model by implementedchance of surviving until 1996 for the respective stages ofinfection. In parallel, a prediction is made of the true (nottruncated) curve of the number of new diagnoses (Fig. 2a).Data from 2005 are still incomplete, and they are thus notincluded in the current study.

Source of infectionBy the start of 2005, 5516MSM diagnosed with HIV hadbeen included in the ATHENA observational cohort. Ofthese registered infections, 8% were reported to haveacquired the infection while abroad and 62% from apartner within the Netherlands. Of those born in theNetherlands, 4% were infected abroad and of those bornabroad, 41% were infected abroad. We assume that theremaining 30% of infections with an unknown country ofinfection are split according to these respective ratiosrespective to their country of birth. Thus, we estimatethat overall 14% of diagnosed infections are imported.

For model verification, we compared the model numberof prevalent cases with number of living HIV-positiveMSM in the ATHENA database. Also, data on thepredicted annual number of deaths with documentedannual AIDS deaths in the Netherlands [14] were usedfor model outcome verification. These data contain

1074 AIDS 2008, Vol 22 No 9

500

(a)

(b)

(c)

(d)

400

300

200

100

400

300

200

100

0

0.5

0

250

500

750

1

1.5

2

2.5

3

0

1.2

1

0.8

0.6

0.4

0.2

0 0

2

4

6

8

01984 1989 1994

Year

1980 1985 1990 1995 2000 2005

Year1980 1985 1990 1995 2000 2005

Year

Year1999 2004 1980 1985 1990 1995 2000 2005

Num

ber

of n

ew d

iagn

oses

Ris

k be

havi

our

rate

β(t

)R

epro

duct

ion

num

ber

R(t

)

Num

ber

of A

IDS

cas

es

Tim

e to

dia

gnos

is (

year

s)N

ew in

fect

ion

per

year

Fig. 2. Data and model fit. (a) Number of new diagnoses of HIV. Thick lines and dots, cases acquired within the Netherlands; thinlines and triangles, cases acquired abroad. Empty symbols represent years when data are only available for patients surviving until1996, and dashed lines represent estimated actual number of diagnoses. (b) Number of new diagnoses of AIDS. Data from DutchHealth Inspectorate (black dots) used in model fit, and ATHENA (empty dots) for model verification (not fitted). (c) Estimates ofthe risk behaviour rate b(t) (solid line, left axis; 1.30, 0.56, 0.66, 0.93) and the mean time between infection and diagnosis (dashedline, right axis; 7.88, 3.71, 3.16, 2.90). (d) Estimate of the reproduction number R(t) (solid line, left axis; 2.39, 0.89, 0.76, 1.04) andof the number of new infections acquired within Netherlands (dashed line, right).




information only on sex and no more specificinformation on risk group. Hence, AIDS deaths amongMSM were predicted using the percentage of MSMamong male ATHENA participants in 1996 as anestimate, predicting that about 2000 MSM had died ofAIDS before 1996 [14] (Fig. 3a). We also compared theestimated proportion of newly diagnosed patients in eachdisease stage as estimated by our best model fit, with dataon CD4 cell count at diagnosis.

Results

Figure 2a and b shows the model curves that fitted best tothe observed time series of annual new diagnoses andAIDS diagnoses (data on AIDS at diagnosis not shown).Figure 2d shows the estimated absolute number of newinfections per year in the Netherlands. This peaked in1983 with 802 new infections, and in 2004 with 554new infections.Estimates for the risk behaviour rate b(t), the reproduc-tion number R(t) and the mean time to diagnosis areshown in Fig. 2c and d. Over the initial period (1980–1983), the estimate for the reproduction number R(t) is2.39 [95% CI (confidence interval) 2.17–2.76]. Between

1984 and 1995, the risk behaviour rate declined by 2.3-fold (95% CI 2.03–2.83), indicating large reductions inrisk behaviour, and thereby reduced the reproductionnumber R(t) below one to 0.89 (95% CI 0.85-0.93), thatis, just below the epidemic threshold.

After 1995, when HAART was introduced, thereproduction number declined yet further to 0.76(95% CI 0.7–0.86), but the reduction was not as greatas it could have been due to a 18% (95% CI 3–34%)increase of the risk behaviour rate, b(t). The riskbehaviour rate is estimated to have increased yet furtherover the period 2000–2004 and returned to only 29%(95% CI 22–72%) below its value in the initial period1980–1983. Reductions in the estimated mean time frominfection to diagnosis [from 3.71 years (95% CI 3.49–3.97) in 1984–1995 to 2.90 years (95% CI. 2.84–3.03) in2000–2004] with consecutive reductions in risk beha-viour and widespread treatment with HAARTresulted inthe reproduction number being much lower than in theinitial time period 1980-1983. Still, R(t) for the last timeperiod 2000–2004 is estimated to be 1.04 (95% CI 0.98–1.09), near or above the critical epidemic threshold, andthus indicating that HIV may once again be spreadingepidemically among MSM in the Netherlands.

From the best fit model, we estimated that 24% of allliving HIV-positive MSM were unaware of their HIV-positive status at the start of 2005 and that they account for90% of new infections. Without both the increase of therisk behaviour rate and the decrease of time to diagnosis,the reproduction number R(t) would have decreased by24% from 0.89 to 0.68 due to the introduction ofHAART. The risk behaviour rate would need to increaseby 32% to offset this benefit, with 43% in order to offsetthe simultaneous benefits of the increase in testingbehaviour and with 59% in order to get R(t) equal to one,that is, to revert to epidemic growth. An increase of 66%was measured to have occurred. On the basis of thesemodel estimates, we conclude that HAART has played animportant role in limiting transmission but that any gainsmade have been more than offset by increases in the riskbehaviour rate. Had these increases not occurred in theHAARTera, the reproduction number R(t) would havedeclined to 0.6, and the epidemic would have been inconvinced decline.

We verified our predictions subjectively for consistencywith approximated data on annual number of AIDSdeaths in MSM (see Methods), and on the number ofcurrently living diagnosed individuals in the nationalpatient database ATHENA [33], shown in Fig. 3, and onthe number of annual AIDS diagnoses after 1996(Fig. 2b). We considered the quality of fit acceptablegiven that the model was not fitted to these data. Aqualitative comparison of CD4 cell counts at diagnosiswith model predictions in terms of disease stages showssimilar trends.


300

(a)

(b)

200

100

0

6000

5000

4000

3000

2000

1000

0 0

25

50

75

100

1980 1985 1990 1995

Year2000 2005

1980 1985 1990 1995

Year2000 2005

Num

ber

of d

eath

sN

umbe

r of

pre

vale

nt c

ases

Per

cent

age

undi

agno

sed

Fig. 3. Consistency of model fit. (a) Number of deaths causedby HIV. Seventy percent of number of AIDS deaths amongmale (black dots, see methods) and model prediction of AIDSdeaths among MSM (thick line). Deaths among MSM inATHENA (empty dots). (b) Number of prevalent cases. HIVþMSM in ATHENA (black dots) and model prediction (thickline). Predicted number (thin line) and proportion (dashedline, right axis) of cases that are unaware of their infectedstatus.




In a sensitivity analysis, the results on the key outcomesb(t) and R(t) appear to be very robust to a wide range ofmodel variants. In particular, model results wereconsistent when assumptions about the relative infec-tiousness of disease stages, effect of diagnosis onbehaviour, and time from diagnosis to start of therapywere varied. In all model variants, R(t) for 2000–2004 isestimated to be near or above the critical threshold(R¼ 1), thus implying uncontrolled epidemic spread,with estimates of the current reproduction numberranging between 0.95 and 1.33, depending on thescenario (details available on request).

Discussion

The joint effect of HAART and risk behaviour on HIVincidence has been previously studied using mathematicalmodels and empirical data [16–21,35]. Although basedon different assumptions, all these studies come to thesame conclusion regarding the potential for an increasein risk behaviour to offset the benefits of HAART inreducing transmission. Our study provides new evidencethat this has actually occurred and quantifies its magnitudeand timing within a well studied population of MSM.

A key feature of our study is the existence of severalnational databases recording diagnoses of HIV infectionand AIDS, and deaths, allowing the diagnosis rate to beestimated reliably by simultaneously fitting to these timeseries within a robust inference framework. We were thusable to confirm that there has indeed been a recentincrease in the diagnosis rate, reflecting a more frequenttesting as was reported recently, but this was not suffi-cient to explain the recent increases in the number ofpeople newly diagnosed. Rather, the recent increase inthe number of new diagnoses reflects a substantial increasein transmission. Our estimates were corroborated bychanging trends in CD4 cell count at diagnosis, where arecent increase in the proportion of newly diagnosedindividuals with high CD4 cell counts is apparent.

Testing rates are low in the Netherlands when comparedwith other developed countries [36,37], and the potentialof intervention by frequent testing with the rapid test isnot yet fully explored [38]. Our model, however, suggeststhat the only way to reverse epidemic spread, and getR well below one, is to reduce the risk behaviour ratefrom current levels. The potential effects of routine use ofnew diagnostic methods that target primary HIVinfection were not explored here and should be exploredin future models [39].

The most likely factor driving changes in the riskbehaviour rate parameter b(t) is changing the sexual riskbehaviour, both within partnerships and in partnerchange rates [12], though related factors such as otherSTIs acting to enhance transmission, saturation of the

susceptible population or even evolution of infectivitycould also play a role. Our analysis made it possible tocompare the relative changes over time in risk behaviourrate between infectious and negative MSM, the ’hidden’information that cannot be measured by survey data, andour results indicate that whatever measures individuals aretaking to ‘serosort’ [40] are not proving effective at thepopulation level and have not offset epidemic spread.

The introduction of HAART was accompanied by adecrease in the percentage of resistant strains among newinfections [33,41]. However, the recent increase in annualnew infections could in turn result in an increasingabsolute number of resistant infections [42].

The widespread use of HAART has led to largereductions in AIDS morbidity and mortality (Figs. 2and 3). Sustaining these reductions into the future willrequire either further improvements in treatment efficacyor a response to limit resurgent epidemic spread.

In conclusion, there is an increase in HIV transmissionamong MSM in the Netherlands, in spite of earlierdiagnosis and subsequent effective treatment. The mosteffective intervention is to bring risk behaviour back topre-HAART levels.

Acknowledgements

D.B. was supported by grant 7014 from AIDSFund Netherlands and by a travel grant from NWO(Netherlands Organisation for Scientific Research). C.F.is funded by the Royal Society. The funders had no rolein study design, data collection and analysis, decision topublish or preparation of the manuscript.

References

1. Prummel MF, ten Berge RJ, Barrowclough H, Cejka V. Kaposi’ssarcoma and fatal opportunistic infections in a homosexualman with immunodeficiency. Ned Tijdschr Geneeskd 1983;127:820–824.

2. Gupta P, Mellors J, Kingsley L, Riddler S, Singh MK, Schreiber S,et al. High viral load in semen of human immunodeficiencyvirus type 1-infected men at all stages of disease and itsreduction by therapy with protease and nonnucleoside reversetranscriptase inhibitors. J Virol 1997; 71:6271–6275.

3. Goudsmit J, Weverling GJ, van der Hoek L, de Ronde A,Miedema F, Coutinho RA, et al. Carrier rate of zidovudine-resistant HIV-1: the impact of failing therapy on transmission ofresistant strains. AIDS 2001; 15:2293–2301.

4. Vandamme AM, Van Laethem K, De Clercq E. Managingresistance to anti-HIV drugs : an important consideration foreffective disease management. Drugs 1999; 57:337–361.

5. Hammer SM, Katzenstein DA, Hughes MD, Gundacker H,Schooley RT, Haubrich RH, et al. A trial comparing nucleosidemonotherapy with combination therapy in HIV-infected adultswith CD4 cell counts from 200 to 500 per cubic millimeter.N Engl J Med 1996; 335:1081–1090.

6. Smit C, Geskus R, Uitenbroek D, Mulder D, van den Hoek A,Coutinho RA, et al. Declining AIDS mortality in Amsterdam :contributions of declining HIV incidence and effective therapy.Epidemiology 2004; 15:536–542.

1076 AIDS 2008, Vol 22 No 9




7. Quinn TC, Wawer MJ, Sewankambo N, Serwadda D, Li CJ,Wabwire-Mangen F, et al. Viral load and heterosexual trans-mission of human immunodeficiency virus type 1.N Engl J Med2000; 342:921–929.

8. Fideli US, Allen SA, Musonda R, Trask S, Hahn BH, Weiss H,et al. Virologic and immunologic determinants of heterosexualtransmission of human immunodeficiency virus type 1 inAfrica. AIDS Res Hum Retroviruses 2001; 17:901–910.

9. White PJ, Ward H, Garnett GP. Is HIV out of control in the UK?An example of analysing patterns of HIV spreading usingincidence-to-prevalence ratios. AIDS 2006; 20:1898–1901.

10. UNAIDS. Report on the global AIDS epidemic, executive sum-maryUNAIDS/06.29E.Geneva: JointUnitedNationsProgrammeon HIV/AIDS (UNAIDS); 2006. http://www.unaids.org/en/HIV_data/2006GlobalReport/default.asp

11. Stolte IG, Dukers NHTM. Response to ’High-risk sexual beha-viour increases amongLondongaymenbetween1998and2001:what is the role of HIV optimism?’. AIDS 2003; 17:2011–2012.

12. Stolte IG, Dukers NHTM, Geskus RB, Coutinho RA, DeWit JBR.Homosexual men change to risky sex when perceiving lessthreat of HIV/AIDS since availability of highly active antire-troviral therapy: a longitudinal study. AIDS 2004; 18:303–309.

13. Truong HHM, Kellogg T, Klausner JD, Katz MH, Dilley J,Knapper K, et al. Increases in sexually transmitted infectionsand sexual risk behaviour without a concurrent increase in HIVincidence amongmenwho have sex withmen in San Francisco:a suggestion of HIV serosorting? Sex Transm Infect 2006;82:461–466.

14. van de Laar MJW, de Boer IM, Koedijk FDH, Op de Coul ELM.HIV and Sexually Transmitted Infections in the Netherlands in2004, RIVM report 441100022/2005. Bilthoven: National In-stitute for Public Health and the Environment; 2005. http://www.rivm.nl/bibliotheek/rapporten/rapporten2005.html

15. van der Bij AK, Stolte IG, Coutinho RA, Dukers NH. Increase ofsexually transmitted infections, but not HIV, among younghomosexual men in Amsterdam: are STIs still reliable markersfor HIV transmission? Sex Transm Infect 2005; 81:34–37.

16. Blower SM, Gershengorn HB, Grant RM. A tale of two futures:HIV and antiretroviral therapy in San Francisco. Science 2000;287:650–654.

17. Nagelkerke NJD, Jha P, de Vlas SJ, Korenromp EL, Moses S,Blanchard JF, et al. Modelling HIV/AIDS epidemics in Botswa-na and India: impact of interventions to prevent transmission.Bull World Health Organ 2002; 80:89–96.

18. Velasco-Hernandez JX, Gershengorn HB, Blower SM. Couldwidespread use of combination antiretroviral therapy eradi-cate HIV epidemics? Lancet Infect DisV 2 2002:487–493.

19. Clements MS, Prestage G, Grulich A, Van de Ven P, Kippax S,Law MG. Modeling trends in HIV incidence among homosex-ual men in Australia 1995–2006. J Acquir Immune Defic Syndr2004; 35:401–406.

20. Law MG, Prestage G, Grulich A, Van de Ven P, Kippax S.Modelling the effect of combination antiretroviral treatmentson HIV incidence. AIDS 2001; 15:1287–1294.

21. Xiridou M, Geskus R, deWit J, Coutinho R, Kretzschmar M. Thecontribution of steady and casual partnerships to the incidenceof HIV infection among homosexual men in Amsterdam. AIDS2003; 17:1029–1038.

22. Baggaley RF, Ferguson NM, Garnett GP. The epidemiologicalimpact of antiretroviral use predicted bymathematical models:a review. Emerg Themes Epidemiol 2005; 2:9.

23. Hosseinipour M, Cohen MS, Vernazza PL, Kashuba ADM. Canantiretroviral therapy be used to prevent sexual transmission ofhuman immunodeficiency virus type 1? Clin Infect Dis 2002;34:1391–1395.

24. van Griensven GJ, Tielman RA, Goudsmit J, Van der Noordaa J,de Wolf F, de Vroome EM, et al. Risk factors and prevalence ofHIV antibodies in homosexual men in the Netherlands. Am JEpidemiol 1987; 125:1048–1057.

25. Wawer MJ, Gray RH, Sewankambo NK, Serwadda D, Li XB,Laeyendecker O, et al. Rates of HIV-1 transmission per coitalact, by stage of HIV-1 infection, in Rakai, Uganda. J Infect Dis2005; 191:1403–1409.

26. Gras L, van Sighem A, van Valkengoed I, Zaheri S, de Wolf F.Monitoring of human immunodeficiency virus (HIV) infectionin the Netherlands, Chapter 10, HIV Monitoring Foundation.Amsterdam: Stichting HIV Monitoring; 2003. http://www.hiv-monitoring.nl

27. van Sighem AI, van de Wiel MA, Ghani AC, Jambroes M, ReissP, Gyssens IC, et al. Mortality and progression to AIDS afterstarting highly active antiretroviral therapy. AIDS 2003; 17:2227–2236.

28. Barroso PF, Schechter M, Gupta P, Bressan C, Bomfim A,Harrison LH. Adherence to antiretroviral therapy and persis-tence of HIV RNA in semen. J Acquir Immune Defic Syndr2003; 32:435–440.

29. Havlir DV, Bassett R, Levitan D, Gilbert P, Tebas P, Collier AC,et al. Prevalence and predictive value of intermittent viremiawith combination hiv therapy. JAMA 2001; 286:171–179.

30. Marks G, Crepaz N, Senterfitt JW, Janssen RS. Meta-analysis ofhigh-risk sexual behavior in persons aware and unaware theyare infectedwithHIV in theUnited States: implications for HIVprevention programs. J Acquir Immune Defic Syndr 2005;39:446–453.

31. Anderson RM, May RM. Infectious diseases of humans: dy-namics and control. Oxford, UK: Oxford University Press;1991. ISBN 0 19854040-X.

32. Fraser C. Estimating individual and household reproduc-tion numbers in an emerging epidemic. PLoS ONE 2007;2:e758.

33. Gras L, van Sighem A, Smit C, Zaheri S, de Wolf F. Monitoringof human immunodeficiency virus (HIV) infection in theNetherlands, Chapter 6, 9&10. Amsterdam: HIV MonitoringFoundation; 2006. http://www.hiv-monitoring.nl.

34. Postma MJ, Jager JC, Dijkgraaf MG, Borleffs JC, Tolley K, LeidlRM. AIDS scenarios for The Netherlands: the economic impacton hospitals. Health Policy 1995; 31:127–150.

35. Katz MH, Schwarcz SK, Kellogg TA, Klausner JD, Dilley JW,Gibson S, et al. Impact of highly active antiretroviraltreatment on HIV seroincidence among men who have sexwith men: San Francisco. Am J Public Health 2002; 92:388–394.

36. Dukers NH, Fennema HS, van der Snoek EM, Krol A, GeskusRB, Pospiech M, et al. HIV incidence and HIV testing beha-vior in men who have sex with men: using three incidencesources, The Netherlands, 1984–2005. AIDS 2007; 21:491–499.

37. Stolte IG, de Wit JB, Kolader ME, Fennema HS, Coutinho RA,Dukers NH. Low HIV-testing rates among younger high-riskhomosexual men in Amsterdam. Sex Transm Infect 2007; 83:387–391.

38. Hutson S. Is home test for HIV a good idea? New Sci 2005;188:16.

39. Pilcher CD, Fiscus SA, Nguyen TQ, Foust E,Wolf L,WilliamsD,et al.Detection of acute infections during HIV testing in NorthCarolina. N Engl J Med 2005; 352:1873–1883.

40. van der Bij AK, Kolader ME, de Vries HJ, Prins M, CoutinhoRA, Dukers NH. Condom use rather than serosortingexplains differences in HIV incidence among men who havesex with men. J Acquir Immune Defic Syndr 2007; 45:574–580.

41. Bezemer D, Jurriaans S, Prins M, van der Hoek L, Prins JM, deWolf F, et al. Declining trend in transmission of drug-resistantHIV-1 in Amsterdam. AIDS 2004; 18:1571–1577.

42. Sanchez MS, Grant RM, Porco TC, Getz WM. HIV drug-resistant strains as epidemiologic sentinels. Emerg Infect Dis2006; 12:191–197.



47

VI 27 years of the HIV epidemic

amongst men having sex with men in the Netherlands: an in depth

mathematical model-based analysis1

Daniela Bezemer1, Frank de Wolf1,2, Maarten C. Boerlijst3, Ard van Sighem1, T. Deirdre Hollingsworth4, and Christophe Fraser4

Abstract

Background: There has been increasing concern about a resurgent epidemic of HIV-1 amongst men having sex with men in the Netherlands, which has parallels with similar epidemics now occurring in many other countries. methods: A transmission model applicable to HIV-1 epidemics, including the use of antiret-roviral therapy, is presented in a set of ordinary differential equations. The model is fitted by maximum likelihood to national HIV-1 and AIDS diagnosis data from 1980-2006, estimating parameters on average changes in unsafe sex and time to diagnosis. Robustness is studied with a detailed univariate sensitivity analysis, and a range of hypothetical scenarios are explored for the past and next decade.results: With a reproduction number around the epidemic threshold one, the HIV-1 epidemic among amongst men having sex with men in the Netherlands is still not under control. Scenario analysis showed that in the absence of limiting infectiousness in treated patients, the epidemic could have been more than double its current size. Ninety percent of new HIV transmissions is estimated to take place before diagnosis. Decreasing time from infection to diagnosis, which was 2.5 years on average in 2006, can prevent many future infections.conclusions: Sexual risk-behaviour of men having sex with men who are not aware of their infection is the most likely factor driving this epidemic.

1HIV Monitoring Foundation, Amsterdam, the Netherlands; 2Department of Infectious Disease Epidemiology, and 4Medical Research Council Centre for Outbreak Analysis and Modelling, Department of Infectious Disease Epide-miology, Imperial College London, UK; 3Institute for Biodiversity and Ecosystem Dynamics, University of Amster-dam, the Netherlands;


48

Background

Despite the success of combination antiretroviral treatment (cART), the HIV-1 epidemic has been increasing amongst men having sex with men (MSM) in many industrialised countries, including the Netherlands [1-3]. Increases in syphilis and gonorrhoea diagnoses have also been documented in populations of MSM in several developed countries [4-6]. Surveys of risk behaviour have noticed increases in the number of casual partners and decreases in reliable condom use [5-10]. It is not clear what the best interventions are to get the HIV-1 epidemic under control.

Earlier mathematical modelling studies showed that an increase in risk behaviour has the potential to counterbalance the beneficial effect of cART [11-19]. Models have varied substan-tially with respect to whether transmission occurred from treated, diagnosed or undiagnosed individuals. With a recently developed mathematical model, which explicitly takes into account that the majority of people on cART are successfully treated and maintain very low HIV RNA levels, and are thus presumed largely uninfectious, we found evidence that increased risk behav-iour amongst undiagnosed individuals may have counterbalanced the beneficial effect of cART among MSM in the Netherlands [1]. Other models looking at the impact of cART treatment have found that treatment of the HIV infected population in an advanced stage of disease progression alone might not halt epidemic spread but that expanded and earlier access to cART can reduce the growth of the epidemic [19-21]. Besides pre-exposure prophylaxis of high risk HIV- negative MSM [22], and testing for acute primary infection, more routine health care interventions such as earlier treatment and more frequent testing are still not fully exploited [23]. Here, we apply our mathematical model to study the impact of cART use and – efficacy, risk-behaviour, and time from infection to diagnosis in a well documented epidemic among MSM.

First, we compare predictions based on our earlier analysis carried out with data till 2003 with an updated dataset covering the period 2004-2006, and we subsequently update our analysis. We present a detailed univariate sensitivity analysis which highlights some of the dependencies under the assumptions and structure of our model. We present data on CD4 count at diagnosis as a verification for the estimated changing patterns of time from infection to diagnosis. We complete with counterfactual hypothetical scenarios to explore the impact of changes in cART use, behaviour change and time to diagnosis. These scenario analyses were applied retrospec-tively to the past decade, and in addition to the coming decade.

Our development and application of a model framework for the interpretation of longitudinal surveillance data from the Netherlands may be adapted to similar situations in other settings (for other risk groups and other countries). In the appendix we present the mathematical details

vI 27 years of the hIv epIdemIc amongst men havIng sex wIth men In the netherlands

49

of the model, present derivations of key quantities such as the reproduction number [24] and the time to diagnosis, and of the maximum likelihood framework used to fit the model.

model and data description

The model code was formulated in a set of ordinary differential equations implemented in the program Berkeley Madonna [25]. The model describes the rate of new HIV-1 infections as a function of the number of currently infected individuals, and individuals entering the coun-try with infections acquired abroad. The framework captures the natural process of disease progression, as well as the effects of therapy. Key parameters relate to the relative changes in infectiousness: relative rate of risky sex; stage of infection; diagnosis, as diagnosed individuals might take less risks [26]; and cART. Parameters on cART use and failure were obtained from the ATHENA national observational cohort [27]. From this cohort we found that most people are long-term successfully treated on cART, with no detectable viral load. We separated this group from another smaller group that is only temporally successful before failing. This is dif-ferent from most other models which use a constant rate of failure applicable to all patients, an assumption which only adequately describes the impact of ineffective monotherapy. We assume that people with no detectable viral load are not infectious, an assumption based on a study with discordant heterosexual couples [28], that is currently being investigated for discord-ant couples with access to cART. We fitted our model simultaneously to longitudinal data on annual HIV-1 diagnoses [29] and annual new AIDS cases [29-31] among MSM in the Neth-erlands in order to estimate the average changes over calendar time in risk-behaviour b(t) and time to diagnosis needed to explain these data. From our parameter estimates the reproduction number R(t) can be calculated. The analysis was stratified into five distinct historical intervals. Parameter values are summarized in Table 1. The protocols for handling the different sources of surveillance data, and for assigning cases with uncertain status, were described in an earlier paper [1]. Model details are in the appendix at the end of this paper.


50

table 1. Descriptions and values of Model Parameters.

Parameter Value Denotation Comment

1/a P 0.24 Duration (years) of Primary Infection [28]1/a 1.89 Duration (years) of each disease stage

s ; (s = 1,..,5)Maximum likelihood

r P 2.77 Relative infectiousness during Primary Infection

[28]

r s, (s = 1,…,4) 0.11 Relative infectiousness of asympto-matic disease stage s (s = 1,..,4)

[28]

r 5 0.35 Relative infectiousness of disease stage 5 (AIDS)

[28]

s 0.5 Reduction in Relative infectiousness after diagnosis of each disease stages s

[26, 41]

b(t) Estimated in five time intervals:(1980-1983); (1984-1995); (1996-1999); (2000-2003); (2004-2006)

Net standardized transmission rate Estimated by fitting model to data

d s (t), (s = 1,…,4) If time < 1996 then 0 else estimated in four time intervals: (1984-1995); (1996-1999); (2000-2003); (2004-2006)

Rate per year at which people are diagnosed when in stage sThis version of the model does not account for diagnosis during primary stage.

Estimated by fitting model to data

d 5 (t) 1/12 Rate per year at which people are diagnosed when in stage 5 (AIDS)

Assump-tion

A(t) A0 if 1980 < t < 1984A1 if t > 1984

Imported cases Estimated by fitting model to data

f P 1 Proportion of imported cases which are in Primary Infection

Assump-tion

g s, (s = 1,…,4) If time < 1996 then 0 elseif time < 1998 then 0.7 ⋅ (time-1996)/2 else if time >1998 then 0.7

Rate per year of starting treatment and suppressing viral load at stages when cART naïve

[29]

g5 If time < 1995 then 0 elseif time < 1996 then 1 else then 2

Rate per year of starting treatment and suppressing viral load at stage 5 (AIDS) when cART naïve

[29]

t s, (s = 1,…,5) 0.5 Fraction that fails treatment at stage s [29] s

f , (s = 1,…,5, f = 1,2)

1.4 Rate per year of starting subsequent treatment and suppressing viral load at stage s, failing round f+1

[29]

s

f , (s = 1,…,5, f = 1,…,3)

0.5 Rate per year of failing treatment at stage s, treatment round f

[29]

m 0 (short study period) Basic death rate when no disease progression i.e. when enduringly or temporally successfully treated

[42]


51

results

model fitWe start by describing the best fit of our baseline model to the national surveillance data and what this reveals in terms of changing patterns of incidence, diagnosis and risk behaviour. Figures 1a and b, show the model fit to the HIV and AIDS diagnosis data. Figure 1c shows the best fit parameters, with a similar risk behaviour over the past six years, and a further decreasing time to diagnosis to 2.5 years on average in 2006. The risk behaviour rate b(t) is standardised by setting it equal to 1.0 for untreated undiagnosed individuals in the asymptomatic stage of infec-tion during the first phase of the epidemic (1980-1983), so that all other values are measured relative to this [1]. Figure 1d shows the resulting reproduction number is still around one, i.e. its epidemic threshold. The resulting prediction in the absolute numbers of new infections have been steadily increasing since 2000, as shown in figure 1e. A total of 620 MSM are estimated to be infected in 2006, which is close to the estimated 777 infections when the epidemic was at its peak around 1983. In figure 1f, the known and estimated unknown prevalence is shown. The percentage of the undiagnosed of the total infected population has decreased to 24%, but only so due to an increase in survival of the diagnosed population. In absolute numbers around 1600 MSM were undiagnosed at the end of 2006, estimated to be responsible for 90% of new HIV transmissions.

Update of analysis to include the period 2004-2007Figure 2a shows that the number of new diagnoses in the period 2004–2006 are within the prediction interval of the model fit in our previous iteration which focused on data till 2003 [1]. Figure 2b shows the prediction interval of the model fit till 2006. The number of diagnoses in 2007 are within the prediction interval, but will likely be higher as there is a delay in data avail-ability. This year was not included in the model fit as testing policy changed to an opting-out strategy at the STI clinic in Amsterdam.

sensitivity AnalysisBecause our baseline model was built on assumptions which underpinned the inferences drawn from the epidemic surveillance data, we explored the impact of varying assumptions in both the input parameters, interpretation of incomplete data, and model structure, refitting our model each time an assumption was varied. The results are shown in table 2. It shows the best model fit results and how these depend on parameters and assumptions in the model after refitting the model with implemented changes. The results on the key outcomes b(t) and R(t) appear to be very robust to a wide range of model variants. In particular, the model results were consistent till 2004 when assumptions about the relative infectiousness of disease stages, about the effect of diagnosis on behaviour, and about the time from diagnosis to start of therapy were varied. To examine whether the results could be due to delays in starting therapy over calendar time,


52

0

200

400

600

800

1000

1200

1400

1980 1983 1986 1989 1992 1995 1998 2001 2004

0

0,5

1

1,5

2

2,5

1980 1983 1986 1989 1992 1995 1998 2001 2004

0

0,2

0,4

0,6

0,8

1

1,2

1980 1983 1986 1989 1992 1995 1998 2001 20040

1

2

3

4

5

6

7

8

0

1000

2000

3000

4000

5000

6000

7000

1980 1983 1986 1989 1992 1995 1998 2001 20040

20

40

60

80

100

120

0

50

100

150

200

250

300

350

400

1980 1983 1986 1989 1992 1995 1998 2001 2004

Year

Year

Year Year

Nu

mb

er

of

an

ua

l ne

w A

IDS

ca

ses

Nu

mb

er

of

an

ua

l ne

w in

fect

ion

s MS

M d

eso

ng

aid

gnivil f

o re

bm

uN Pe

rce

nta

ge

un

dia

gn

ose

d M

SM

0

100

200

300

400

500

600

1980 1983 1986 1989 1992 1995 1998 2001 2004

siso

ng

aid

we

n la

un

a fo r

eb

mu

N Ave

rage

tim

e fro

m in

fect

ion

to d

iagn

osis

)t(R

Year

Sta

ndard

ised n

et tr

ansm

issio

n r

ate

Year

a c

b d

e f

figure 1. This whole figure is an update of previously published figures [1].a, Black lines are best model fit to filled symbols of number of annual diagnosis separated for infected in the Netherlands in dots, and infected abroad in triangles, 2007 is incomplete data not included in fit. b, Black lines are best model fit to filled symbols of annual number of new AIDS cases in blue dots, and simultaneous HIV and AIDS diagnosis in red triangles, 2007 is incomplete data not included in fit. c, Fitting parameters with confidence intervals, on left axis in blue is the risk behaviour rate, in red on the right axis is the average time from infection to diagnosis. The risk behaviour rate b(t) is standardised by setting it equal to 1.0 for untreated undiagnosed individuals in the asymptomatic stage of infection during the first phase of the epidemic (1980-1983), so that all other values are measured relative to this [1]. d, Resulting estimated Reproduction number, R(t) over the whole study period, including the confidence interval. e, Estimated annual new infections over the whole study period, including the confidence interval. f, Blue dots is the living diagnosed MSM population, and the blue line the model estimation, The red line is the model estimation for the undiagnosed HIV infected MSM population, and the dotted red line is the percentage undiagnosed of the total prevalence.


53

0

250

500

750

1000

1995 2000 2005 2010

Ann

ual

diag

nose

s

a

0

250

500

750

1000

1995 2000 2005 2010

Ann

ual

diag

nose

s

b

Year

Year

figure 2. a, Model fit to total annual diagnosis in black filled dots till 2003, including prediction interval, and data update in black circles. b, Model fit to total annual diagnosis till 2006, including prediction interval, and incomplete diagnosis in 2007.


54

we considered a model with a switch in policy after 2000 where individuals were no longer treated until the last stages of infection (model #29 in Table 2), but our conclusions remained unaffected. The assumption which has largest impact on our predictions relates to the case where the risk behaviour rate b(t) was assumed not to have increased in the cART era (model #19), but this model does not fit the data. If we assume (somewhat implausibly, and also resulting in a much worse model fit) that all infections with unknown source of infection are acquired outside of the Netherlands, then the epidemic would have been mainly driven by import and R(t) within the Netherlands would be smaller than one (model #9). Estimates of R(t) are higher (and model fit better) if all ‘unknown’ cases are assumed to be the result of transmission within the Neth-erlands (model #8). None of the other variations studied in the sensitivity analysis considerably altered the model predictions after a refit to the data. In all model variants, R(t) for 2000-2004 is estimated to be near or above the critical threshold (R(t)=1), thus implying uncontrolled epidemic spread, with estimates of the reproduction number ranging between 0.95 and 1.33, depending on the scenario under consideration. One fit is done using data from the literature on the transmission probability for receptive anal intercourse instead of relative infectiousness, resulting in b(t) giving an indication of the aver-age number of unsafe sex-acts per year between infectious- and negative MSM (model #20).

The data update on the period 2005–2006 is shown in model #42 to #59. It shows that adjust-ments to the data due to a delay in monitoring did not change our previous results. Whether R(t) is just below or above the threshold one in the last four years of study depends largely on how effective cART really was and on how the time periods were defined. However, the confi-dence interval and all sensitivity analysis show that under any set of assumptions, it is still very close to one. Under all assumptions risk-behaviour has stayed the same over the past decade or even increased a little. The average time from infection to diagnosis has decreased under all assumptions, to 2.46 in the best fit.


55

table 2. Uncertainty and sensitivity analysis. Main results after refitting the model under different parameter values then in table 1 or to alternative datasets.

What if

R(t)respectively

in: 1980-1983; 1984-1995; 1998-1999; 2000-2003

b(t)respectively

in: 1980-1983; 1984-1995; 1996-1999; 2000-2003

time to diagno-sis respectively in: 1980-1983;

1984-1995; 1996-1999; 2000-2003 Dev/2* %

of H

IV p

osit

ives

un

awar

e at

the

end

of

the

mod

el fi

t

% n

ew in

fect

ions

fr

om u

naw

are

at t

he

end

of t

he m

odel

fit

Best fit 2.39; 0.89; 0.76; 1.04 1.30; 0.56; 0.66; 0.93 7.88; 3.71; 3.16; 2.90 58.98 24.0 90.3

Lower 95% confidence limit**

2.17; 0.85; 0.70; 0.98 1.18; 0.53; 0.60; 0.87 ***; 3.49; 3.00; 2.84 60.90 **** ****

Upper 95% confidence limit**

2.76; 0.93; 0.86; 1.09 1.50; 0.58; 0.71; 0.97 ***; 3.97; 3.41; 3.03 60.90 **** ****

1 Infectiousness primary infection 50% smaller

3.37; 0.91; 0.72; 1.10 2.23; 0.71; 0.88; 1.38 7.88; 3.72; 3.16; 2.89 57.41 24.0 85.6

2 Infectiousness primary infection 75% smaller

4.70; 0.91; 0.69; 1.15 3.50; 0.83; 1.06; 1.84 7.88; 3.72; 3.16; 2.89 55.04 23.9 80.8

3 No transmission from primary phase

10.15; 0.92; 0.65; 1.26 8.63; 0.97; 1.33; 2.73 7.88; 3.75; 3.17; 2.89 49.00 23.9 71.5

4 No reduction in risk behaviour after diagnosis

2.93; 0.94; 0.71; 1.09 1.36; 0.43; 0.54; 0.84 7.88; 3.67; 3.17; 2.89 66.74 23.9 82.3

5 Ceasing risk behaviour after diagnosis

1.77; 0.81; 0.84; 0.98 1.16; 0.75; 0.85; 1.02 7.88; 3.90; 3.10; 2.92 52.02 24.1 100

6 No reduction in risk behaviour after diagnosis in cART era

2.39; 0.89; 0.72; 1.09 1.30; 0.56; 0.55; 0.84 7.88; 3.71; 3.17; 2.89 58.84 23.9 82.2

7 All diagnosed HIV infections as local

3.49; 1.0; 0.91; 1.17 1.89; 0.62; 0.79; 1.04 7.88; 3.92; 3.16; 2.88 49.04 25.2 91.0

8 Unknown country infection with local

2.62; 0.95; 0.80; 1.09 1.42; 0.59; 0.69; 0.97 7.88; 3.77; 3.19; 2.86 53.46 24.2 90.5

9 Unknown country infection with import

1.15; 0.63; 0.57; 0.82 0.62; 0.39; 0.50; 0.72 7.88; 3.57; 3.07; 3.01 103.48 24.4 89.6

10 10% more AIDS cases pre -cART

2.41; 0.87; 0.74; 1.05 1.30; 0.54; 0.61; 0.93 7.88; 3.83; 3.19; 2.89 76.26 24.0 90.4

11 10% more AIDS cases at diagnosis post-cART

2.39; 0.90; 0.75; 1.05 1.30; 0.56; 0.65; 0.92 7.88; 3.80; 3.29; 3.03 58.58 25.1 90.6

12 Added diagnosis data on people with unknown transmission route to locally acquired cases

2.41; 0.93; 0.82; 1.10 1.21; 0.58; 0.70; 0.94 7.88; 3.74; 3.27; 3.22 61.68 28.2 91.4


56

13 Added diagnosis data on people with unknown transmission route to imported cases

2.45; 0.81; 0.69; 0.95 1.33; 0.50; 0.59; 0.83 7.88; 3.79; 3.34; 3.13 78.62 24.9 90.1

14 All import divided among 5 disease stages, none in primary stage

3.00; 0.96; 0.90; 1.04 1.62; 0.61; 0.82; 0.97 7.88; 3.26; 2.64; 2.43 130.52 17.9 87.1

15 Time to diagnosis at AIDS stage: 1 week

2.4; 0.90; 0.76; 1.04 1.32; 0.56; 0.66; 0.93 7.82; 3.66; 3.11; 2.85 61.29 23.6 90.2

16 Time to diagnosis at AIDS stage same as time to diagnosis in earlier stages

2.08; 0.74; 0.65; 1.15 0.96; 0.39; 0.39; 0.72 9.69; 7.2; 6.4; 6.07 203.53 52.6 96.0

17 Fixed time to diagnosis

2.40; 0.89; 0.67; 0.99 1.31; 0.55; 0.59; 0.96 7.88; 4.00; 3.00; 2.00 102.91 16.5 88.1

18 Diagnosis rate unchanged after implementation of cART

2.34; 0.91; 0.80; 1.07 1.27; 0.58; 0.68; 0.91 7.88; 3.35; 3.35; 3.35 76.83 27.36 91.1

19 Behaviour unchanged after implementation of cART

2.18; 0.99; 0.75; 0.70 1.19; 0.62; 0.62; 0.62 7.88; 3.55; 3.69; 2.97 156.78 13.9 82.4

20 Transmission probability per sex act (instead of relative infectivity): primary infection and AIDS 0.183; chronic 0.014 [43]

4.27; 1.05; 0.61; 1.33 12.95; 3.56; 4.51; 10.06 7.88; 3.68; 3.13; 2.99 101.50 28.2 78.2

21 Full cART implementation in 2000 (instead of 1998)

2.39; 0.89; 0.74; 1.04 1.30; 0.56; 0.64; 0.92 7.88; 3.72; 3.15; 2.90 58.44 24.0 90.2

22 0% of AIDS cases treated in 1995

2.39; 0.89; 0.74; 1.05 1.30; 0.56; 0.64; 0.93 7.88; 3.70; 3.18; 2.89 59.28 24.5 90.4

23 Full cART implementation in 1996, and 0% of AIDS cases treated in 1995

2.39; 0.89; 0.75; 1.05 1.30; 0.56; 0.65; 0.93 7.88; 3.71; 3.19; 2.89 62.93 24.4 90.4

24 75% failing treatment each round

2.39; 0.89; 0.83; 1.13 1.30; 0.56; 0.64; 0.88 7.88; 3.71; 3.15; 2.90 58.87 25.6 85.6

25 75% failing treatment each round before 2000 and 25% thereafter

2.39; 0.89; 0.83; 0.95 1.30; 0.56; 0.64; 0.90 7.88; 3.72; 3.12; 2.90 58.44 24.2 90.11

26 10 rounds of treatment

2.39; 0.89; 0.73; 1.01 1.30; 0.56; 0.66; 0.93 7.88; 3.71; 3.16; 2.90 58.93 24.0 91.2

27 No cART in first disease stage

2.39; 0.89; 0.77; 1.05 1.30; 0.56; 0.65; 0.90 7.88; 3.71; 3.16; 2.90 58.88 24.1 88.2


57

28 Only cART in last two stages within 0.5 year

2.41; 0.91; 0.76; 1.08 1.31; 0.56; 0.59; 0.86 7.88; 3.87; 3.23; 2.91 102.95 23.9 83.6

29 After 2000 only cART in last two stages within 0.5 year

2.39; 0.89; 0.76; 1.14 1.30; 0.56; 0.66; 0.90 7.88; 3.71; 3.16; 2.90 59.47 24.7 85.8

30 Rate of starting subsequent treatment = 0.5

2.39; 0.89; 0.76; 1.04 1.30; 0.56; 0.64; 0.89 7.88; 3.72; 3.15; 2.90 58.72 24.7 88.0

31 Rate of failing treatment = 2 and Rate of starting subsequent treatment= 0.5

2.39; 0.89; 0.73; 1.02 1.30; 0.56; 0.62; 0.88 7.88; 3.72; 3.15; 2.90 58.41 25.3 87.2

32 Rate of failing treatment = 2 and Rate of starting subsequent treatment = 0.5 and 75% failing treatment each round and rate between stages in cART era = 1/3

2.43; 0.93; 0.90; 1.22 1.32; 0.57; 0.55; 0.74 7.88; 4.26; 4.56; 4.51 82.84 35.6 81.8

33 Rate of starting first treatment = 0.5

2.39; 0.89; 0.76; 1.04 1.30; 0.56; 0.64; 0.91 7.88; 3.71; 3.15; 2.90 58.52 24.1 89.0

34 5 time intervals: 1980-1984-1990- 1995-2000-2003

2.26; 0.98; 0.84; 0.76; 1.06

1.23; 0.61; 0.52; 0.66; 0.95

7.88; 3.75; 3.86; 3.12; 2.81

56.43 24.2 90.5

35 Time intervals: 1980-1984-1996- 1999-2004

2.51; 0.89; 0.68; 1.02 1.36; 0.56; 0.60; 0.90 7.9; 3.8; 3.1; 3.0 58.40 24.2 90.1

36 Time intervals: 1980-1984-1996- 2001-2004

2.39; 0.89; 0.80; 1.03 1.30; 0.56; 0.69; 0.94 7.88; 3.70; 3.19; 2.67 60.40 22.0 89.8

37 Rate between stages in cART era = 1/3

2.42; 0.93; 0.83; 1.13 1.32; 0.57; 0.61; 0.84 7.88; 4.25; 4.59; 4.48 85.14 32.0 91.6

38 Rate between stages in all eras = 1/2

2.34; 0.89; 0.74; 1.06 1.22; 0.53; 0.63; 0.92 8.32; 4.13; 3.36; 3.02 58.02 24.7 90.5

39 Rate of failing treatment = 2

2.39; 0.89; 0.73; 1.03 1.30; 0.56; 0.64; 0.91 7.88; 3.71; 3.16; 90 58.72 24.5 89.4

40 Death rate = 0.03 2.39; 0.89; 0.74; 1.04 1.30; 0.56; 0.66; 0.93 7.88; 3.71; 3.16; 2.90 58.99 26.4 90.7

41 4th interval: 2000-2004

2.38; 0.90; 0.74; 1.05 1.29; 0.56; 0.64; 0.94 7.88; 3.72; 3.14; 2.80 61.46 23.8 90.4


58

Data update What if

R(t)respectively

in: 1980-1983; 1984-1995; 1998-1999; 2000-2003; 2004-2006

b(t)respectively

in: 1980-1983; 1984-1995; 1996-1999; 2000-2003; 2004-2006

time to diagnosis respectively

in: 1980-1983; 1984-1995; 1996-1999; 2000-2003; 2004-2006

dev/2 * %

of H

IV p

osit

ives

un

awar

e at

the

end

of

the

mod

el fi

t

% n

ew in

fect

ions

fr

om u

naw

are

at t

he

end

of t

he m

odel

fit

42 4th period: 2000-2003; 5th period 2004-2006

2.37; 0.89; 0.75; 1.01; 1.00

1.28; 0.56; 0.65; 0.90; 0.93

7.88; 3.64; 3.14; 2.83; 2.47

67.92 20.3 89.1

Lower 95% confidence limit**

2.10; 0.86; 0.70; 0.97; 0.91

1.14; 0.53; 0.60;0.87; 0.84

***; 3.41; 2.96; 2.63; 2.21

69.84 **** ****

Upper 95% confidence limit**

2.56; 0.92; 0.81;1.06; 1.05

1.39; 0.58; 0.71;0.96; 0.97

***; 3.88; 3.25; 3.03; 2.73

69.84 **** ****

43 1st period: 1980 – 1984 1.92; 0.86; 0.77; 1.02; 0.99

1.04; 0.53; 0.67; 0.91; 0.92

7.88; 3.79; 3.12; 2.82; 2.43

45.74 19.7 88.8


2.33; 0.89; 0.75;1.01; 1.03

1.26; 0.56; 0.65;0.92; 0.94

7.88; 3.62; 3.11; 2.77; 2.65

71.70 21.93 89.9

45 4th period: 2000-2006 2.33; 0.89; 0.74; 1.03 1.26; 0.56; 0.64; 0.93 7.88; 3.63; 3.10; 2.72 72.15 22.4 90.0

46 4th period: 2000-2005 2.32; 0.90; 0.74; 1.05 1.26; 0.56; 0.64; 0.94 7.88; 3.63; 3.11; 2.81 65.45 23.9 90.5

47 4th period: 2000-2004 2.32; 0.90; 0.74; 1.05 1.26; 0.56; 0.64; 0.94 7.88; 3.63; 3.11; 2.81 65.45 24.2 90.5

48 4th period: 2000-2003 2.32; 0.90; 0.74; 1.07 1.26; 0.56; 0.65; 0.96 7.88; 3.63; 3.10; 2.83 62.06 25.2 90.8

49 Only 25% reduction in risk behaviour after diagnosis in 5th period 2004-2006

2.39; 0.89; 0.76;1.01; 1.03

1.30; 0.56; 0.66;0.90; 0.89

7.88; 3.64; 3.13;2.83; 2.46

67.66 20.3 84.64


2.32; 0.89; 0.75; 1.01; 0.99

1.26; 0.56; 0.66; 0.91; 0.92

7.88; 3.62; 3.12; 2.82; 2.46

67.63 19.0 88.6

51 4th period: 2000 – 2007 2.27; 0.90; 0.73; 1.03 1.23; 0.56; 0.64; 0.93 7.88; 3.63; 3.10; 2.72 72.27 22.0 89.9

52 g s= 1.5 from 2004; t s = 0.35 from 2004; 4th period: 2000-2003; 5th period 2004-2006,

2.30; 0.90; 0.77;1.01; 0.93

1.25; 0.56; 0.67;0.91; 0.94

7.88; 3.63; 3.14;2.83; 2.47

69.70 19.6 92.3

53 g s = 1.5 from 2004; t s = 0.35 from 2000; 4th period: 2000-2003; Only 25% reduction in risk behaviour after diagnosis in 5th period 2004-2006

2.33; 0.90; 0.74; 0.98; 0.95

1.26; 0.56; 0.67;0.92; 0.92

7.88; 3.63; 3.13;2.82; 2.46

70.46 19.2 90.5

54 g s = 1.5 from 2003; t s = 0.35 from 2003; 4th period 2000-2002; 5th period 2003-2006

2.32; 0.90; 0.76;1.01; 0.97

1.26; 0.56; 0.66; 0.92; 0.96

7.88; 3.63; 3.12; 2.76; 2.65

72.37 21.5 93.3


2.32; 0.90; 0.78; 1.02; 0.94

1.26; 0.56; 0.68; 0.92; 0.95

7.88; 3.63; 3.14;2.81; 2.46

70.06 19.3 92.8


2.36; 0.90; 0.77; 0.98; 0.95

1.28; 0.56; 0.68; 0.95; 0.96

7.88; 3.68; 3.16; 2.82; 2.47

82.27 19.3 93.6

57 g s= 1.5 from 1995; t s = 0.35 from 2000; 4th period 2000-2006

2.33; 0.89; 0.71; 0.99 1.27; 0.56; 0.67; 0.97 7.88; 3.63; 3.12; 2.72 74.54 21.8 93.9


59


2.33; 0.90; 0.70; 1.00 1.26; 0.56; 0.65; 0.98 7.88; 3.67; 3.16; 2.70 68.76 22.55 94.0


2.33; 0.89; 0.71; 0.99 1.27; 0.56; 0.67; 0.97 7.88; 3.63; 3.12; 2.72 74.48 21.3 93.9

* Analyses are based on minimizing the deviance, denoted Dev. Values reported here are half the deviance. ** 95% confidence intervals obtained by the likelihood ratio applied to the univariate marginal likelihood profile. Confidence intervals for R(t) are based on those for ß(t). *** No confidence intervals are given for the time to diagnosis for the first period, since this is the mean time to AIDS estimated from other data. **** No confidence intervals are given for these as they are not parameters and univariate likelihood profiles could not be derived.

cd4 cell count at diagnosisThough we did not fit our model to CD4 count data, a qualitative comparison of changes in CD4 count at diagnosis is informative as a corroboration of our inferences on the role of chang-ing patterns of diagnosis. We compared the estimated proportion of newly diagnosed patients in each disease stage as estimated by our best fit model with data on CD4 cell count at diagno-sis. Therefore, we describe the five stages in our model as the following CD4 cell count intervals in cells/mm3: stage 1 = > 600; stage 2 = 400-600; stage 3 = 250-400; stage 4 = 100-250; stage 5 = 0-100. Figure 3a shows these stages plotted for MSM in ATHENA as a function of the year of diagnosis. While our model disease stages can only approximately be identified with cat-egories based on CD4 cell count (Figure 3b,c), three observations can be made. First, we have a tendency to overestimate the extent to which individuals are diagnosed in the early stage of infection. Second, apart from this mismatch, our model reproduces satisfactorily the observed temporal trends, and third, the recent increase in the proportion of newly diagnosed individuals with high CD4 cell counts corroborates our model’s inferences in interpreting recent increases in annual number of new HIV diagnosis as rising transmission and increased diagnosis rather than improved diagnosis of people infected many years in the past.


60

0%

25%

50%

75%

100%

1985 1988 1991 1994 1997 2000 2003 2006Year

Per

cent

age

in e

ach s

tage

0%

25%

50%

75%

100%

1985 1988 1991 1994 1997 2000 2003 2006Year

Per

cent

age

in e

ach s

tage

0%

25%

50%

75%

100%

1985 1988 1991 1994 1997 2000 2003 2006Year

Per

cent

age

in e

ach s

tage

a

b

c

figure 3. Using CD4 count at diagnosis as a surrogate of the time since infection at diagnosis. We defined five CD4 cell count intervals for every model stage of infection with average duration of 1.89 years, and show these plotted for MSM in ATHENA as a function of year of diagnosis in a, where the proportion of people diagnosed in each stage is shown ranging from earliest (lightest) to last (AIDS, darkest). These estimates are biased by the fact that that only people who survive until 1996 are included in our study. The subsequent figures (b and c) show the estimated proportion of newly diagnosed patients in each disease stage as estimated by our best fit model. b includes the same process of truncation which generates bias prior to 1996 as the data (and is thus to be compared to a), while c indicates what we estimate actually occurred in the absence of this bias.


61

Hypothetical scenario analysisBased on our best model fit, we explored a number of hypothetical “what if” scenarios, based on the assumption that the relative importance of the other contributing factors remains constant. All of the numbers below should be interpreted as ballpark approximations – exact estimates are available from figure 4, but are provided in the text rounded to the nearest hundred to reduce false accuracy.

0

2000

4000

6000

8000

10000

12000

14000

1995 1998 2001 2004 2007Year

a

0

2000

4000

6000

8000

10000

2007 2010 2013 2016Year

b

Cum

ulat

ive

infe

ctio

ns fr

om 2

006

Cum

ulat

ive

infe

ctio

ns fr

om 1

995

figure 4. Hypothetical scenarios for past and future decade. a, during past decade. Best fit model (thick, R = 1.00). Had there been no treatment with cART (long-dash, R = 1.43). Had there been no increase in diagnosis rate (short-dash, R = 1.12). Had there been no increase in risk behaviour rate (starred, R = 0.60). Had there been no treatment, no increases diagnosis and no increase in risk behaviour rate (thin, R = 0.89). b, during next decade. If nothing changes (thick, R = 1.04). If proportion failing each line of cART is reduced to ten percent (short dash, R = 0.92). If average time to diagnosis is reduced to one year (see methods) (long dash, R = 0.86). If the risk behaviour rate is back to pre- cART levels (thin, R = 0.60). All three interventions (starred, R = 0.46).


62

Firstly, we explored a number of scenarios from 1995 onwards (Figure 4A). In the absence of cART limiting infectiousness in treated patients, the epidemic under current conditions would have been much larger, with an estimated 11,300 infections arising between 1995 and 2006, instead of the estimated 4,900. If on the other hand, cART had been introduced but there had been no increases in the risk behaviour rate, the number of new infections over this period would be only 2,000. If cART had been introduced and there had been the increase in the risk behaviour rate, but no increases in the diagnosis rate, the cumulative number of new infections would be 5,800. Finally, if no changes had occurred since 1995, i.e. no cART and no increas-ing risk and testing behaviour, this number would be 3,900. Thus based on these model esti-mates, we conclude that cART has played an important role in limiting transmission, but that any gains made have been more than offset by increases in the risk behaviour rate. Had these increases not occurred in the cART era, the reproduction number R(t) would have declined to 0.6, and the epidemic would have entered in a convincing decline.

Furthermore, we explore a number of hypothetical scenarios for the coming decade (Figure 4B). If nothing changes, the epidemic will spread uncontrolled and the cumulative number of infections between 2007 and 2016 will reach 7,700. If the frequency of testing is increased such that the mean time from infection to diagnosis (and subsequent treatment) is reduced to one year, this number will be reduced to 4,406. If the quality of treatment on offer is improved, such that the fraction of patients failing each line of therapy is only ten percent, then the cumulative number of new infections will reach 5,600. If the risk behaviour rate is reduced back to pre-cART levels, this number can be reduced to 1,700. Finally if all three interventions could be successfully implemented it is further reduced to only 800 new infections. From this analysis, we conclude that reducing the risk behaviour rate has the greatest impact on the epidemic, though earlier diagnosis and treatment can also prevent almost half of the infections.

discussion and conclusion

The model we presented in this paper is shown to be a robust and valuable tool for study-ing HIV epidemics in times of cART. We simultaneously estimated both changes in average risk-behaviour and time to diagnosis. Using these estimates, we calculated the reproduction number, which summarizes the state of the epidemic. In all sensitivity analyses of the cur-rent situation among MSM in the Netherlands, the reproduction number R(t) is estimated to lie very close to its threshold value of one, indicating current epidemic spread driven by local transmission. Under all assumptions, the net transmission rate b(t) (the per-capita rate at which infectious individuals infect new people) has shown no improvement over the past decade. The time from infection to diagnosis however has steadily decreased to 2.5 years on average in 2006, indicating improvements in testing. The qualitative comparison with CD4 data underscored


63

that the increase in annual diagnosis is due to a recent increase in transmissions. Data on rectal gonorrhoea and recent syphilis diagnoses have the same temporal curve as the incidence curve resulting from our model fit [4]. R(t) is lower now than when the epidemic was at its peak in the beginning of the eighties, but as it is now taking place on a much larger scale almost as many MSM are infected annually. Currently, around 90% of new infections are estimated to be trans-mitted from the undiagnosed infected population, being 24% of the total infected population. Thus, the epidemic is driven by MSM having risky sex not being aware they are infected, and curiously the absolute magnitude of this group has remained similar over time.

That sexual behaviour has increased seems to some extent a normal reaction to the reduced risk, and as shown in the hypothetical scenarios cART has substantially mitigated the effects of this return to pre-AIDS era risk behaviour levels which otherwise would have caused twice as many cases. Hypothetical scenarios further showed that, if nothing changes, twice as many MSM will be in need of health care for being HIV-1 infected in the coming decade than at present. The most effective way to prevent this is to decrease risk-behaviour. A rough estimate of average number of unsafe sex acts needed for current HIV-1 spread is around ten per year. So even though the per-sex act change of transmission might seem small, this clearly shows that epidemic spread is possible, especially since one still cannot easily be diagnosed during the primary stage when one is most infectious [28].

There is a lot of scope to improve the rate of diagnosis during asymptomatic infection [5, 32-34] and the potential for diagnosing individuals during the acute stage of infection [35]. This model analysis does not produce reliable estimates of what proportion of transmission is due to acute (primary) infection, and what proportion occurs during asymptomatic set-point infection. This proportion needs to be estimated by other means. As the partner change rate during the acute stage of infection is unknown, we could not predict to what level the epidemic can be reduced without testing during that window-period. From phylogenetic studies it seems that about 25% of onward transmission occurs during that stage [36], implying that using classical serological diagnosis to limit infections after the acute phase of infection could be very effective in control-ling the epidemic. The role of contact tracing and virological diagnosis to diagnose people in acute stage of infection should be explored.

In the Netherlands there has been a health care policy towards promoting safe sex rather then HIV testing [33]. It would be interesting to do qualitative comparison with other countries on CD4 counts at diagnosis to see if the different policies in the past indeed still lead to different stages of infection at diagnosis.

A key assumption, which is biologically plausible but remains unvalidated in practice, is that individuals who are successfully treated, and who maintain very low levels of virus in blood, do


64

not contribute to onwards infections at all. Individuals who fail therapy are assumed to be as infectious as untreated individuals. While it is unlikely that mathematical model-based analy-ses of the type used here can validate this assumption, alternative scenarios which consider the effects of relaxing this assumption in data-driven analyses should be explored in detail in future work [19, 37-39]. Previous modelling studies focussing on hypothetical scenarios have empha-sised the potential that ongoing transmission during incompletely suppressive antiretroviral therapy could have in driving epidemics [11-19, 37, 38]. We cannot rule out such a contribution to the Dutch HIV epidemic, an issue that could be best addressed by empirical discordant pair studies.

In this paper we performed a univariate sensitivity analysis to explore the impact of different model assumptions. This has the benefit of being easy to understand, as the impact of every assumption can be checked for and understood separately. However a multivariate sensitivity analysis will be developed in the future to more rigorously test the dependence of the model pre-dictions to different assumptions. The effect of the opting-out testing policy can also be evalu-ated in the near future. This will provide an interesting population-based test of the hypothesis that more active diagnosis and treatment programs can contribute to mitigating what is cur-rently a re-emerging epidemic. It would also be interesting to use the model we presented in this paper to study other well monitored epidemics, and to extend this model for studying the transmission of resistance [40].

Acknowledgements

DB was supported by grant 7014 from AIDS Fund Netherlands and by a travel grant from NWO (Netherlands Organisation for Scientific Research). CF is funded by the Royal Society. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

reference List

1. Bezemer D, de Wolf F, Boerlijst MC, van Sighem A, Hollingsworth TD, Prins M, et al. A resurgent HIV-1

epidemic among men who have sex with men in the era of potent antiretroviral therapy. AIDS 2008;

22(9):1071-1077.

2. Jaffe HW, Valdiserri RO, De Cock KM. The reemerging HIV/AIDS epidemic in men who have sex with

men. JAMA 2007; 298(20):2412-2414.

3. White PJ, Ward H, Garnett GP. Is HIV out of control in the UK? An example of analysing patterns of HIV

spreading using incidence-to-prevalence ratios. AIDS 2006; 20(14):1898-1901.


65





HIV, among young homosexual men in Amsterdam: are STIs still reliable markers for HIV transmission? Sex

Transm Infect 2005; 81(1):34-37.






2004; 18(14):1943-1949.

8. Stolte IG, Dukers NH. Response to ‘High-risk sexual behaviour increases among London gay men between

1998 and 2001: what is the role of HIV optimism?’. AIDS 2003; 17(13):2011-2012.

9. Stolte IG, Coutinho RA. Risk behaviour and sexually transmitted diseases are on the rise in gay men, but

what is happening with HIV? Current Opinion in Infectious Diseases 2002; 15(1):37-41.

10. Truong HHM, Kellogg T, Klausner JD, Katz MH, Dilley J, Knapper K, et al. Increases in sexually trans-

mitted infections and sexual risk behaviour without a concurrent increase in HIV incidence among men who

have sex with men in San Francisco: a suggestion of HIV serosorting? Sexually Transmitted Infections 2006;

82(6):461-466.




cisco. Science 2000; 287(5453):650-654.






15(17):2293-2301.







17(7):1029-1038.




66


22 Suppl 1:S149-S159.

20. Abbas UL, Anderson RM, Mellors JW. Potential impact of antiretroviral therapy on HIV-1 transmission and

AIDS mortality in resource-limited settings. Jaids-Journal of Acquired Immune Deficiency Syndromes 2006;

41(5):632-641.

21. Lima VD, Johnston K, Hogg RS, Levy AR, Harrigan PR, Anema A, et al. Expanded access to highly active

antiretroviral therapy: A potentially powerful strategy to curb the growth of the HIV epidemic. Journal of

Infectious Diseases 2008; 198(1):59-67.

22. Desai K, Sansom SL, Ackers ML, Stewart SR, Hall HI, Hu DJ, et al. Modeling the impact of HIV chemo-

prophylaxis strategies among men who have sex with men in the United States: HIV infections prevented and

cost-effectiveness. AIDS 2008; 22(14):1829-1839.

23. Granich RM, Gilks CF, Dye C, De Cock KM, Williams BG. Universal voluntary HIV testing with imme-

diate antiretroviral therapy as a strategy for elimination of HIV transmission: a mathematical model. Lancet

2009; 373(9657):48-57.

24. Fraser C. Estimating individual and household reproduction numbers in an emerging epidemic. PLoS ONE

2007; 2(1):e758.

25. Macy RI, Oster GF. Berkeley Madonna, version 8.3.8, http://www.berkeleymadonna.com/. In: 2009.

26. Marks G, Crepaz N, Senterfitt JW, Janssen RS. Meta-analysis of high-risk sexual behavior in persons aware

and unaware they are infected with HIV in the United States: implications for HIV prevention programs.

J Acquir Immune Defic Syndr 2005; 39(4):446-453.

27. Gras L, van Sighem A, van Valkengoed I, Zaheri S, de Wolf F. Monitoring of human immunodeficiency virus

(HIV) infection in the Netherlands, Chapter 10, http://www.hiv-monitoring.nl , HIV Monitoring Founda-

tion. In. Amsterdam: Stichting HIV Monitoring; 2003.

28. Hollingsworth TD, Anderson RM, Fraser C. HIV-1 transmission, by stage of infection. J Infect Dis 2008;

198(5):687-693.

29. Gras L, van Sighem A, Smit C, Zaheri S, de Wolf F. Monitoring of human immunodeficiency virus (HIV)

infection in the Netherlands, Chapter 6, 9&10, http://www.hiv-monitoring.nl , HIV Monitoring Founda-

tion. In. Amsterdam; 2006.

30. Postma MJ, Jager JC, Dijkgraaf MG, Borleffs JC, Tolley K, Leidl RM. AIDS scenarios for The Netherlands;

the economic impact on hospitals. Health Policy 1995; 31(2):127-150.

31. van Veen M, Koedijk F, van den Broek I, Op de Coul E, de Boer I, van Sighem A, et al. HIV and Sexually

Transmitted Infections in the Netherlands in 2006, http://www.rivm.nl/bibliotheek/rapporten/210261003.

pdf. In. Bilthoven: National Institute for Public Health and the Environment; 2007.

32. Smit C, Hallett TB, Lange J, Garnett G, de WF. Late entry to HIV care limits the impact of anti-retroviral

therapy in The Netherlands. PLoS ONE 2008; 3(4):e1949.

33. Stolte IG, de Wit JB, Kolader ME, Fennema HS, Coutinho RA, Dukers NH. Low HIV-testing rates among

younger high-risk homosexual men in Amsterdam. Sex Transm Infect 2007; 83(5):387-391.


67

34. Dukers NH, Fennema HS, van der Snoek EM, Krol A, Geskus RB, Pospiech M, et al. HIV incidence and

HIV testing behavior in men who have sex with men: using three incidence sources, The Netherlands,

1984-2005. AIDS 2007; 21(4):491-499.

35. Cohen MS, Kaleebu P, Coates T. Prevention of the sexual transmission of HIV-1: preparing for success. J Int

AIDS Soc 2008; 11(1):4.

36. Lewis F, Hughes GJ, Rambaut A, Pozniak A, Leigh Brown AJ. Episodic sexual transmission of HIV revealed

by molecular phylodynamics. PLoS Med 2008; 5(3):e50.

37. Anderson RM, Gupta S, May RM. Potential of community-wide chemotherapy or immunotherapy to control

the spread of HIV-1. Nature 1991; 350(6316):356-359.

38. Wilson DP, Law MG, Grulich AE, Cooper DA, Kaldor JM. Relation between HIV viral load and infectious-

ness: a model-based analysis. Lancet 2008; 372(9635):314-320.

39. Bogaards JA, van Ballegooijen WM, Jan WG, Boerlijst MC, Goudsmit J. Is population-level perversity a

likely outcome of mass vaccination against HIV? Lancet Infect Dis 2005; 5(5):254-256.

40. Phillips AN, Sabin C, Pillay D, Lundgren JD. HIV in the UK 1980-2006: reconstruction using a model of

HIV infection and the effect of antiretroviral therapy. HIV Med 2007; 8(8):536-546.

41. Van Griensven GJP, Devroome EMM, Tielman RAP, Goudsmit J, Van der Noordaa J, de Wolf F, et al.

Impact of Hiv Antibody Testing on Changes in Sexual-Behavior Among Homosexual Men in the Nether-

lands. American Journal of Public Health 1988; 78(12):1575-1577.



43. Vittinghoff E, Douglas J, Judson F, McKirnan D, MacQueen K, Buchbinder SP. Per-contact risk of human

immunodeficiency virus transmission between male sexual partners. Am J Epidemiol 1999; 150(3):306-311.

44. van Griensven GJ, Tielman RA, Goudsmit J, Van der Noordaa J, de Wolf F, de Vroome EM, et al. Risk

factors and prevalence of HIV antibodies in homosexual men in the Netherlands. Am J Epidemiol 1987;

125(6):1048-1057.



191(9):1403-1409.



104(44):17441-17446.

47. Jansen VA, Stollenwerk N, Jensen HJ, Ramsay ME, Edmunds WJ, Rhodes CJ. Measles outbreaks in a popu-

lation with declining vaccine uptake. Science 2003; 301(5634):804.

48. Grassly NC, Fraser C. Mathematical models of infectious disease transmission. Nat Rev Microbiol 2008;

6(6):477-487.

49. Wallinga J, Lipsitch M. How generation intervals shape the relationship between growth rates and reproduc-

tive numbers. Proceedings of the Royal Society B-Biological Sciences 2007; 274(1609):599-604.

50. Amundsen EJ, Stigum H, Rottingen JA, Aalen OO. Definition and estimation of an actual reproduction

number describing past infectious disease transmission: application to HIV epidemics among homosexual

men in Denmark, Norway and Sweden. Epidemiol Infect 2004; 132(6):1139-1149.


68

Appendix on derivation of the model

survival distributionTo model the natural history of infection in untreated individuals, an Erlang survival distri-bution was fitted to data from 130 MSM seroconverters before the cART era in the Amster-dam Cohort Studies [44]. The maximum likelihood estimate was for an Erlang distribution of degree 5 and rate 1/1.89 per year. This best fit survival distribution can be modelled using a compartmental model, with unidirectional flow through five compartments with mean stay in each of 1.89 years (Figure 5). These compartments can roughly be understood as stages of progression of disease, but the main motivation for their use is to replicate in detail the survival distribution of untreated HIV infection.

0%

25%

50%

75%

100%

0 2 4 6 8 10 12

Ye ars following se roconv e rsion

Pro

porti

on s

urvi

ving

figure 5. Best fit survival distribution to data from 130 MSM seroconverters before the cART era in the Amsterdam Cohort Studies (all data were truncated from 22 Nov 1993 when the first person received a protease inhibitor). Solid line, Kaplan-Meier distribution estimated from data and dashed line, best fit Erlang distribution.

relative infectiousness of different stages of disease progressionThe relative infectiousness of different stages was adapted from Hollingsworth et al.[28]. This was obtained by fitting a hazard model to the data from Wawer et al.[45], a study of HIV-1 transmission in serodiscordant heterosexual couples in Uganda. Infectiousness is 25-fold higher for primary infection stage, assumed to last 0.24 years, than during asymptomatic infection (model stages 1-4, mean duration 7.56 years), and 3.2-fold higher during the AIDS stage (model stage 5, mean duration 1.89 years). In the absence of better data, it was assumed here that these relative infectiousness values were intrinsic features of natural infection, and thus similar for


69

this population of predominantly Dutch MSM. Dependence of the modelling results on these values was explored in the sensitivity analysis.

The relative infectiousness values refer to infectiousness within HIV-1 serodiscordant partner-ships. The relative contribution of different stages will depend somewhat on the frequency of partner change, per sex act transmission probabilities, and the details of the sexual network [46]. We find in the sensitivity analysis that our main conclusions are not overly dependent on the value of the relative infectiousness parameters.

Net transmission rateThe net transmission rate b(t) is a time-varying function that determines the rate at which indi-viduals infect other people. It is defined up to an arbitrary constant, and is thus a relative rate. It only makes sense to compare this rate in different time periods rather than to assign meaning to its absolute value. The model also includes modifiers to describe the effects of stage of infection, diagnosis and treatment on infectiousness, described later.

Our main conclusions are based on fitting this net transmission rate through stages of the epi-demic, a method which we expect to be fairly independent of the details of the sexual partner network. Similarly, because the overall prevalence of infection remains very low throughout the observed period, we do not explicitly account for “saturation” (e.g. depletion of the susceptible “pool”). b(t) is primarily intended as a measure of changes in risk behaviour between discordant couples, and one should note that b(t) is a compound measure which is affected by changes in the partner change rate, by the rate and nature of risky sex acts within partnerships, by the effect of saturation of the susceptible population, by the effect of the changing prevalence of other STIs in modulating HIV transmission, by risk-management strategies such as “sero-sorting”, and even by possible secular changes in the infectiousness of the virus.


70

Incidence of InfectionOur model relates the incidence rate of infection to the prevalence of infections, stratified by disease stage and treatment state. The parameters on infectiousness are defined as relative rate modifiers for different stages of disease progression, of diagnosis and of treatment. The total rate at which people in primary infection infect others, IY p, is given by

importlocalIY p = (t) p(Y p + Y p )

where r p is the infectiousness of primary infection relative to other disease stages, importY P and

localY p are the number of people currently in the primary infection stage, who acquired infection abroad (import cases), and in the Netherlands (local cases). After primary infection, people progress through a series of stages labeled s = 1,…,5. The total rate at which people before diag-nosis infect others during these stages is

importlocalIY s = (t) s(Y s + Y s )

where r s is the relative infectiousness of disease stage s, importY s and localY s are the number of people in disease stage s. After diagnosis, people reduce their risk behaviour by a factor s due to aware-ness of their infection status, so that the total rate at which diagnosed, untreated people infect others, diag

IY s , is

diag diagIY s = (t) sY s

where diagY s is the number of diagnosed untreated people in stage s, (s = 1, …,5). People are

assumed not to be infectious while they are successfully treated. For people who have failed treatment f times, where f = 1, …,3 , the total rate at which they infect others is

f fIF s = (t) sF s

where fF s is the number of people who have failed treatment f times who are in disease stage s.

The total incidence rate of infection, I(t), is the sum of these terms

fdiags=1

5

f=1

3I(t) = IY p + Σ IY s+ IY s + Σ IF s


71

Undiagnosed primary Infections – Highly infectiousThe rate of change in the number of locally acquired cases that are in primary infection is given by

locallocal

dY p

dt= I(t) − pY p

where I(t) is the total incidence rate of infection, 1/a p is the average duration of the primary infection stage, and local

Y p are the number of locally acquired cases that are in the primary infec-tion stage. The rate of changes in the number of imported cases that are in primary infection is given by

importimport

dY p

dt= A(t) f p− pY p

where A(t) are the number of imported cases per year, f P is the fraction of imported cases that are in primary stage, and importY P are the number of imported cases in the primary infection stage.

Undiagnosed cases progressing Through stages of Infection - InfectiousChanges in the number of locally acquired undiagnosed untreated people in stage s, (s = 1,…,5) is given by

locallocal local local

dY s

dt= pY p − Y s − s(t)Y s ; (s = 1)

locallocal local local

dY s

dt= Y s-1 − Y s − s(t)Y s ; (s = 2,…,5)

where a describes the rate of progression to each subsequent stage, d s(t) describes the rate at which people are diagnosed when in stage s, (s = 1,…,5) and is defined as a time-varying func-tion, and localY s are the number of locally infected cases in stage s, (s = 1,…,5). Changes in the number of imported undiagnosed untreated people in stage s, (s = 1,…,5) is given by

importimport import

dY s

dt1− f p

5= A(t) + pY p − ( + 1(t))Y s ; (s = 1)

import import1− f P

5= A(t) + Y s-1 − ( + s(t))Y s ; (s = 2,…,5) import

dY s

dt

where importY s are the number of imported cases in stage s (s = 1,…,5).


72

diagnosed cases progressing Through stages of Infection – Infectious, but with reduced risk behaviourChanges in the number of all diagnosed untreated people together in stage s (s = 1,…,5) is given by

diagimport local diag

import local diag diag

dY s

dt

diagdY s

dt= s(t) (Y 1 + Y 1 ) + Y s-1 − ( + s(t))Y s ; (s = 2,…,5)

= s(t) (Y 1 + Y 1 ) − ( + s(t))Y s ; (s = 1)

where diagY s are the number of diagnosed cases in stage s, (s = 1,…,5), and g s(t) the rate per year

of starting treatment and suppressing viral load at stage s when therapy naive.

temporarily successful treated cases – Not InfectiousThe rate of change in the number of temporarily successful treated cases for the three rounds of treatment ( f = 1,…,3) is defined by

ff fdiag

dQ s

dt

fdQ s

dt

= s s(t)Y s − ( + s)Q s ; (s = 1,…,5, f = 1)

f ff-1 f-1= s s (t)F s − ( + s)Q s ; (s = 1,…,5, f = 2,3)

where fQ s are the number of cases in stage s, (s = 1,…,5) on treatment round f, ( f = 1,…,3), t s is the fraction of cases starting therapy at stage s, (s = 1,…,5) that will have only a temporarily successful viral load suppression when starting treatment, f-1 s is the rate of starting treatment and obtain viral load suppression at stage of infection s, (s = 1,…,5) after treatment failure f, ( f = 1,2), f s is the rate of failing treatment at stage s, (s = 1,…,5) after treatment failure f, ( f = 1,…,3), and m is the basic death rate.


73

failing cases – InfectiousThe rate of change in the number of people failing treatment f times ( f = 1,…,3) is defined as

ff ff

dF s

dt= s Q s − ( + s )F sf ; (s = 1; f = 1,2)

ff ff

dF s

dt= s Q s + F s-1 − ( + s )F s f f ; (s = 2,…,5; f = 1,2)

ff ff

dF s

dt= s Q s − F s ; (s = 1; f = 3)

ff ff

dF s

dt= s Q s + F s-1 − F s f ; (s = 2,…,5; f = 3)

where fF s are the number of cases in stage s , (s = 1,…,5) failing treatment round f , ( f = 1,…,3), people failing the third round of therapy ( f = 3) have run out of treatment options and continue disease progression.

enduringly successfully treated cases – Not InfectiousThe rate of change in the number of enduringly successfully treated people is defined by

dTdt

= Σ [(1 − s) s(t)Y s + (1 − s) sF s + (1 − s) sF s ] − T diag 1 1 2 2s=1

5

where T is the number of cases that are enduringly successfully treated.

calculation of the reproduction NumberThe reproduction number R(t) can be defined as a time-varying function which is the average number of people an infected at time t would infect over his whole infectious lifespan if condi-tions remained the same as at time t [24]. R(t) is calculated as the sum of multiplications for all infectious stages of: the average duration in each stage of infection; the probability to get in that stage of infection; the infectiousness of that stage of infection (r s); and the net transmission rate (b(t)).

The average duration in ‘Primary Infection’ is

1pDY p =


74

The probability to get into ‘Primary Infection’ (PY p) is 1.The contribution to the Reproduction Number of the ‘Primary Infection’ stage is

RY p = (t) pDY pPY p

The average duration in ‘Undiagnosed Cases Progressing through Stages of Infection’ is

; (s = 1,…,5)1+ s (t)DY s =

The probability to get into ‘Undiagnosed Cases Progressing through Stages of Infection’ is

; (s = 2,…,5)+ s-1 (t)PY s = PY s-1

; (s = 1)PY s = 1

The contribution to the Reproduction Number of the ‘Undiagnosed Cases Progressing through Stages of Infection’ is

; (s = 1,…,5)RY s = (t) sDY sPY s

The average duration in ‘Diagnosed Cases Progressing through Stages of Infection’ is

; (s = 1,…,5)1+ s (t)DY s =

diag

The probability to get into ‘Diagnosed Cases Progressing through Stages of Infection’ is

; (s = 1) s (t)+ s (t)PY s = PY s

diag

; (s = 2,…,5)s (t)

+ s (t)PY s = PY s + + s-1 (t) PY s-1 diag diag

The contribution to the Reproduction Number of the ‘Diagnosed Cases Progressing through Stages of Infection’ is

diag diag diag ; (s = 1,…,5)RY s = (t) sDY s PY s


75

The average duration in ‘Failing Cases’ is

; (s = 1,…,5; f = 1,2)1+ s DF s =

ff

; (s = 1,…,5; f = 3)1DF s = f

The probability to get into ‘Failing Cases’ is

; (s = 1; f = 1)+ s PF s = 1f

f

f s

+ s (t) PY sdiag

; (s = 2,…,5; f = 1) + s PF s = sf

f f

f s

+ s (t)g s (t)

diagPY s + + s-1 PF s-1 f

; (s = 1; f = 2) + s PF s = 1f

f

f s

f-1+ s PF sf-1

; (s = 2,…,5; f = 2) + s PF s = sf

f f

f sf-1PF s + + s-1 PF s-1

f

; (s = 1; f = 3)+ 1 PF s = sf

f

f s PF sf-1

; (s = 2,…,5; f = 3) + s PF s = sf

f

f sf-1PF s + PF s-1

f

f-1s

f-1 + sf-1 s

f-1 + sf-1 1

f-1 + sf-1 s

The contribution to the Reproduction Number of the ‘Failing Cases’ is

f f f ; (s = 1,…,5; f = 1,…,3)RF s = (t) sDF s PF s

The total Reproduction Number R(t) is defined as the sum of these terms

fR(t) = RY p Σ RY s + RY s + Σ RF s diags=1

5

f=1

3

We verified our formula numerically by checking for correct monotonicity in all the transmis-sion and treatment parameters, and checking that R(t) = 1 is indeed the threshold between epidemic growth and decline when there is no import [47]. In the presence of imports, the threshold separates exponential epidemic growth (when R(t) > 1) from a steady equilibrium of outbreak (when R(t) < 1).


76

For this model of HIV, with its very long infectious period, the number of cases can grow slowly for many years approaching this equilibrium even when R(t) < 1 if R(t) is close to 1, so that the threshold (R(t) = 1) does not in practical terms mark a sudden change in dynamics in the way it does for classical models of acute infections with short infections [48].

We note that in the case of HIV, the epidemic situation can change substantially within the incubation period of a single individual, so that the actual lifetime reproduction number of an individual may differ substantially from the instantaneous value we estimate. Thus, R(t) as defined is in essence a descriptor of the epidemic at a point in time, similar in spirit to the exponential growth rate [49] or the incidence to prevalence ratio [3, 50]. There are however two advantages to R(t) over these alternative measures, namely that it is sensitive to the effects of the changing generation time induced by changing diagnosis rates and cART, and it has the meaningful threshold at R(t) = 1.

fitting to AIds casesThe accumulation of new AIDS cases is defined as

d(AIDS)dt

= 5(t)(Y 5 + Y 5 ) + Y 4 + Σ F 4

local import fdiag f=1

3

where the annual increase is fitted to the annual data on new AIDS cases during the pre-cART era.

In the cART era, the cumulative number of cases diagnosed with AIDS are defined as

dt= 5(t)(Y 5 + Y 5 ) import

d(AIDS )diaglocal

where the annual increase is fitted to the annual data on number of cases having AIDS when first diagnosed with HIV.

fitting to the observed and true Number of diagnoses prior to 1996With the widespread implementation of cART in the Netherlands, ATHENA started moni-toring all HIV patients. Patients who died before 1996 are thus not included in our data on annual cases with a new HIV diagnosis. Together with the data on annual AIDS cases in the pre-cART era and the knowledge of disease progression we were however able to infer from the observed number of diagnoses the underlying real diagnosis curve, including the people who had died prior to 1996. From the model, the total cumulative number of diagnoses are defined straightforwardly as


77

dt= s(t)Y s import

dD simport ; (s = 1,…,5)

dt= s(t)Y s local

dD slocal ; (s = 1,…,5)

The probability of still being alive in 1996 per stage by year is shown in Figure 6.

0

0.2

0.4

0.6

0.8

1

1.2

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

s tage 1s tage 2s tage 3s tage 4s tage 5

Year

Pro

babi

lity

of s

urvi

ving

till

1996

figure 6. Probability of surviving until 1996 given disease stage per year.

The diagnosis curves for local and imported cases, excluding cases dying before 1996, and thus comparable to the observed data are then calculated from the cumulative Erlang distribution,

96

Σ j=s

5

; (s = 1,…,5)96Φ s (t) = exp(- t96)( t96 )

5-j

(5-j )!

96t ≡(1996 − t) if t < 1996

0 if t > 1996

Σ Φs (t) (D s (t) − D s (t − 1)) s=1

5

import import importD obs = 96

Σ Φs (t) (D s (t) − D s (t − 1)) s=1

5

import local localD obs = 96

These curves are then fitted to the annual diagnosis cases, separated for acquired abroad and acquired locally respectively. The imported cases initiate the epidemic and are estimated in two time intervals (see Table 1).


78

Average time from Infection to diagnosisThe average time to diagnosis, TD, is calculated as

Σ s=1

4TD = + +

s−11

P

1

1 +s

1 + 1 +

44 1

1 + 1 + +5 +

The average time from infection to diagnosis is calculated for those infected in the correspond-ing time interval, and is thus not the average time to diagnosis for those diagnosed at that time.

maximum Likelihood model fitting and parameter estimationAll the data we fit to – diagnoses, AIDS cases, AIDS diagnoses – are numbers of individuals. To define the likelihood, we assume these are Poisson (i.e. random) distributed around a mean defined by the model. If we define Xi to be the data, and xi(q) the model predictions, dependent on a set of parameters q , then the likelihood can be defined as

L( ) = Xi !i

Xi

Π(xi ( )) exp (-xi( ))

Since some of our data are truncated, we estimate the probability pi that the actual data are included in the database using the “survival until 1996” function ( 96Φ s (t))above. If Yi were the actual number of diagnoses at point i, then the probability of being included in the data is Binomial distributed with probability pi and count Yi and the expected mean for Yi is xi( )/ pi . The likelihood for these truncated observations is thus

Yi ≥ Xi

Yi

piL( ) = Σ

i(pi ) (1− pi )

Xi Yi − Xi

Yi !

Yi(xi( )/pi ) exp (-xi( )/pi )Π

but this is exactly equal to the simple Poisson likelihood defined above, so that data truncation does not need to be accounted for explicitly.

The maximum value the likelihood could ever take is obtained when the model predictions xi( ) are exactly equal to the data values Xi, i.e. the likelihood for the saturated model

Lsat = Xi !i

(Xi ) exp (-Xi )Xi

Π


79

In our estimations, we use for convenience the equivalent deviance measure, given by

Dev( ) = 2 ln(Lmax/L( ))

= 2 (Xi (ln(Xi ) − ln(xi ( ))) + xi ( ) − Xi )

which is minimized to find the best fit parameters. This is equivalent to maximizing the likeli-hood, and gives a number standardized relative to the best possible fit (which would yield Dev = 0).

Confidence intervals are determined by the likelihood ratio method. The deviance is minimized using the custom Optimize function in Berkeley Madonna which uses the relatively robust downhill simplex optimization algorithm to find the minimum. The algorithm was started from numerous starting values to ensure both that the optimization was robust and that a glo-bal minimum had been found.

81

VII transmission networks of HIV-1 among men having sex with men

in the Netherlands2

Daniela Bezemer1, Ard van Sighem1, Vladimir V. Lukashov2, Lia van der Hoek2, Nicole Back2, Rob Schuurman3, Charles A.B. Boucher3,4, Eric C.J. Claas5, Maarten C. Boerlijst6,

Roel A. Coutinho7,8, Frank de Wolf1,9 for the ATHENA observational cohort

Abstract

objective: To obtain insight in the HIV-1 transmission networks among men having sex with men in the Netherlands.design: A phylogenetic tree was constructed from polymerase sequences isolated from 2877 HIV-1 subtype B infected patients monitored in one of the 24 HIV treatment centres in the Netherlands as part of the ATHENA national observational cohort.methods: For men having sex with men with a known date of infection, the most similar sequences were selected as potential transmission pairs when they clustered with bootstrap value ≥ 99%. Time from infection to onward transmission was estimated as the median time between dates of infection for each transmission pair. The source of infections with a resistant strain was traced using the entire phylogenetic tree.results: Of sequences from 403 men having sex with men with a known date of infection between 1987 and 2007, 175 (43%) formed 63 clusters. Median time to onward transmission was 1.4 years (IQR 0.6-2.7). Twenty-four (6%) men having sex with men carried a virus with resistance-related mutations, 13 of these were in 8 clusters together with sequences from 28 other patients in the entire phylogenetic tree. Six clusters contained sequences obtained from 29 men all presenting the same resistance-related mutations.conclusions: Onward transmission of HIV-1 from infected men having sex with men happens both during and after primary infection. Transmission of resistant strains from the antiretro-viral-therapy-treated population is limited, but strains with resistance-related mutations have formed sub-epidemics.

1HIV Monitoring Foundation, Amsterdam, The Netherlands; 2Department of Medical Microbiology, Academic Medical Centre, Amsterdam, The Netherlands; 3Department of Virology, University Hospital Utrecht, The Neth-erlands; 4Department of Virology, Erasmus Medical Center, Rotterdam, The Netherlands; 5Department of Medical Microbiology, Leiden University Medical Center, The Netherlands; 6Institute for Biodiversity and Ecosystem Dy-namics, University of Amsterdam, The Netherlands; 7Department of Internal Medicine, Academic Medical Center, Amsterdam, The Netherlands; 8Center for Infectious Disease Control, National Institute of Public Health and the Environment, Bilthoven, The Netherlands; 9Department of Infectious Disease Epidemiology, Imperial College Lon-don, United Kingdom


82

Introduction

Despite the success of combination antiretroviral treatment (cART) in reducing viral load, HIV-1 transmission continues among men having sex with men (MSM) in industrialised coun-tries, including the Netherlands [1, 2]. Using a mathematical model to describe trends in the transmission dynamics of HIV-1 amongst MSM in the Netherlands, we recently estimated that 90% of onward transmission in this risk group takes place from the undiagnosed group. The average time from infection to diagnosis was 2.7 years [1]. However, the impact of primary infection on onward transmission remains unclear. As usually no diagnostic tests are performed during this initial phase of the infection when the viral load peaks and infectiousness is high [3, 4], the rate of partner change may be crucial for epidemic spread [5].

Despite discrepancies, HIV sequence analysis can reveal information on contact networks [6]. Based on the phylogenetic clustering of HIV-1 polymerase (pol) sequences obtained from pri-mary infections, previous studies suggested that 25% to 50% of transmissions among MSM take place during primary infection [7-9]. However, clustering of sequences obtained from primary infections does not necessarily represent transmission during the period of primary infection [10]. Lewis et al applied a Bayesian Monte Carlo Markov Chain method on sequences obtained from 402 patients without a known date of infection, and estimated that 25% of transmissions took place within the first 6 months of infection [11]. But in this study assump-tions needed to be made to estimate the dates of infection. We estimated the median time between infection and onward transmission for potential transmission pairs selected from a phylogenetic tree of HIV-1 subtype B pol sequences obtained from 403 MSM shortly after their known date of infection.

Further insight in transmission dynamics of HIV-1 can be obtained by investigating the trans-mission networks of strains with resistance-related mutations [12]. Transmission soon after infection facilitates transmission of resistant HIV-1 strains, which would revert to the original wild type within a few months in absence of antiretroviral therapy [8, 13-15]. Certain resist-ance-related mutations revert to a new wild type, a process that is well documented for amino acid position 215 in reverse transcriptase (RT) [13, 16]. In previous studies, we found that 6% of new infections presented resistance-associated mutations, and that 23% of ART-naive patients failing cART were infected with a resistant strain [17, 18]. Several studies reported phylogenetic clustering of resistant strains obtained from ART-naive patients [9, 15, 19]. We used the set of HIV-1 subtype B pol sequences obtained from 403 MSM with a known date of infection to monitor the transmission of resistant stains over calendar time, and linked these resistant strains to the source of infection in a phylogenetic tree of all 4090 HIV-1 subtype B pol gene sequences obtained from 2877 infected patients, both before and after starting ART. All patients were monitored in one of the 24 HIV treatment centres in the Netherlands as part of the ATHENA national observational cohort.

vII transmIssIon networks of hIv-1 among men havIng sex wIth men In the netherlands

83

methods

databaseThe ATHENA cohort encompasses all patients infected with HIV-1 followed longitudinally in one of the 24 HIV treatment centres in the Netherlands [20]. Demographic data were collected at entry in the cohort. At each follow-up visit, clinical, virological, and immunological data were collected, as well as data on the use of cART [1, 18, 21]. HIV-1 pol gene sequences were obtained as part of the screening for resistance to antiretroviral drugs, both before and during treatment with cART [22-24].

sequencesPatients were eligible for this study if at least one pol sequence was available containing at least the first 251 amino acids of the RT gene. Population-based nucleotide sequencing of the HIV-1 pol gene was performed as described in detail previously [17]. Sample contamination was checked for at the respective sequencing sites. Multiple sequence alignment was done by hand and using the default parameters of the ClustalX 1.83 program. Subtype B was identified by phylogenetic analysis of combined RT and protease sequences, using reference sequences from the Los Alamos database [25] and our own database. The percentage of ambiguous sites was estimated for all sequences. Sequences were screened for major resistance-conferring mutations at the amino acid positions described by the International AIDS Society-USA, including alter-native substitutions at position 215 [26].

New HIV-1 infectionsFor this study, new HIV-1 infections were defined as those infections with either a seroconver-sion interval of ≤18 months between the last negative and the first positive HIV-1 serology test, or documented evidence of a primary infection. A diagnosis of primary HIV-1 infection was defined as detectable HIV-1 RNA and/or detectable serum p24 antigen in plasma combined with either a negative HIV-1 antibody-test or a positive HIV-1 antibody- test with a negative, incomplete, or indeterminate HIV-1 Western Blot. The estimated date of infection was defined as the midpoint between the last seronegative and the first seropositive sample, or the date of the last seronegative but RNA-positive sample, or the date of an indeterminate result on the Western blot [17]. Sequences corresponding to new infections were obtained from a sample taken ≤18 months after the estimated date of seroconversion.

distance methodThe percentage pairwise sequence distances between all available entire protease sequences and RT sequences cut to equal lengths of 251 amino acids were calculated taking into account ambiguous sites according to a mixed weighted distance method [27]. To study the level of mixing of MSM HIV-1 transmission networks with transmission via other routes, we used a


84

pairwise sequence distance ≤1.5% between RT sequences as a selection criterion for potential transmission pairs [18, 28]. Transmission clusters were defined as groups of linked potential transmission pairs.

phylogenetic analysisPhylogenetic trees were constructed of pol gene RNA sequences that included at least the first 251 amino acids of the RT gene. Trees were constructed with the Neighbor-Joining method [29, 30] within the MEGA program [31], and ambiguous sites were ignored. To prevent false clus-tering due to convergent evolution, 36 amino acid sites associated with major drug-resistance were excluded [26]. When available, multiple sequences per person were included. All trees were rooted against an HIV-1 subtype K sequence (Los Alamos Database accession number AJ249235).

potential transmission pairsThe first phylogenetic tree included all HIV-1 subtype B pol sequences obtained from therapy-naïve MSM within 18 months after their estimated date of infection. The Maximum Composite Likelihood method was used to compute evolutionary distances, and a bootstrap analysis with 1000 replications was performed. From this tree, a selection was made of patients in clusters with a bootstrap value ≥99% [18, 28, 32]. Each patient in this selection was combined with the other patient in the same cluster with the smallest pairwise sequence distance to form the most likely transmission pair. The time between infection and onward transmission was estimated as the median time between dates of infection of all these potential transmission pairs. The correlation between the pairwise sequence distance and time between the dates of infection of transmission pairs was estimated from linear regressions.

tracing the source of resistant strainsIn order to trace the infecting source of the transmitted resistant strains with a known date of infection, a second phylogenetic tree was made that contained all subtype B sequences avail-able in the ATHENA database. For the construction of this tree, the Kimura 2-parameter model was used [33], and a bootstrap analysis with 100 replications was performed. The clusters observed were confirmed in a smaller tree of all sequences that clustered with a resistant strain with a known date of infection. Evolutionary distances in this smaller tree were computed using the Maximum Composite Likelihood method with a gamma distribution of the substitu-tion rate with shape parameter 1.0 [34], and 1000 bootstrap replications were performed. The clusters were studied with the aim to identify whether patients on treatment transmitted the resistant strain.


85

results

patient characteristicsBy June 2007, 12,951 persons infected with HIV-1 were included in the ATHENA national observational cohort; 6845 (53%) were reported to have been infected by MSM contact. In total, 4090 HIV-1 subtype B pol gene sequences were obtained from 2877 persons, of whom 2022 (70%) were MSM, 486 (17%) were infected via heterosexual contact, 167 (6%) via drug injection, and 202 (7%) via other or unknown transmission routes. The first sequence of each patient was obtained between 1987 and 1996 for 101 (3.5%) patients, between 1996 and 2000 for 503 (17.5%), between 2001 and 2004 for 1114 (38.7%), and after 2004 for 1159 (40.3%) patients. Of 832 patients who had a non-B HIV-1 infection and a sequence available, only 68 (8%) were MSM.

transmission of HIV-1 between msm and other risk groupsHIV-1 subtype B RT sequences from 817 (28%) out of 2877 persons clustered within a sequence distance of 1.5% from another person’s RT sequence. Of these 817 patients, 603 (74%) were MSM, 91 (11%) were heterosexuals, 51 (6%) were infected via injection drug use, and 72 (9%) were infected via other or unknown transmission routes. Only 8% of MSM had an infection with HIV-1 that was most similar to that found in patients from other risk-groups.

table 1. Characteristics of 403 MSM identified with an HIV-1 subtype B infection with a known date of infection in the period 1987 – 2007.

N 403

Median age at estimated time of infection, in years 34.8 (IQR 29.9 – 41.7) (range 21.0 – 61.9)

Percentage born in The Netherlands 81

Number of acute infections 107

Median interval between the last antibody-negative and the first RNA positive visit, in months

6.4 (IQR 4.9 – 9.9), for 296 seroconverters

Median interval between the estimated date of sero-conversion and the sequenced sample, in months

3.3 (IQR 1.3 – 6.2)

Median interval between the first RNA-positive visit and the sequenced sample, in months

0.4 (IQR 0.0 – 1.4)

Median plasma HIV-1 RNA concentration at first RNA positive sample (copies/ml)

78495 (IQR 16976 – 250729)

Median CD4 count at diagnosis (106 cells/l) 0.49 (IQR 0.36 – 0.67)

Percentage ambiguous sites 0.08 (IQR 0.0 – 0.4)

Median percentage pairwise nucleotide sequence distance of 403 new infections’ pol sequences

5.5% (IQR 4.9 – 6.1; range 0.0 – 9.7)


86

New infectionsFrom 403 (20%) of the 2022 MSM with a subtype B pol sequence, the approximate date of HIV-1 antibody seroconversion was known (Table 1). The number of new infections by year of seroconversion is shown in Figure 1A. For 393 MSM, the pol gene was sequenced including both protease and RT, whereas for 10 MSM only RT was available. The median percentage of ambig-uous sites among the 403 sequences was low (0.08%, interquartile range [IQR] 0.0 – 0.4), which allowed for calculation of pairwise sequence distances either including or excluding these sites. Of the 403 MSM, 292 (72%) were monitored in an HIV treatment centre in Amsterdam, and 294 (73%) reported that they were most likely infected in the Netherlands. Of the 42 MSM infected before 1996, 64% were prospectively identified within the Amsterdam Cohort Studies [17].

clustering among new HIV-1 subtype B infectionsWe constructed a phylogenetic tree of 499 available pol sequences from 403 MSM with at least one RT sequence within 18 months of their estimated date of infection. The tree showed 63 transmission clusters with a bootstrap value ≥99%, including 175 (43%) patients (Figure 2A). Both HIV-1 protease and RT sequences were obtained for all 175 patients. The clusters were confirmed in a separate phylogenetic tree containing sequences of the 393 MSM with both protease and RT sequenced. The size of the clusters varied from 2 to 8 patients (median 3, IQR 2 – 4). The median minimum number of nucleotide substitutions per site per person within each cluster was low (0.0035, IQR 0.0014 – 0.0065; range 0 – 0.0158), indicating clustering of potential transmission pairs [28]. The time between the two most distant dates of infection in each cluster ranged from 0.05 to 9.7 years (median 2.0, IQR 0.9-4.2). The median difference in time between the dates of infection of the most likely transmission pair for all patients in a clus-ter was 1.4 years (IQR 0.6-2.7, range 0.03 – 9.05). The corresponding median pairwise sequence distance was 0.9% (IQR 0.4 – 1.5), and increased by 0.33% (95% confidence interval [CI] 0.28-0.38; p<0.0001) per year of separation in time. The distribution of the median difference in time between the dates of infection of the most likely transmission pair for all patients in a cluster is shown in Figure 2B. Constraining the analysis to transmission pairs with a pairwise sequence distance ≤1.5% or a synonymous sequence distance ≤4.5%, showed that the median difference in time varied between 1.1 and 1.4 years (Table 2, #2-7). Undetected transmissions related to individuals with unknown date of infection do not impact these estimates (Table 2, #9).

transmission of resistant HIV-1 subtype B strainsFigure 1A shows the annual percentage of infections with a resistant virus strain among all 403 MSM with a new HIV-1 subtype B infection. In total, 24 patients (6.0%, 95% CI 3.8 – 8.7%) were infected with an antiretroviral drug-resistant strain. Of these 24 patients, 18 (4.5%) had one or more mutations associated with resistance to nucleoside RT inhibitors, 2 (0.5%) patients were resistant to non-nucleoside RT inhibitors, 2 (0.5%) to a protease inhibitor, and 2 (0.5%) to two or three drug classes. Fourteen of 24 resistant strains (58%) had a mutation at position


87

0

5

10

15

20

25

199410

19959

199612

199713

19985

199918

200013

200125

200233

200347

200458

200574

200659

year seroconversion n

perc

enta

ge

m215r215

0

5

10

15

20

25

30

35

19871

19891

19908

19916

19924

19933

199410

19959

199612

199713

19985

199918

200013

200125

200233

200347

200458

200574

200659

20074

year seroconversionn

perc

enta

ge pimdnnrtnrt

5 – 45

5 – 28

0 – 11

2 - 14

0 - 60 - 4

2 - 8

8 - 16

1 - 6

A

B

figure 1. Annual proportion of sequences with resistance-associated mutations among 403 MSM with a new HIV-1 subtype B infection.A. Percentage of patients resistant to antiretroviral drugs from each drug class. NRT: nucleoside RT inhibitor, NNRTI: non-nucleoside RT inhibitor, PI: protease inhibitor, MD: Multidrug. The 95% confidence intervals are given above each bar. The annual number of infections is denoted below the calendar year on the horizontal axis. Eight of 10 patients with only a RT sequence were infected in 1994 and earlier; the other two were infected in 1997 and 2002, thus only in years when no resistant mutations were found in the available protease sequences of new infections. Thus, the percentages in Figure 2 are correct, even though the corresponding number of new infections (n) in these years does not refer to the number of people with a complete corresponding pol sequence.B. Percentage of patients infected with a strain harbouring a resistance-associated mutation (m215) or a revertant mutation (r215) at position 215 in RT.

215 in RT. Figure 1B shows the annual percentage of infections with HIV-1 strains with a resistance-conferring mutation at RT position 215, separated into 215 resistant mutations (215 Y or F) and 215 revertant mutations (215 C, D, E, or S). After 1996, only revertant mutations were found at position 215.


88

0.0140

0

5

10

15

20

25

3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 69 72 84 90 93 105108

months

% t

rans

mis

sion

B

A

figure 2. Transmission clusters of HIV-1 strains obtained from MSM with a known date of infectionA. Phylogenetic tree of HIV-1 subtype B polymerase sequences belonging to 403 MSM patients with an identified new HIV-1 subtype B infection. The significant transmission clusters with bootstrap values ≥99 are shown in red. Blue denotes the reference subtype K sequence. Phylogenetic analyses were conducted in MEGA4.B. The distribution in time between the dates of infections of the most likely transmission pair for all patients in a significant cluster. Light grey bars represent the percentage per category of 3 months, and dark grey bars per 6 months.


89

Networks of transmitted resistant HIV-1 subtype B strainsBased on a phylogenetic analysis of all 4090 HIV-1 subtype B pol sequences in the ATHENA database, we selected sequences from 88 persons that clustered with the sequences obtained from the 24 MSM who were newly infected with a resistant HIV-1 strain. Phylogenetic analysis of this subset of 152 sequences of 112 patients showed 8 significant clusters (Figure 3). The 8 clusters contained sequences of 13 (54%) of the 24 newly infected patients who had a resistant strain, including 7 with a bootstrap value ≥99, and 1 with a bootstrap value of 95 but a pairwise sequence distance of the closest sequence pair of only 1.0%. The same clusters were also iden-tified when clusters where based on connections between potential transmission pairs with a mixed weighted distance ≤1.5% at RT, except for 3 patients (M24, M31, and M54). Sequences from the other 11 patients infected with a resistant strain did not cluster significantly with other

table 2. Time interval between dates of infection corresponding with pol sequences from different patients with minimum distance to each other in various subsets of 403 new infections amongst MSM.

# Data selectiondistance between sequences (%)

time difference between dates of infection (years)* N rate ** P

median (IQR) median (IQR)1 All 2.3 (1.0- 3.5) 2.9 (1.1 – 7.3) 403 0.36 <0.0001

2 In cluster with bootstrap ≥99% 0.9 (0.5 – 1.5) 1.4 (0.6 – 2.7) 175 0.33 <0.0001

3 Distance to closest sequence ≤1.5% 0.7 (0.4 – 1.1) 1.2 (0.5 – 2.0 ) 157 0.27 <0.0001

4 Distance to closest sequence ≤ 1.5% and in cluster with bootstrap ≥99%

0.7 (0.3 – 1.1) 1.1 (0.5 – 2.0 ) 133 0.25 <0.0001

5 Synonymous distance to closest sequence in cluster with bootstrap ≥99%

2.8 (0.9 – 4.5)*** 1.4 (0.5 – 2.9) 175 0.92*** <0.0001

6 Synonymous distance to closest sequence ≤4.5%

0.7 (0.4 – 1.1) 1.2 (0.5 – 2.0 ) 145 0.25 <0.0001

7 Synonymous distance to closest sequence ≤ 4.5% and in cluster with bootstrap ≥99%

0.7 (0.3 – 1.0) 1.2 (0.5 – 2.1 ) 129 0.24 <0.0001

8 Distance to closest sequence of any infection in the database

1.4 (0.7 – 2.7) 2.1 (0.7 – 5.7) * 403 0.30 <0.0001

9 Distance to closest sequence of any infection in the database ≤1.5%

0.7 (0. 4 – 1.1) 1.2 (0.4 – 2.4) * 220 0.22 <0.0001

10 Any infection in the database clos-est to any infection in the database ≤1.5% for RT only

0.8 (0.5 – 1.2) **** 1.4 (0.4 – 3.0) * 817 0.16**** <0.0001

* Absolute time difference between estimated dates of infection, except for patients in selection 8, 9 and 10 for whom it represents the absolute time between the dates of first HIV diagnosis;** rate in % distance / year;*** Here the % sequence distance refers to the synonymous distance only;**** sequence distance refers to distance between RT sequences only; P is the correlation between distance and time; IQR: inter-quartile range.


90

sequences in the database. One of them (M30), appeared to be infected with an HIV-1 strain resistant to three drug classes and reported the possibility of having been infected by someone from outside the Netherlands. Both M33 (infected in 1994) and M34 (infected in 2002) might have been infected by someone on treatment as their HIV strains contain the RT mutations 70R and 184V, respectively, known for mutating back to wild type in the absence of antiviral therapy [13].

figure 3. transmission networks of HIV-1 strains with resistance-related mutations.A phylogenetic tree constructed of HIV-1 subtype B polymerase sequences from 24 new resistant infections among MSM and 88 infections selected from all 2877 subtype B infections. Analyses were conducted in MEGA4. New infections are coded as M#-RCYEAR, which codifies respectively; M = MSM; # = unique number; R = resistant; C = the code to which phase the sequences belong: P = primary infection, S = seroconverter, N = therapy-naïve, T = during treatment, I = during treatment interruption; and YEAR = the estimated calendar year of infection. (N)NRT = (non-) nucleoside reverse transcriptase; PR = protease; µ = amino acid mutation conferring antiretroviral-treatment resistance. Undated infections are coded similarly but without YEAR. H instead of M represents heterosexual transmission. In the tree, the 24 new resistant infections from Figure 1 are represented in bold, and resistance-conferring mutations are given either in the table when part of significant cluster, or following the branch name in the tree. In the table, the dates (SEQUENCE) and mutations (MUTATION) of the clustering sequences are specified, as well as the dates of the first HIV-positive test, the dates of the first ART, and the dates of the first cART regimen of the patients. Patients selected from Figure 1 are indicated in yellow, and the dates of sequences corresponding to new infections are shown in yellow. Sequences labelled ‘RT only’ in the first column do cluster, but they are not part of the phylogenetic tree presented.


91

100

100

100

100

100

100

100 100

95

100

100

100

82

73

97

65

80

57

27

90

55

49

98

95

93

100 88

65

100

100

43

77

66

100 52

83

96

99 94

96

76

8994

87

100

100

0.005

4

H3-RS2005 M4-RN2004

M5-RN2004

M7-RS2005 M4-RT2004 M8-RN

M9-RT M9-RT

M9-RT M9-RT

M10-RT M10-RT M10-RT

M11-RT1989

M16-RN2001

M19-RT M20-N

M21-N

M23-N2005 M24-N

M25-N M26-S2004

M27-S2005

M30-RN2004

M31-S2001 M32-RN2001

M36-RT1996

M38-RN

M40-RN M40-RT

M41-N

M44-RN M45-RN

M46-RN M47-RN

M48-N M49-I

M50-N M51-RT

M51-RT M51-RN

M52-S1989 M52-RT1989

M59-RN

M53-RN1994 M53-RT1994

M54-RT M55-RN1996

M57-RN M58-RN2003

K AJ249235

M15-N2005

M28-RS1999

M29-RS2005

M35-RS2005

M1-RS2006 M2-RS2005

M6-RS2004

M12-RP2005

M13-RP1994

M14-RS2005

M17-RS2001

M18-RS2005

M22-RP2006

M30-RS2004

M32-RS2001

M34-RS2002

M36-RP

M37-RS2003

M39-RS2005

M42-RS2005 M43-RS2005

M55-RS1996

M56-RS2004

M33-RP1994

M53-RS1994

I

II

III

CLUSTER PATIENT HIV+ TEST FIRST ART FIRST cART SEQUENCE MUTATIONS DRUG CLASS

M9 15-3-1992 5-1-1993 20-7-1998 3-4-2000 70R 184 215 219E

46 82A NRT

R

IM9 15-3-1992 5-1-1993 20-7-1998 24-9-2001 70R 215S/ 219E

103N NRT

I M9

15-3-1992 5-1-1993 20-7-1998 3-10-2001 70R 215 219E

190A NRT

NNRT

I M9

15-3-1992 5-1-1993 20-7-1998 22-5-2003 70S/R 215 219E

181 NRT

NNRT I M6 15-3-2005 15-8-2006 15-8-2006 15-3-2005 215 219E NRT I M4 18-4-2005 14-11-2006 14-11-2006 26-4-2005 215 219E NRT

I M4 18-4-2005 14-11-2006 14-11-2006 25-4-2007 70R 184 215 219E

103N NRT

NNRT I H3 1-7-2005 13-7-2005 215 219E NRT I M2 13-9-2005 4-10-2005 215 219E NRT I M7 11-10-2005 18-12-2006 18-12-2006 4-11-2005 215 219E NRT I M5 28-12-2005 2-2-2006 215 219E NRT I M1 13-4-2006 13-4-2006 215 219E NRT I M8 9-1-2007 10-1-2007 215 219E NRT RT only M11 5-12-1989 8-10-1996 31-1-1997 8-10-1996 I b M11 5-12-1989 8-10-1996 31-1-1997 24-12-1996 184 NRT

I b M10 21-7-1992 21-5-1996 26-9-1996 3-4-2000

23-9-2002 24-6-2004

67N 184 210 215Y 219E 46I 84 90M

NRT R

M16 21-6-2002 13-10-2002 13-10-2002 9-10-2002 219 NRT M14 7-3-2005 24-3-2005 219 NRT M15 7-11-2006 7-11-2006

M19 1-7-1992 18-8-1997 18-8-1997 12-7-2001 54 84 R M20 3-3-2003 9-2-2004 9-2-2004 21-3-2003 M25 15-8-2004 3-9-2004

M21 27-8-2004 8-9-2004 M26 11-10-2004 25-10-2004 M24 15-11-2005 5-12-2005 M27

15-3-2006 12-5-2006 M23 26-4-2006 10-5-2006 M22 17-8-2006 21-8-2006 17-8-2006 33F R

M31 27-6-2001 30-3-2006 30-3-2006 27-6-2001 M32 15-8-2001 13-6-2002 210 NRT M32 15-8-2001 8-10-2002 210 NRT

M37 17-2-2004 3-6-2004 215S NRT M41 15-4-2005 31-1-2007 215S NRT M45 15-6-2005 2-8-2005 2-8-2005 13-7-2005 215S NRT M43 30-8-2005 30-8-2005 215S NRT M40 15-11-2005 14-2-2006 14-2-2006 24-1-2006 215S NRT M40 15-11-2005 14-2-2006 14-2-2006 22-12-2006 184 215S NRT M39 28-12-2005 1-2-2006 215S NRT M44 15-5-2006 1-6-2006 215S NRT M42 15-5-2006 21-8-2006 21-8-2006 7-8-2006 215S NRT M46 13-11-2006 4-12-2006 215S NRT M38 15-12-2006 17-1-2007 215S NRT

M49 1-7-1996 1-7-1996 1-7-1996 7-1-1999 M48 28-4-1999 15-9-1999 15-9-1999 14-9-1999

M50 19-7-1999 12-8-1999 12-8-1999 12-8-1999 M51 8-7-1998 17-8-1998 17-8-1998 12-8-1998 108I NNRT

M51 8-7-1998 17-8-1998 17-8-1998 21-2-2000

184 108I 30N

NRT NNRT

R

M51 8-7-1998 17-8-1998 17-8-1998 2-7-2001

184 103N 108I

30N

NRT NNRT

R M47 11-11-2005 23-8-2006 215T/S NRT

M52 16-1-1990 30-5-1990 4-4-1996 16-1-1990 M52 16-1-1990 30-5-1990 4-4-1996 22-8-1995 41 67N 70R 184 215Y NRT

RT only M53 22-3-1995 27-10-2003 27-10-2003

22-3-1995 8-5-1995

15-1-1996 41 215Y

NRT

M53 22-3-1995 27-10-2003 27-10-2003 3-11-1998

2-9-2002 41 215 NRT

M53 22-3-1995 27-10-2003 27-10-2003 28-6-2004 41 65R 184 215

181 NRT

NNRT

M54 28-12-1994 13-3-1995 31-7-1996 2-11-2001 67N 70R 184 215F 219

48 82M 90M NRT

R M55 11-11-1996 27-5-1997 27-5-1997 11-11-1996 67N 70R 215F 219 NRT

M55 11-11-1996 27-5-1997 27-5-1997 12-5-1997 67N/ 70R/K 215F/ 219 NRT

M58 26-8-2003 22-11-2006 215 NRT M57 27-10-2003 28-10-2003 215 NRT M56 31-1-2005 31-1-2005 215 NRT M59 8-11-2006 22-11-2006 215 NRT

I b

I

IIIIII

IIIIIIIII

III

III

IIIIII

IIIIII

27-6-2001 13-6-2002 8 -10-2002


92

Twenty-one persons in three clusters had sequences with a revertant mutation at position 215 in RT, specifically 215C in cluster 1 and 8, and 215S in cluster 5. The mutant strain with 215C and 219E at RT in cluster 1 might have been circulating for a period of 15 years, i.e. the period between the earliest and the latest date of diagnosis in the cluster. No pre-treatment HIV-1 RT sequence was available from the earliest diagnosed patient (M9) in this cluster. However, the sequence obtained from M9 at the time of therapy failure harboured the same 215 revertant mutant as the other sequences in the cluster. A pre-treatment RT sequence from patient M11 diagnosed in 1989 that was linked to cluster 1 with a bootstrap value of 80 did not show any resistant mutation. Cluster 2 was linked to one other antiretroviral-naive patient (M16) infected with a strain that harboured a 219Q mutation. In cluster 3, the resistance-associated mutation 33F in HIV-1 protease obtained from patient M22 was not present in any of the other 8 persons in the same cluster, and was possibly a natural polymorphism [35]. Cluster 4 was neither linked to other persons with a sequence with a 210W mutation in RT, nor to persons on cART at the time of infection. Clusters 6 and 7 possibly show a direct transmission from someone failing treatment with a 215Y mutation in cluster 6, and a 215F mutation in cluster 7. In cluster 6, the selection of a resistant strain in the potential initial source failing therapy (M52) was observed.

discussion

Our study on transmission networks of HIV-1 subtype B among MSM in the Netherlands indicates that 25% of onward transmissions occur within 7 months after infection, half of transmissions within 17 months, and 75% within 2.7 years. This finding is compatible with our previous analysis of the dynamics of the HIV-1 epidemic among MSM in the Netherlands, in which we estimated that individuals who were unaware of their infection were the source of 90% of new infections, with an average of 2.7 years between infection and diagnosis for those infected after 2000 [1]. Our estimate of the median time between transmissions, for which we used only sequences corresponding to infections with an approximate known date of infection, is in agreement with a study by Lewis et al, who estimated the time between the nodes in a tree of sequences with an unknown date of infection [11]. They reported an estimated 25% of transmissions taking place within the first 6 months of infection and 50% within 14 months after infection. This might indicate that the HIV-1 transmission dynamics amongst MSM in the Netherlands and the United Kingdom are similar.

However, phylogenetic studies on transmission dynamics have shortcomings. Discrepancies have been demonstrated to arise when comparing the viral phylogeny with known sexual-con-tact networks [6]. Our study contained a subset of infections with an estimated date of infection that clustered and formed likely transmission pairs, although it cannot be excluded that there were intermediate transmissions. In addition, distinguishing people who infected more than


93

one person is not always feasible, and individuals that were infected by a common source within a short time period might cluster as a transmission pair. Furthermore, transmission dynamics are known to vary over calendar time [1], yet we estimated an average transmission rate. The search for new infections was intensified in the later years of our study, resulting in more avail-able sequences, and thus the identification of more transmission pairs in these years. In addi-tion, a significant proportion of patients identified as newly infected were familiar with their HIV-positive status after their first positive test, and thus soon after infection. Consequently, their behaviour may differ from those diagnosed at a later stage of infection.

In several countries, the overall transmission of resistant HIV-1 strains is reported to have decreased since the introduction of cART in 1996 [17, 36-38]. This has been explained by the efficacy of cART and the lower transmission potential of resistant HIV-1 strains [8, 9, 13, 39-41]. We found that 6% of HIV-1 subtype B strains in new infections among MSM had resistance-related mutations. Tracing the source of these resistant strains showed clusters with mainly transmission of HIV-1 with a revertant mutation at position 215 of RT. These observa-tions are compatible with the estimate that the 215 RT revertant mutations have no significant fitness effect on the fitness of the virus [13, 16]. In contrast to an incidence of 20% among new transmissions in 1994, the 215 mutants were not found after the introduction of cART in 1996 until 2003, except for one case in 1999 [40]. However, from 2003 onwards, revertant mutants were present among new infections in all subsequent years. This could be due to the introduction of baseline sequencing around that time, but it might also be a result of chang-ing transmission dynamics. Previously, we estimated that transmission decreased in the early cART era, but a resurgence of the epidemic occurred in later years [1]. Thus the reappearance of the revertant 215 RT mutation among recent transmissions may be a sentinel for changing transmission dynamics [42, 43]. Additionally, the HIV-1 incidence is increasing among older MSM, suggesting the possibility of a re-opened reservoir [44]. However, the initial resistant mutations at position 215 were only observed before the introduction of cART, which reflects the contribution of patients failing monothearpy.

Sampling methods might influence the monitoring of transmitted resistance since viruses from failing patients are often sequenced retrospectively. Contact tracing on an individual basis might also have an impact as recent partners of a person are identified. The cluster of sequences obtained that way could have an impact on the percentage of resistance found in the respective year of infection (whether or not with a resistant strain) that was larger than expected when sampling was random. We reduced these effects by only selecting infections with a known date of infection.

In conclusion, our study indicates that onward HIV-1 transmission from infected MSM takes place both during and after primary infection. Strains with resistance-related mutations have


94

formed sub-epidemics, and transmission of resistant strains from the antiretroviral-treated pop-ulation is limited. However, with the current changes in risk behaviour among people using cART [45-47] transmission from the treated population might increase. Intensifying contact tracing and facilitating frequent testing could help to identify people earlier in their infection and prevent onward transmission.

Acknowledgements

DB was supported by grant 7014 from AIDS Fund Netherlands. The authors thank Maus Sabe-lis for useful discussions, Maria Prins for helping in identifying new infections, Luuk Gras for statistical assistance, and Sally Ebeling for English revision. The ATHENA national observa-tional cohort is a collaboration between 24 HIV treatment centres in the Netherlands: Treating physicians (*Site coordinating physicians) Academisch Medisch Centrum bij de Universiteit van Amsterdam – Amsterdam: Dr. J.M. Prins*, Drs. J.C. Bos, Dr. J.K.M. Eeftinck-Schattenkerk, Dr. S.E. Geerlings, Dr. M.H. Godfried, Prof. dr. J.M.A. Lange, Dr. J.T.M. van der Meer, Dr. F.J.B. Nellen, Drs. D.P. Olszyna, Dr. T. van der Poll, Prof. dr. P. Reiss, Drs. S.U.C. Sankats-ing, Drs. R. Steingrover, Drs. M. van der Valk, Drs. J.N. Vermeulen, Drs. S.M.E. Vrouenraets, Dr. M. van Vugt, Dr. F.W.M.N. Wit. Academisch Ziekenhuis Maastricht – Maastricht: Dr. G. Schreij*, Dr. S. van der Geest, Dr. A. Oude Lashof, Dr. S. Lowe, Dr. A. Verbon. Catharina Ziekenhuis – Eindhoven: Dr. B. Bravenboer*, Drs. M.J.H. Pronk. Emma Kinderziekenhuis – AMC Amsterdam: Prof. dr. T.W. Kuijpers, Drs. D. Pajkrt, Dr. H.J. Scherpbier. Erasmus MC – Rotterdam: Dr. M.E. van der Ende*, Drs. H. Bax, Drs. M. van der Feltz, Dr. L.B.S. Gelinck, Drs. Mendoca de Melo (until September 1, 2008), Dr. J.L. Nouwen, Dr. B.J.A. Rijnders, Dr. E.D. de Ruiter, Dr. L. Slobbe, D rs. C.A.M. Schurink, Dr. T.E.M.S. de Vries. Erasmus MC – Sophia – Rotterdam: Dr. G. Driessen, Dr. M. van der Flier, Dr. N.G. Hartwig. Flevoziekenhuis – Almere: Dr. J. Branger. Haga Ziekenhuis, locatie Leyenburg – Den Haag: Dr. R.H. Kauff-mann*, Drs. K. Pogány (until August 1, 2008), Dr. E.F. Schippers (from May 1, 2008). Isala Klinieken – Zwolle: Dr. P.H.P. Groeneveld*, Dr. M.A. Alleman. Kennemer Gasthuis – Haar-lem: Prof. dr. R.W. ten Kate*, Dr. R. Soetekouw. Leids Universitair Medisch Centrum – Lei-den: Dr. F.P. Kroon*, Dr. S.M. Arend, Drs. M.G.J. de Boer, Prof. dr. P.J. van den Broek, Prof. dr. J.T. van Dissel, Drs. C. van Nieuwkoop. Maasstadziekenhuis – locatie Clara – Rotterdam: Dr. J.G. den Hollander*. Medisch Centrum Alkmaar – Alkmaar: Dr. W. Bronsveld*. Medisch Centrum Haaglanden -locatie Westeinde – Den Haag: Dr. R. Vriesendorp*, Dr. F.J.F. Jeuris-sen, Dr. E.M.S. Leyten. Medisch Centrum Leeuwarden – Leeuwarden: Dr. D. van Houte*, Dr. M.B. Polée. Medisch Spectrum Twent e – Enschede: Dr. C.H.H. ten Napel*, Dr. G.J. Kootstra. Onze Lieve Vrouwe Gasthuis – Amsterdam: Prof. dr. K. Brinkman*, Drs. G.E.L. van den Berk, Dr. W.L. Blok, Dr. P.H.J. Frissen, Drs. W.E.M. Schouten.St. Medisch Centrum Jan van Goyen – Amsterdam: Dr. A. van Eeden*, Dr. D.W.M. Verhagen. Slotervaart Ziekenhuis – Amsterdam:


95

Dr. J.W. Mulder*, Dr. E.C.M. van Gorp, Dr. A.T.A. Mairuhu, Drs. R. Steingrover, Dr. J. Wagenaar. St. Elisabeth Ziekenhuis – Tilburg: Dr. J.R. Juttmann*, Dr. M.E.E. van Kasteren. St. Lucas Andreas Ziekenhuis – Amsterdam: Dr. J. Veenstra*, Dr. W.L.E. Vasmel. Universitair Medisch Centrum St. Radboud – Nijmegen: Dr. P.P. Koopmans*, Drs. A.M. Brouwer, Dr. A.S.M. Dofferhoff, Prof. dr. R. de Groot, Drs. H.J.M. ter Hofstede, Dr. M. Keuter, Dr. A.J.A.M. van der Ven. Universitair Medisch Centrum Groningen – Groningen: Dr. H.G. Sprenger*, Dr. S. van Assen, Dr. J.T.M. van Leeuwen, Dr. C.J. Stek. Universitair Medisch Centrum Gro-ningen – Beatrix Kliniek – Groningen: Dr. R. Doedens, Dr. E.H. Scholvinck. Universitair Medisch Centrum Utrecht – Utrecht: Prof. dr. I.M. Hoepelman*, Dr. M.M.E. Schneider, Prof. dr. M.J.M. Bonten, Dr. P.M. Ellerbroek, Drs. C.A.J.J. Jaspers, Drs. L.J. Maarschalk-Ellerbroek, Dr. J.J. Oosterheert, Dr. E.J.G. Peters, Dr. T. Mudrikova, Drs. M.W.M. Wassenberg, Dr. S. Weijer. WilhelminaKinderziekenhuis – UMC Utrecht: Dr. S.P.M. Geelen, Dr. T.F.W. Wolfs. VU Medisch Centrum – Amsterdam: Prof. dr. S.A. Danner*, Dr. M.A. van Agtmael, Drs. W.F.W. Bierman, Drs. F.A.P. Claessen, Drs. M.E. Hillebrand, Drs. E.V. de Jong, Drs. W. Kort-mann, Dr. R.M. Perenboom, Drs. E.A. bij de Vaate. Ziekenhuis Rijnstate – Arnhem: Dr. C. Richter*, Drs. J. van der Berg, Dr. E.H. Gisolf. Ziekenhuis Walcheren – Vlissingen: Dr. A.A. Tanis*. St. Elisabeth Hospitaal/Stichting Rode Kruis Bloedbank – Willemstad, Curaçao: Dr. A.J. Duits, Dr. K. Winkel. Virologists: Academisch Medisch Centrum bij de Universiteit van Amsterdam – Amsterdam:Dr. N.K.T. Back, Dr. M.E.G. Bakker, Dr. H.L. Zaaijer.Prof. dr. B. Berkhout, Dr. S. Jurriaans. CLB Stichting Sanquin Bloedvoorziening – Amsterdam: Dr. Th. Cuijpers.Onze Lieve Vrouwe Gasthuis – Amsterdam: Dr. P.J.G.M. Rietra, Dr. K.J. Roozendaal; Slotervaart Ziekenhuis – Amsterdam: Drs. W. Pauw, Drs. P.H.M. Smits, Dr. A.P. van Zanten.VU Medisch Centrum – Amsterdam: Dr. B.M.E. von Blomberg, Dr. A. Pettersson, Dr. P. Savelkoul; Ziekenhuis Rijnstate – Arnhem: Dr. C.M.A. Swanink; HAGA, ziekenhuis, locatie Leyenburg – Den Haag: Dr. P.F.H. Franck, Dr. A.S. Lampe; Medisch Centrum Haaglanden, locatie Westeinde – Den Haag: Drs. C.L. Jansen.; Streeklaboratorium Twente – Enschede: Dr. R. Hendriks.Streeklaboratorium Groningen – Groningen: Dr. C.A. Benne;.Streeklaboratorium Volksgezondheid Kennemerland – Haarlem: Dr. J. Schirm, Dr. D. Veenendaal.Laboratorium voor de Volksgezondheid in Friesland – Leeuwarden: Dr. H. Storm, Drs. J. Weel, Drs. J.H. van Zeijl; Leids Universitair Medisch Centrum – Leiden: Dr. H.C.J. Cla as, Prof. dr. A.C.M. Kroes.Academisch Ziekenhuis Maastricht – Maastricht: Prof. dr. C.A.M.V.A. Bruggeman, Drs. V.J. Goossens.Universitair Medisch Centrum St. Radboud – Nijmegen: Prof. dr. J.M.D. Galama, Dr. W.J.G. Melchers, Dr. Verduyn-Lunel.Erasmus MC – Rotterdam:Dr. G.J.J. van Doornum, Dr. H.G.M. Niesters, Prof. dr. A.D.M.E. Osterhaus, Dr. M. Schutten.St. Elisabeth Ziekenhuis – Tilburg: Dr. A.G.M. Buiting.Universitair Medisch Centrum Utrecht – Utrecht: Dr. C.A.B. Boucher, Dr. E. Boel, Dr. R. Schuurman.Catharina Ziekenhuis – Eindhoven: Dr. A.F. Jansz, drs. M. Wulf; Pharmacologists: Medisch Centrum Alkmaar – Alkmaar: Dr. A. Veldkamp.Slotervaart Ziekenhuis – Amsterdam: Prof. dr. J.H. Beijnen, Dr. A.D.R. Huitema.Universitair Medisch Centrum St. Radboud – Nijmegen: Dr. D.M.Burger.Academisch Medisch Centrum bij de Universiteit van Amsterdam – Amsterdam: Drs. H.J.M. van Kan.


96

reference List

1. Bezemer D, de Wolf F, Boerlijst MC, van Sighem A, Hollingsworth TD, Prins M, et al. A resurgent HIV-1

epidemic among men who have sex with men in the era of potent antiretroviral therapy. AIDS 2008;

22(9):1071-1077.

2. Jaffe HW, Valdiserri RO, De Cock KM. The reemerging HIV/AIDS epidemic in men who have sex with

men. JAMA 2007; 298(20):2412-2414.



191(9):1403-1409.

4. Hollingsworth TD, Anderson RM, Fraser C. HIV-1 Transmission, by Stage of Infection. J Infect Dis 2008;

198(5):687-693.

5. Xiridou M, Geskus R, de WJ, Coutinho R, Kretzschmar M. Primary HIV infection as source of HIV trans-

mission within steady and casual partnerships among homosexual men. AIDS 2004; 18(9):1311-1320.

6. Resik S, Lemey P, Ping LH, Kouri V, Joanes J, Perez J, et al. Limitations to contact tracing and phyloge-

netic analysis in establishing HIV type 1 transmission networks in Cuba. AIDS Res Hum Retroviruses 2007;

23(3):347-356.

7. Brenner BG, Roger M, Routy JP, Moisi D, Ntemgwa M, Matte C, et al. High rates of forward transmission

events after acute/early HIV-1 infection. J Infect Dis 2007; 195(7):951-959.

8. Yerly S, Vora S, Rizzardi P, Chave JP, Vernazza PL, Flepp M, et al. Acute HIV infection: impact on the spread

of HIV and transmission of drug resistance. AIDS 2001; 15(17):2287-2292.

9. Yerly S, Jost S, Telenti A, Flepp M, Kaiser L, Chave JP, et al. Infrequent transmission of HIV-1 drug-resistant

variants. Antivir Ther 2004; 9(3):375-384.

10. Brown AE, Gifford RJ, Clewley JP, Kucherer C, Masquelier B, Porter K, et al. Phylogenetic Reconstruction

of Transmission Events from Individuals with Acute HIV Infection: Toward More-Rigorous Epidemiological

Definitions. J Infect Dis 2009; 199(3):427-431.



12. Drumright LN, Frost SD. Sexual networks and the transmission of drug-resistant HIV. Curr Opin Infect Dis

2008; 21(6):644-652.

13. Bezemer D, de Ronde A, Prins M, Porter K, Gifford R, Pillay D, et al. Evolution of transmitted HIV-1 with

drug-resistance mutations in the absence of therapy: effects on CD4+ T-cell count and HIV-1 RNA load.

Antivir Ther 2006; 11(2):173-178.

14. Little SJ, Frost SDW, Wong JK, Smith DM, Pond SLK, Ignacio CC, et al. Persistence of transmitted drug

resistance among subjects with primary human immunodeficiency virus infection. Journal of Virology 2008;

82(11):5510-5518.

15. Brenner BG, Roger M, Moisi DD, Oliveira M, Hardy I, Turgel R, et al. Transmission networks of drug resis-

tance acquired in primary/early stage HIV infection. AIDS 2008; 22(18):2509-2515.


97




17. Bezemer D, Jurriaans S, Prins M, van der Hoek L, Prins JM, de Wolf F, et al. Declining trend in transmission

of drug-resistant HIV-1 in Amsterdam. AIDS 2004; 18(11):1571-1577.

18. Bezemer D, van Sighem A, de Wolf F, Cornelissen M, van der Kuyl AC, Jurriaans S, et al. Combination

antiretroviral therapy failure and HIV super-infection. AIDS 2008; 22(2):309-311.

19. Hue S, Gifford RJ, Dunn D, Fernhill E, Pillay D. Demonstration of Sustained Drug-Resistant HIV-1 Line-

ages Circulating amongst Treatment-Naive Individuals. J Virol 2009.

20. Gras L, van Sighem A, Smit C, Zaheri S, Schuitemaker H., de Wolf F. Monitoring of Human Immunodefi-

ciency Virus (HIV) Infection in the Netherlands. In. Amsterdam: Stichting HIV Monitoring; 2007.



22. Pozniak A, Gazzard B, Anderson J, Babiker A, Churchill D, Collins S, et al. British HIV Association

(BHIVA) guidelines for the treatment of HIV-infected adults with antiretroviral therapy. HIV Med 2003; 4

suppl 1:1-41.

23. Oette M, Kaiser R, Daumer M, Petch R, Fatkenheuer G, Carls H, et al. Primary HIV drug resistance and

efficacy of first-line antiretroviral therapy guided by resistance testing. J Acquir Immune Defic Syndr 2006;

41(5):573-581.

24. Gras L, van Sighem A, Smit C, Zaheri S, Schuitemaker H, de Wolf F. Monitoring of human immunode-

ficiency virus (HIV) infection in the Netherlands, Chapter 6, 9&10, http://www.hiv-monitoring.nl, HIV

Monitoring Foundation. In. Amsterdam; 2007.

25. Kuiken C, Korber B, Shafer RW. HIV sequence databases. AIDS Rev 2003; 5(1):52-61.

26. Johnson VA, Brun-Vezinet F, Clotet B, Kuritzkes DR, Pillay D, Schapiro JM, et al. Update of the drug resis-

tance mutations in HIV-1: Fall 2006. Top HIV Med 2006; 14(3):125-130.

27. Gonzales MJ, Delwart E, Rhee SY, Tsui R, Zolopa AR, Taylor J, et al. Lack of detectable human immu-

nodeficiency virus type 1 superinfection during 1072 person-years of observation. J Infect Dis 2003;

188(3):397-405.

28. Hue S, Clewley JP, Cane PA, Pillay D. HIV-1 pol gene variation is sufficient for reconstruction of transmis-

sions in the era of antiretroviral therapy. AIDS 2004; 18(5):719-728.

29. Saitou N, Nei M. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol Biol

Evol 1987; 4(4):406-425.

30. Tamura K, Nei M, Kumar S. Prospects for inferring very large phylogenies by using the neighbor-joining

method. Proc Natl Acad Sci U S A 2004; 101(30):11030-11035.

31. Tamura K, Dudley J, Nei M, Kumar S. MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) soft-

ware version 4.0. Mol Biol Evol 2007; 24(8):1596-1599.

32. Felsenstein J. Confidence limits on phylogenies: An approach using the bootstrap. Evolution 1985;

39:783-791.


98

33. Kimura M. A simple method for estimating evolutionary rates of base substitutions through comparative

studies of nucleotide sequences. J Mol Evol 1980; 16(2):111-120.

34. Hue S, Pillay D, Clewley JP, Pybus OG. Genetic analysis reveals the complex structure of HIV-1 transmission

within defined risk groups. Proc Natl Acad Sci U S A 2005; 102(12):4425-4429.

35. Shafer RW, Rhee SY, Bennett DE. Consensus drug resistance mutations for epidemiological surveillance:

basic principles and potential controversies. Antivir Ther 2008; 13 suppl 2:59-68.

36. Yerly S, von Wyl V, Ledergerber B, Boni J, Schupbach J, Burgisser P, et al. Transmission of HIV-1 drug resis-

tance in Switzerland: a 10-year molecular epidemiology survey. AIDS 2007; 21(16):2223-2229.

37. Lohse N, Obel N, Kronborg G, Jorgensen LB, Pedersen C, Larsen CS, et al. Declining prevalence of HIV-

infected individuals at risk of transmitting drug-resistant HIV in Denmark during 1997-2004. Antiviral

Therapy 2006; 11(5):591-600.

38. Burns S, Cameron S, Cane P, Down P, Chrystie I, Churchill D, et al. Evidence of a decline in transmitted

HIV-1 drug resistance in the United Kingdom. AIDS 2007; 21(8):1035-1039.

39. Routy JP, Machouf N, Edwardes MD, Brenner BG, Thomas R, Trottier B, et al. Factors associated with a

decrease in the prevalence of drug resistance in newly HIV-1 infected individuals in Montreal. AIDS 2004;

18(17):2305-2312.



15(17):2293-2301.

41. Brown AJL, Frost SDW, Mathews WC, Dawson K, Hellmann NS, Daar ES, et al. Transmission fitness of


population. Journal of Infectious Diseases 2003; 187(4):683-686.

42. Sanchez MS, Grant RM, Porco TC, Getz WM. HIV drug-resistant strains as epidemiologic sentinels.

Emerging Infectious Diseases 2006; 12(2):191-197.

43. Sanchez MS, Grant RM, Porco TC, Gross KL, Getz WM. A decrease in drug resistance levels of the HIV

epidemic can be bad news. Bulletin of Mathematical Biology 2005; 67(4):761-782.

44. Dukers NH, Fennema HS, van der Snoek EM, Krol A, Geskus RB, Pospiech M, et al. HIV incidence and

HIV testing behavior in men who have sex with men: using three incidence sources, The Netherlands,

1984-2005. AIDS 2007; 21(4):491-499.



2004; 18(14):1943-1949.

46. Stolte IG, De Wit JBF, Kolader M, Fennema H, Coutinho RA, Dukers NHTM. Association between ‘safer

sex fatigue’ and rectal gonorrhea is mediated by unsafe sex with casual partners among HIV-positive homo-

sexual men. Sexually Transmitted Diseases 2006; 33(4):201-208.

47. Stolte IG, Dukers NHTM, Geskus RB, Coutinho RA, De Wit JBR. Homosexual men change to risky sex

when perceiving less threat of HIV/AIDS since availability of highly active antiretroviral therapy: a longitu-

dinal study. AIDS 2004; 18(2):303-309.

99

VIII discussion

Before combination antiretroviral therapy (cART) became available in 1996, HIV-infected patients were treated with nucleoside reverse transcriptase inhibitors only, and viral rebounds of resistant strains emerged within several weeks [1]. Blocking various steps of the viral replication cycle with cART was a promising strategy, although there was a risk that resistance would again take over sooner or later. However after more than a decade in most HIV-infected patients treated with cART, HIV production is still successfully turned down to a very low level [2]. As a result the amount of HIV in peripheral blood and semen, and thus infectiousness is reduced [3]. To use cART as an effective prevention method to decrease HIV transmission, obviously HIV-infected patients first need to be diagnosed, and transmittable resistant strains should not evolve. Since successful cART treatment results in an increased life expectancy, these treated patients might still contribute substantially to new transmission when experiencing episodes of viral production, even when very low [4-6].

Men having sex with man (MSM) is the largest risk group for HIV-1 infection in the Nether-lands, and annual diagnoses are continually increasing [2]. From patient data we assume that only 8% of MSM have been infected outside the Netherlands. Analysing polymerase (pol) gene sequences revealed that only 8% of MSM had an infection with HIV-1 similar to that found in patients from other risk-groups. In this thesis we studied the impact of cART on the trans-mission dynamics of HIV-1 among MSM in the Netherlands. We developed a mathematical transmission model for the interpretation of longitudinal HIV-1 surveillance data and studied pol sequences obtained from 2877 HIV-1 subtype B infected patients for information on trans-mission networks.

HIV-1 epidemic amongst msm – mathematical model approach

In chapter 5 and 6 we describe the HIV-1 transmission model framework developed for the interpretation of HIV-1 surveillance data on MSM in the Netherlands. The model includes the distribution in disease progression and rates of cART success and failure. Using HIV-1 and AIDS surveillance data, the model allowed for estimating changes in average time from infec-tion to diagnosis and the average net-transmission rate per HIV-1 infected MSM. The estimated net-transmission rate represents mainly changes in average level and frequency of risk behavior between sero-discordant couples, but the estimated changes in this transmission rate might


100

in part also be due to saturation of the contact network with HIV-1, or even due to evolving virulence.

We estimated that risk-behaviour has increased with 66% since the introduction of cART in 1996. Time from infection to diagnosis was estimated to be 2.5 years on average in 2006. Around 1600 undiagnosed MSM at the end of 2006 are estimated to account for 90% of new HIV-1 infections. Interestingly, earlier diagnosis and increasing risk-behaviour resulted in a fairly constant number of undiagnosed infections over the years. Scenario analysis shows that the increase in risk-behaviour has completely offset the benefits of cART in reducing epidemic spread. The reproduction number is currently around the threshold for epidemic spread, mean-ing that every newly infected MSM will on average infect one other MSM over his life time. If the risk behaviour would not have increased with the introduction of cART, the reproduction number would be 0.6, and the epidemic would be in decline. Hypothetical scenario analysis further showed that in the absence of cART limiting infectiousness in treated patients, the epidemic could be more than twice as large as it is at present.

When we stared the research described in this thesis, the prevailing explanation for the increas-ing number of new HIV-1 diagnoses among MSM was the delayed testing for HIV and that most of these men had been infected many years ago. However, we noticed that the number of new diagnoses was increasing since 1998. Our model results show that indeed patients are tested earlier in their infection but that simultaneously the number of annual new infections increases. The recent increase in the proportion of newly diagnosed individuals with high CD4 cell counts corroborates our model’s inferences in interpreting recent increases in annual number of new HIV diagnoses as rising transmission and increased diagnosis rather than improved diagnosis of people infected many years in the past. The estimated incidence curve is qualitatively very similar to longitudinal data on syphilis infections at the STI clinic in Amsterdam [7-9]. The data update, and the univariate sensitivity analysis that explored the impact of varying assump-tions in both the input parameters, interpretation of incomplete data, and model structure show that the main model predictions are robust.

testingWhen the average time from infection to diagnosis is reduced from 2.5 years on average to one year, the reproduction number will still be close to one, but this will correspond to 43% reduc-tion in new infections in the coming decade. Thus, together with all other possible prevention measures early testing is extremely important. With the current rapid HIV tests, it is possible to reduce the time from infection to diagnosis to 3 months or less, since antibodies to HIV-1 are detectable after two to four weeks in most people [10]. Previous but also current intervention policy likely delays HIV testing. Before the introduction of cART people were not encouraged to get tested because there was no treatment available, and currently intervention with rapid

vIII dIscussIon

101

tests although available, is still not approved and not actively promoted. Think for instance of a user-friendly and easy available home test including website and telephone number for sup-port. Currently in high risk-taking MSM sub-populations it may in fact be more risky to have unprotected intercourse with someone who is not tested than with someone who is on success-ful cART. If MSM adapt their risk-behaviour to the knowledge of their and their partners HIV status taking into account the window period of the HIV test, the reproduction number will decrease even further. But other sexually transmitted diseases should still be tested regularly as the spread of these facilitates HIV to spread.

Key assumption A biologically plausible, currently debated, but non-validated key assumption is that success-fully treated individuals maintain very low amounts of virus particles in peripheral blood and therefore do not contribute to onward infections. Individuals who fail therapy are assumed to be as infectious as untreated individuals. While it is unlikely that mathematical model-based analyses of the type used here can alone validate this assumption, alternative scenarios which consider the effects of relaxing this assumption in data-driven analyses should be explored in detail in future work [11-14]. Previous modelling studies have emphasised the potential effect of ongoing transmission during incompletely suppressive antiretroviral therapy in driving the epidemic [11-13, 15-22]. We cannot rule out such a contribution to the Dutch HIV epidemic, an issue that could be best addressed by empirical studies on discordant pairs. This is currently being investigated in the HPTN052 trial (http://www.hptn.org/research_studies/HPTN052.asp).

future modelling researchFuture model adaptations should include the results of such discordant pair studies, as well as diagnosis and treatment in primary stage. In addition, multivariate sensitivity analyses should be further developed. To extend the model with transmission from the treated group would on the long run need a death rate included, which is not a straightforward extension to make. The simplest way to incorporate this would be in the form of a fixed rate constant times the number of infected patients, whereas including a bookkeeping of the age distribution and age-dependent death rates would make the model considerably more complex to analyse. When using the model to study transmission dynamics in a certain risk group, it is crucial to evaluate model assumptions and mirror those to the risk group under study. For instance, when using the model to study transmission dynamics in the heterosexual population one should consider adding the HIV incidence in the drug user population as local import. The exact role of pri-mary infections could not be identified as this depends on the unknown rate of partner change during that period. Findings from transmission network studies could possibly improve the parameters in a future version.


102

HIV-1 epidemic amongst msm – phylogenetic approach

time between infection and onwards transmissionIn chapter 7 we estimated the median time between infection and onward transmission for a selection of potential transmission pairs selected from a phylogenetic tree of HIV-1 subtype B pol sequences obtained from 404 MSM shortly after their known date of infection. The results indicate that 25% of onward transmissions occurred within 7 months after infection, half of the transmissions within 17 months, and 75% within 2.7 years. These findings are compatible with those obtained by the mathematical transmission model described in chapter 5 and 6. Irrespective all artifacts described, the results indicate that transmission of HIV-1 among MSM happens primarily in the years early after primary infection. Our estimate of the median time between transmissions, for which we used only sequences corresponding to infections with a well documented date of infection, is in agreement with a study in the United Kingdom [23]. Lewis et al., estimated the time between the nodes in a phylogenetic tree of sequences obtained from patients with an unknown date of infection, and additional assumptions were needed to estimate the respective dates of infection of the sequences used [23]. However it would be interesting to compare both methods in our study population. Our results further indicate that 11% of onward transmissions occurred within 3 months after infection. In comparison with the reproduction number of primary infection as resulted from our mathematical model (chapter 5 and 6) this could give an indication on the average partner change rate in MSM during primary infection [3]. Our finding was in agreement with a previous mathematical model study within the Amsterdam Cohort Studies [24]. Xiridou et al estimated from behavioural data of MSM that on average 11% of infections take place during the first 2.5 months of infection.

transmission of resistant HIV-1 strainsIn a selection of new infections (100 in chapter 2 and 404 in chapter 7) we observed resistant HIV-1 being transmitted in up to 30% of the cases during the years of antiretroviral therapy based on one drug class. The level of transmission of HIV-1 resistant to one drug class decreased with the introduction of cART to around 6%. Since 2004, 1% of infections among MSM con-cerned a multidrug resistant strain. A similar trend is found in the ATHENA cohort among the newly diagnosed chronic infections [2]. Although the level of transmission of resistant HIV-1 substantially varied between risk groups and countries that implemented cART, the efficacy of cART and the lower transmission potential of multidrug resistant HIV-1 strains was reported in several studies [2, 18, 25-36]. Variations in the combination of antiretroviral drugs used and differences in healthcare approaches may in part have caused those differences. Trends in resis-tance should preferably be evaluated within the context of one well defined risk group, which would make comparisons of results obtained in different studies more feasible as well.

vIII dIscussIon

103

Once a resistant strain is transmitted from those failing cART, it may also transmit onwards within the undiagnosed population. Thus resistant strains are transmitted from an additional group of patients than the wildtype virus, and based on that are expected to dominate the epi-demic on the long term. This seemed to be occurring quickly in the pre-cART era, although numbers in those years where small. In the cART era the almost absence of transmission of multidrug resistant strains reflects the efficacy of cART in controlling the virus, the efficacy of healthcare in changing a patients therapy combination in time when failing, and likely also the lower transmission capacity of those strains that did manage to escape from people on failing therapy.

On the other hand the low percentage of resistant infections reflects our model findings on delayed diagnosis of HIV-infected MSM in the Netherlands [37, 38]. If all patients were diag-nosed and subsequently treated with cART, and only one occasional new infection with multi drug resistance escaped, 100% of transmitted HIV would be resistant. Differences in propor-tion of undiagnosed infections between countries can account in part for differences in the percentage resistance measured. Important is the number of infections with a resistant strain and the reproduction number of the resistant strains. In addition, the overall number of new infections is of importance, since for every new infection there is a chance to fail cART and evolve fitter resistant strains. The model estimates in chapters 5 and 6, on an increasing HIV-1 incidence, together with an overall 6% of new infections is with a resistant strain, imply that the number of infections with HIV-1 conferring resistant mutations is increasing.

fitness of transmitted resistant strainsThe surrounding environment of HIV, including the presence or absence of cART, determines which strain is most fit [39]. When a patient ceases therapy after cART failure due to develop-ment of multidrug resistant strains, archived wild type virus will again be dominant within weeks [40]. This reflects the fitness loss from mutations conferring multidrug resistance. Study-ing the evolution of resistant strains that are transmitted to persons that are antiretroviral ther-apy-naive (chapter 3), we found that several mutations at known resistance sites reverted back to wild type, but amino acid position 215 of reverse transcriptase (RT) that also mutated to other amino acids. This happened due to evolution (positive selection), as no wild type was present upon infection. One exception may occur when yet infected with a mix, but from sequence comparisons this did not seem the case in our study group. Confirming results from other groups reporting reduced fitness of HIV-1 with these mutations [41-43]. Initial HIV infec-tion consists of a very homogeneous virus population [44, 45]. Mutations at resistant sites that mutate back to wildtype shortly after infection might not be detected when sequencing at a later stage during infection [46], and will not be transmitted further, but when archived in reservoirs might come up more easily under certain therapies. Other mutations did not change, among which the amino acid position 41L and 215 revertant of RT, which confirmed studies on their


104

comparable fitness to wildtype [42, 47, 48]. Mutations do not change, either because the muta-tion did not affect or even increased their fitness (purifying selection), because the study period was too short, or because of compensatory mutations [49]. In addition it is less straightforward to mutate back to wild type if this involves intermediate phenotypes that have lower fitness. Moreover, what matters is not just being fit as a mutant (fitness peak next to a steep valley), but as the surrounding cloud of mutants forming the quasispecies (smoother landscape) [48].

Our data suggest an indication that ART-naïve patients infected with an HIV-1 strain with resistant mutations that change have a slower CD4 decline after seroconversion than individu-als infected with viruses with mutations that remain. However the fitness of various mutant combinations should be studied in more detail in a much larger study group, preferably with longitudinal data including comparisons to wildtype infections, and in combination with in vitro replication capacity experiments [42, 47, 50].

failing cArtIn chapter 4, we did not find evidence for superinfection with a resistant strain to contribute to patients failing therapy after initially being successfully treated. But at least 23% of those starting cART without previous monotherapy were initially infected with a resistant strain. Patients who received monotherapy before the introduction of cART were also found to have a significantly higher chance of failing cART. Untill now multidrug resistant strains evolving in patients on cART still have a lower viral load and transmission fitness [51], but compensa-tory mutations are likely to evolve when inadequate therapy is continued [52]. Under full sup-pression the virus is not able to evolve but, in every replication round the chance of adaptive mutations can increase its replication capacity under the changed conditions. Over the past years treatments have improved towards less side-effects so that patients can adhere better to the treatment [2].

Superinfection with a resistant strain may occur more likely in sero-concordant couples on the same cART combinations [53]. For both individual health care and HIV care on a population level, it is important to sequence HIV pol when patients are diagnosed with HIV in order to prescribe effective treatment. Regular monitoring of patients should continue, with the aim to change antiretroviral therapy in time in case of treatment failure. This is important because risk-behavior is shown to increase amongst patients who think they are successfully treated, even if this is not necessarily so [54].

transmission networks of resistant HIV-1 infectionsIn the selection in chapter 7, of HIV-1 subtype B pol sequences obtained from 404 MSM shortly after their known date of infection, 24 had resistant mutations. We studied the networks of the 24 new infections with a resistant strain in a phylogenetic tree of pol sequences isolated from

vIII dIscussIon

105

2877 HIV-1 subtype B infected patients. We found that a significant proportion of resistant transmissions are from people who themselves are infected with a resistant strain. The revertant mutant at amino acid position 215 of RT is as fit as wildtype [42], and transmits depending on the networks it is in. Only before 1997 transmission of the full resistant 215 RT mutant was observed, indicating the contribution to transmission of patients failing AZT monotherapy. Two new infections in 2004 and 2005 were with a multi-drug resistant strain, with resistance to dual- and triple drug classes respectively. Multidrug resistant strains seem only to be transmit-ted occasionally from someone failing treatment, and no evidence is found for these strains to be part in onwards transmissions from the undiagnosed population. The virus strain with muta-tions against all three drug classes was likely not transmitted from within the Dutch epidemic. This emphasizes that the HIV epidemic cannot be stopped locally because it is a global problem. In the cART era there was also an infection with a mutation at amino acid position 184 of RT, which normally reverts back to wildtype in the absence of selection pressure by antiretroviral therapy (chapter 3). This might indicate that treated patients in the cART era transmit strains with resistance to only one drug class, implying poor adherence to the complete therapy combi-nation. Several other studies also reported phylogenetic clustering of resistant strains obtained from ART naive patients [33, 55, 56], but we were able to study new infections in networks of all sequences available from both treated and untreated patients, and to link this to the meas-ured percentage of resistant strain transmission.

future phylogenetic researchOur next step is to use all available pol sequences to study the networks resistant strains are in, so to monitor possibly more rare transmissions of multi-drug resistance. Risk-behaviour from HIV positive MSM in the Amsterdam Cohort Studies will be coupled. The study should also be expanded to investigate transmission networks of other risk-groups and non-B subtypes. Lately many research questions have been studied comparing characteristics of infections early in the epidemic with more recent infections. To prevent comparing apples and oranges such research questions should preferably be studied in the light of transmission clusters; in vitro fitness exper-iments on evolution towards a more virulent virus [50]; the adaptation of HIV-1 to the human immune system at the population level (paper submitted by Schellens et al.,2009); and on her-itability and evolution of viral load [57] (paper submitted by Gras et al.,2009). Currently, we also study the rate of molecular evolution at the population level of phylogenetically connected infections, to assess if the changes in reproduction number and the introduction of cART can be verified at the molecular level. The rate of molecular evolution on population level is known to differ per risk-group [58]. When transmissions occur repeatedly during the initial stage of host infection the rate of HIV-1 evolution on the population level decreases, conversely, in slow spread the virus evolution at the population level will increase. Comparing the rate of molecular evolution on population versus intra-host level could indicate the average number of bottlenecks a virus passes in order to cause the measured delay in evolution at the population level.


106

conclusion

In conclusion there is a resurgent epidemic amongst MSM. Increasing risk-behaviour has offset the benefits of cART in reducing HIV-1 transmission. HIV-1 among MSM spreads mostly from the undiagnosed population, and so do resistant strains. Transmission of multidrug resist-ance is rare. Early diagnosis should be achieved by contact tracing and improved and frequent rapid testing. Together with early access to cART and a decrease in risk behaviour this may effectively contain epidemic spread amongst MSM.

reference List

1. Boucher CA, O’Sullivan E, Mulder JW, Ramautarsing C, Kellam P, Darby G, et al. Ordered appearance of

zidovudine resistance mutations during treatment of 18 human immunodeficiency virus-positive subjects. J

Infect Dis 1992; 165(1):105-110.

2. Gras L, van Sighem A, Smit C, Zaheri S, de Wolf F, Schuitemaker H, et al. Monitoring of human immuno-

deficiency virus (HIV) infection in the Netherlands. In. Amsterdam: Stichting HIV Monitoring; 2008.

3. Hollingsworth TD, Anderson RM, Fraser C. HIV-1 transmission, by stage of infection. J Infect Dis 2008;

198(5):687-693.

4. Sedaghat AR, Siliciano RF, Wilke CO. Low-level HIV-1 replication and the dynamics of the resting CD4+ T

cell reservoir for HIV-1 in the setting of HAART. BMC Infect Dis 2008; 8:2.

5. Shen L, Siliciano RF. Viral reservoirs, residual viremia, and the potential of highly active antiretroviral

therapy to eradicate HIV infection. J Allergy Clin Immunol 2008; 122(1):22-28.

6. Nettles RE, Kieffer TL, Kwon P, Monie D, Han Y, Parsons T, et al. Intermittent HIV-1 viremia (Blips) and

drug resistance in patients receiving HAART. JAMA 2005; 293(7):817-829.





HIV, among young homosexual men in Amsterdam: are STIs still reliable markers for HIV transmission? Sex

Transm Infect 2005; 81(1):34-37.




10. Richman DD. Human Immunodeficiency Virus, International Medical Press 1-901769-09-7; 2003.

11. Anderson RM, Gupta S, May RM. Potential of community-wide chemotherapy or immunotherapy to control

the spread of HIV-1. Nature 1991; 350(6316):356-359.

12. Wilson DP, Law MG, Grulich AE, Cooper DA, Kaldor JM. Relation between HIV viral load and infectious-

ness: a model-based analysis. Lancet 2008; 372(9635):314-320.

vIII dIscussIon

107


22 Suppl 1:S149-S159.

14. Bogaards JA, van Ballegooijen WM, Jan WG, Boerlijst MC, Goudsmit J. Is population-level perversity a

likely outcome of mass vaccination against HIV? Lancet Infect Dis 2005; 5(5):254-256.




cisco. Science 2000; 287(5453):650-654.






15(17):2293-2301.







17(7):1029-1038.





24. Xiridou M, Geskus R, de WJ, Coutinho R, Kretzschmar M. Primary HIV infection as source of HIV trans-

mission within steady and casual partnerships among homosexual men. AIDS 2004; 18(9):1311-1320.

25. Yerly S, von Wyl V, Ledergerber B, Boni J, Schupbach J, Burgisser P, et al. Transmission of HIV-1 drug resist-

ance in Switzerland: a 10-year molecular epidemiology survey. AIDS 2007; 21(16):2223-2229.

26. Lohse N, Obel N, Kronborg G, Jorgensen LB, Pedersen C, Larsen CS, et al. Declining prevalence of HIV-

infected individuals at risk of transmitting drug-resistant HIV in Denmark during 1997-2004. Antiviral

Therapy 2006; 11(5):591-600.

27. Burns S, Cameron S, Cane P, Down P, Chrystie I, Churchill D, et al. Evidence of a decline in transmitted

HIV-1 drug resistance in the United Kingdom. AIDS 2007; 21(8):1035-1039.

28. Grant RM, Hecht FM, Warmerdam M, Liu L, Liegler T, Petropoulos CJ, et al. Time trends in primary

HIV-1 drug resistance among recently infected persons. Jama-Journal of the American Medical Association

2002; 288(2):181-188.

29. Vella S, Palmisano L. The global status of resistance to antiretroviral drugs. Clinical Infectious Diseases 2005;

41:S239-S246.


108

30. Routy JP, Machouf N, Edwardes MD, Brenner BG, Thomas R, Trottier B, et al. Factors associated with a

decrease in the prevalence of drug resistance in newly HIV-1 infected individuals in Montreal. AIDS 2004;

18(17):2305-2312.

31. Yerly S, Vora S, Rizzardi P, Chave JP, Vernazza PL, Flepp M, et al. Acute HIV infection: impact on the spread

of HIV and transmission of drug resistance. AIDS 2001; 15(17):2287-2292.

32. Brown AJL, Frost SDW, Mathews WC, Dawson K, Hellmann NS, Daar ES, et al. Transmission fitness of


population. Journal of Infectious Diseases 2003; 187(4):683-686.

33. Yerly S, Jost S, Telenti A, Flepp M, Kaiser L, Chave JP, et al. Infrequent transmission of HIV-1 drug-resistant

variants. Antivir Ther 2004; 9(3):375-384.

34. Chaix ML, Descamps D, Wirden M, Bocket L, Delaugerre C, Tamalet C, et al. Stable frequency of HIV-1

transmitted drug resistance in patients at the time of primary infection over 1996-2006 in France. AIDS

2009; 23(6):717-724.

35. Wensing AM, van d, V, Angarano G, Asjo B, Balotta C, Boeri E, et al. Prevalence of drug-resistant HIV-1

variants in untreated individuals in Europe: implications for clinical management. J Infect Dis 2005;

192(6):958-966.

36. Wensing AM, Boucher CA. Worldwide transmission of drug-resistant HIV. AIDS Rev 2003; 5(3):140-155.

37. Sanchez MS, Grant RM, Porco TC, Gross KL, Getz WM. A decrease in drug resistance levels of the HIV

epidemic can be bad news. Bulletin of Mathematical Biology 2005; 67(4):761-782.

38. Sanchez MS, Grant RM, Porco TC, Getz WM. HIV drug-resistant strains as epidemiologic sentinels.

Emerging Infectious Diseases 2006; 12(2):191-197.

39. van Opijnen T., de RA, Boerlijst MC, Berkhout B. Adaptation of HIV-1 depends on the host-cell environ-

ment. PLoS ONE 2007; 2(3):e271.

40. Deeks SG, Wrin T, Liegler T, Hoh R, Hayden M, Barbour JD, et al. Virologic and immunologic conse-

quences of discontinuing combination antiretroviral-drug therapy in HIV-infected patients with detectable

viremia. New England Journal of Medicine 2001; 344(7):472-480.

41. Back NK, Nijhuis M, Keulen W, Boucher CA, Oude Essink BO, van Kuilenburg AB, et al. Reduced replica-

tion of 3TC-resistant HIV-1 variants in primary cells due to a processivity defect of the reverse transcriptase

enzyme. EMBO J 1996; 15(15):4040-4049.




43. Paredes R, Sagar M, Marconi VC, Hoh R, Martin JN, Parkin NT, et al. In vivo fitness cost of the M184V

mutation in multidrug-resistant human immunodeficiency virus type 1 in the absence of lamivudine. J Virol

2009; 83(4):2038-2043.

44. Keele BF, Giorgi EE, Salazar-Gonzalez JF, Decker JM, Pham KT, Salazar MG, et al. Identification and char-

acterization of transmitted and early founder virus envelopes in primary HIV-1 infection. Proc Natl Acad Sci

U S A 2008; 105(21):7552-7557.

vIII dIscussIon

109

45. Kearney M, Maldarelli F, Shao W, Margolick JB, Daar ES, Mellors JW, et al. Human immunodeficiency

virus type 1 population genetics and adaptation in newly infected individuals. J Virol 2009; 83(6):2715-2727.

46. Johnson JA, Li JF, Wei X, Lipscomb J, Irlbeck D, Craig C, et al. Minority HIV-1 drug resistance mutations

are present in antiretroviral treatment-naive populations and associate with reduced treatment efficacy. PLoS

Med 2008; 5(7):e158.

47. Little SJ, Frost SDW, Wong JK, Smith DM, Pond SLK, Ignacio CC, et al. Persistence of transmitted drug

resistance among subjects with primary human immunodeficiency virus infection. Journal of Virology 2008;

82(11):5510-5518.

48. Goudsmit J, de RA, de RE, de BR. Broad spectrum of in vivo fitness of human immunodeficiency virus type

1 subpopulations differing at reverse transcriptase codons 41 and 215. J Virol 1997; 71(6):4479-4484.

49. van Maarseveen NM, Wensing AMJ, de Jong D, Taconis M, Borleffs JCC, Boucher CAB, et al. Persistence of

HIV-1 variants with multiple protease inhibitor (PI)-resistance mutations in the absence of PI therapy can be

explained by compensatory fixation. Journal of Infectious Diseases 2007; 195(3):399-409.

50. Gali Y, Berkhout B, Vanham G, Bakker M, Back NK, Arien KK. Survey of the temporal changes in HIV-1

replicative fitness in the Amsterdam Cohort. Virology 2007; 364(1):140-146.

51. Leigh Brown AJ, Frost SD, Mathews WC, Dawson K, Hellmann NS, Daar ES, et al. Transmission fitness of


population. J Infect Dis 2003; 187(4):683-686.

52. Martinez-Picado J, Martinez MA. HIV-1 reverse transcriptase inhibitor resistance mutations and fitness: a

view from the clinic and ex vivo. Virus Res 2008; 134(1-2):104-123.

53. Pagie L, Hogeweg P. Colicin diversity: a result of eco-evolutionary dynamics. Journal of Theoretical Biology

1999; 196(2):251-+.



2004; 18(14):1943-1949.

55. Brenner BG, Roger M, Moisi DD, Oliveira M, Hardy I, Turgel R, et al. Transmission networks of drug resist-

ance acquired in primary/early stage HIV infection. AIDS 2008; 22(18):2509-2515.

56. Hue S, Gifford RJ, Dunn D, Fernhill E, Pillay D. Demonstration of Sustained Drug-Resistant HIV-1 Line-

ages Circulating amongst Treatment-Naive Individuals. J Virol 2009.



104(44):17441-17446.

58. Maljkovic B, I, Ribeiro R, Kothari M, Athreya G, Daniels M, Lee HY, et al. Unequal evolutionary rates in

the human immunodeficiency virus type 1 (HIV-1) pandemic: the evolutionary rate of HIV-1 slows down

when the epidemic rate increases. J Virol 2007; 81(19):10625-10635.

111

summary

The first AIDS cases in the Netherlands were diagnosed in 1982 among men having sex with men (MSM). In 1984 serological testing was possible and from 1991 antiretroviral mono ther-apy was available. However resistant strains emerged within several weeks on therapy, and were also shown to be transmitted. Since 1996 effective combination antiretroviral therapy (cART) is available. cART proved effective as morbidity and mortality among HIV infected patients strongly decreased. cART reduces the viral load and therewith also the infectiousness. However annual HIV-1 diagnoses among MSM have been increasing over the past decade. We wanted to know the impact that cART has had on the transmission dynamics of HIV-1 among MSM. For this we used data till 2007 from the prospective Amsterdam Cohort Studies and 24 HIV treatment centres in the Netherlands as part of the ATHENA national observational cohort.First we developed a mathematical HIV-1 transmission model including the distribution in disease progression and parameters on cART use and estimated the changes in risk behaviour and time from infection to diagnosis, needed to explain annual diagnoses of HIV-1 and AIDS. Herewith we could calculate the reproduction number R(t), being the number a newly infected MSM at time t will on average infect over his whole infectious lifespan if conditions remain the same as at time t: when larger then one the epidemic will increase, when smaller then one the epidemic will contract. We show that together with a 57% decrease in risk-behaviour R(t) declined early on from initial values above two and was maintained below one from 1985 to 2000. Since 1996, when highly active antiretroviral therapy became widely used, the risk behaviour rate has increased 66%, resulting in an increase of R(t) to around the threshold one for a self-sustaining epidemic in the latest period 2000 tot 2006. The percentage of the undiag-nosed HIV positive MSM of the total number of infected MSM has decreased to 24%, but only so due to an increase in survival of the diagnosed population. In absolute numbers around 1600 HIV positive MSM were undiagnosed at the end of 2006, estimated to be responsible for 90% of new HIV-1 transmissions. Decreasing time from infection to diagnosis, 2.5 years on average in 2006, with subsequent cART can prevent thousands of future infections. The recent increase in the proportion of newly diagnosed individuals with high CD4 cell counts corroborates our model’s inferences in interpreting recent increases in annual number of new HIV-1 diagnosis as rising transmission and increased diagnosis rather than improved diagnosis of people infected many years in the past. If the risk behaviour would not have increased with the introduction of cART, R(t) would be 0.60, and the epidemic would be in decline. Hypothetical scenario analy-sis further showed that in the absence of cART limiting infectiousness in treated patients, the epidemic could be more than twice as large as it is at present.Next we wanted to obtain better insight in the time span from infection to onwards transmis-sion. For that we performed a phylogenetic study of polymerase sequences isolated from 2877


112

HIV-1 subtype B infected patients. These RNA sequences are obtained as part of the screening for resistant strains to antiretroviral drugs, and their mutual similarities have proved valuable in reflecting the underlying transmission networks. Among MSM 25% of onward transmissions was estimated to occur within 7 months after infection, half of transmissions within 17 months, and 75% within 2.7 years. This finding is compatible with the results of the transmission model. Transmission of resistant strains from the cART treated population showed to be limited. But 23% of people failing cART, without previous monotherapy, were initially themselves infected with a resistant strain. Strains with stable resistance-related mutations in the absence of cART have formed sub-epidemics. On average 6% [95% confidence interval (CI), 3.8 – 8.7 %] of infections among MSM had drug-resistant mutations between 1987 – 2007. First after 2004, 1% of infections among MSM considered a multidrug resistant strain. The apparent pressure to mutate in the absence of cART at several resistance-associated positions confirms a decreased viral fitness of those mutations. Only when a sequence is obtained shortly after infection the presence of unstable mutations can be monitored.In conclusion there is a resurgent epidemic amongst MSM. Increasing risk-behaviour has off-set the benefits of cART in reducing HIV transmission in The Netherlands. HIV-1 among MSM spreads mostly from the undiagnosed population, and so do resistant strains. Transmis-sion of multidrug resistance is rare. Early diagnosis should be achieved by contact tracing and improved and frequent rapid testing. Together with early access to cART and a decrease in risk behaviour this may effectively contain epidemic spread amongst MSM.

113

samenvatting

De eerste AIDS gevallen in Nederland werden in 1982 gediagnosticeerd bij mannen die sex hebben met mannen (MSM). Vanaf 1984 werd het mogelijk om een serologische test te doen op antistoffen tegen infectie met het verantwoordelijke HIV virus, en vanaf 1991 was er antiretro-virale monotherapie. Echter resistente strains kwamen binnen enkele weken van therapie op, en werden ook overgedragen. Sinds 1996 is er effectieve antiretrovirale combinatietherapie (cART) beschikbaar. cART bleek effectief, de morbiditeit en mortaliteit bij HIV patiënten daalde. cART reduceert de virale load en daarmee ook de infectiositeit. Toch steeg het aantal HIV-1 diagnosen bij MSM over het laatste decennium. Wij wilden weten wat de impact van cART is geweest op de transmissiedynamiek van HIV-1 bij MSM. Hiervoor hebben we gebruik gemaakt van data tot 2007 van de prospectieve Amsterdamse Cohort Studies en de 24 HIV behandelcentra in Nederland zoals verzameld in het ATHENA nationale observationele cohort.Allereerst hebben we een mathematisch HIV-1 transmissiemodel ontwikkeld met daarin de distributie in tijd van infectie tot AIDS en inclusief parameters over het gebruik van cART. Hiermee was het mogelijk om de veranderingen in risicogedrag en tijd van infectie tot diagnose zo te bepalen dat de veranderingen over de tijd in aantallen HIV-1 diagnoses en AIDS bij MSM verklaard konden worden. Met deze resultaten was het vervolgens mogelijk om het reproductie-aantal over de tijd, R(t), uit te rekenen. R(t) is het aantal nieuwe infecties dat een MSM geïn-fecteerd op tijdstip t maakt gedurende de rest van zijn leven onder gelijkblijvende condities: als groter dan één zal de epidemie groeien, en als kleiner dan één zal de epidemie krimpen. We laten zien dat samen met een daling van 57% in risicogedrag, R(t) onder de één daalde van 1985 tot 2000. Sinds cART vanaf 1996 voor iedereen beschikbaar kwam, is het risicogedrag echter weer gestegen met 66%. Dit resulteert in een grotere R(t) van rond de epidemische drempel één, wat betekent dat de epidemie in de periode 2000 tot 2006 zichzelf in stand houdt. Het per-centage ongediagnosticeerde MSM van het totaal aantal HIV-1 positieve MSM is gedaald naar 24%. Dit komt doordat de gediagnosticeerde populatie langer leeft. In absolute getallen waren er eind 2006 ongeveer 1600 HIV-1 positieve MSM niet gediagnosticeerd, welke verantwoorde-lijk zijn voor 90% van de HIV-1 transmissie. Een daling in de tijd tussen infectie en diagnose, gemiddeld 2,5 jaar in 2006, samen met een eerdere start van cART, zou duizenden toekomstige infecties kunnen voorkomen. De recente stijging in de proportie van nieuw gediagnosticeerde individuen met hoge CD4 cel waarden bevestigt de resultaten van ons model in het interprete-ren van recente stijgingen in jaarlijkse aantallen HIV-1 als toename van HIV-1 transmissie en vroegere diagnose, en niet als weerspiegeling van diagnose bij mensen die jaren geleden al geïn-fecteerd zijn. Als het risicogedrag onder MSM niet was gestegen na de introductie van cART, zou R(t) 0.60 zijn, en zou de epidemie afnemen. Hypothetische scenarioanalyse liet verder zien


114

dat in afwezigheid van de lagere infectiviteit door cART in behandelde patiënten, de epidemie twee keer zo groot zou zijn als nu.Vervolgens wilden we beter inzicht verkrijgen in de tijd tussen infectie en transmissie. Hiervoor deden we een fylogenetische studie met de HIV-1 subtype B polymerase sequenties van 2877 patiënten. Deze sequenties zijn afgenomen voor de screening van resistentie tegen antiretrovirale drugs. De mate van gelijkheid van deze RNA sequenties van verschillende personen reflecteert de onderliggende transmissienetwerken. Onder MSM vonden we dat 25% van de transmis-sies plaats vindt binnen 7 maanden na infectie, 50% binnen 17 maanden, en 75% binnen 2,7 jaar. Deze bevinding is compatibel met de resultaten van het mathematische transmissiemodel. Transmissie van resistent HIV-1 vanuit de met cART behandelde populatie bleek gering. Ech-ter tenminste 23% van de mensen die falen op cART, zonder eerst monotherapie te gebruiken, waren reeds geïnfecteerd met resistent HIV-1. Transmissie van HIV-1 met stabiele resistentie gerelateerde mutaties hebben subepidemieën gevormd. Gemiddeld 6% van de HIV-1 positieve MSM geïnfecteerd tussen 1987 – 2007 had resistentie gerelateerde mutaties. Na 2004, had 1% van de HIV-1 infecties bij MSM resistentie tegen 2 of 3 drugs types. De druk om te muteren in afwezigheid van cART op verschillende resistentie gerelateerde posities duid op een lagere fitness van HIV-1 met deze mutaties. Alleen wanneer een sequentie wordt afgenomen kort na infectie kan de aanwezigheid van deze mutaties aangetoond worden.Concluderend: er is opnieuw een HIV-1 epidemie gaande bij MSM in Nederland. De stijging in risicogedrag heeft de voordelen van cART in het reduceren van HIV transmissie teniet gedaan. Bij MSM verspreid zowel HIV-1 met als zonder resistente mutaties zich voornamelijk vanuit de ongediagnosticeerde populatie. Transmissie van HIV-1 met resistentie tegen meerdere drugs typen is vooralsnog incidenteel. Vroegere diagnose zou mogelijk kunnen zijn door partner waarschuwing en beter, makkelijker, en vaker te testen. Samen met vroege start van cART en een daling in risicogedrag zou dit effectief de HIV-1 epidemie onder MSM reduceren.

115

curriculum vitae

Name Daniela Olga Bezemeremail [email protected] of birth 14 September 1973 place of birth Willemstad, Curaçao, Netherlands Antilles

publicationsdaniela Bezemer, Ard van Sighem, Vladimir V. Lukashov, Lia van der Hoek, Nicole Back, Rob Schuur-man, Charles A.B. Boucher, Eric C.J. Claas, Maarten C. Boerlijst, Roel A. Coutinho, Frank de Wolf for the ATHENA observational cohort, Transmission networks of HIV-1 among men having sex with men in the Netherlands, submitted (2009)

Luuk Gras, Suzanne Jurriaans, Ard van Sighem, daniela Bezemer, Margreet Bakker, Christophe Fraser, Jan Prins, Ben Berkhout, Frank de Wolf, Mean viral load at set point is higher in more recent calendar years in patients from the ATHENA cohort, submitted (2009)

daniela Bezemer, Frank de Wolf, Maarten C. Boerlijst, Ard van Sighem, T. Deirdre Hollingsworth, and Christophe Fraser, 27 years of the HIV epidemic amongst men having sex with men in the Netherlands: an in depth mathematical model-based analysis, submitted (2009)

daniela Bezemer, Frank de Wolf, Maarten C. Boerlijst, Ard van Sighem, T. Deirdre Hollingsworth, Maria Prins, Ronald B. Geskus, Luuk Gras, Roel A. Coutinho, and Christophe Fraser, A resurgent HIV-1 epidemic among men who have sex with men in the era of potent antiretroviral therapy, AIDS 2008 22(9):1071-1077

daniela Bezemer, Ard van Sighem, Frank de Wolf, Marion Cornelissen, Suzanne Jurriaans, Maria Prins, Roel A. Coutinho, and Vladimir V. Lukashov, cART Failure and HIV Super-Infection, AIDS (2008) 22(2):309-311

Carlos Llorens, Ricardo Futami, daniela Bezemer, and Andres Moya, The Gypsy Database (GyDB) of Mobile Genetic Elements, Nucleic Acids Research (2008); 36(Database issue):D38-D46

daniela Bezemer, Anthony de Ronde, Maria Prins, Kholoud Porter, Robert Gifford, Deenan Pillay, Bernard Masquelier, Hervé Fleury, Francois Dabis, Nicole Back, Suzanne Jurriaans, Lia van der Hoek on behalf of the CASCADE collaboration, Evolution of transmitted HIV-1 with drug-resistance mutations in the absence of therapy: effects on CD4+ T-cell count and HIV-1 RNA load, Antiviral Therapy (2006) 11(2):173-178

daniela Bezemer, Suzanne Jurriaans, Maria Prins, Lia van der Hoek, Jan M. Prins, Frank de Wolf, Ben Berkhout, Roel A. Coutinho, Nicole K.T. Back, Declining trend in transmission of drug-resistant HIV-1 in Amsterdam, AIDS (2004) 18:1571-1577

Mario A. Fares, daniela Bezemer, Andres Moya, and Ignacio Marin, Selection on coding regions deter-mined HOX7 genes evolution, Molecular Biology and Evolution (2003) 20(12):2104-2112


116

oral presentations 2008 2nd NCHIV, Tropical Institute, Amsterdam, the Netherlands2008 15th International HIV Dynamics & Evolution Meeting, Santa Fe, USA2007 14th Conference on Retroviruses and Opportunistic Infections, Los Angeles, USA.

Attendance funded by a student travel award of the CROI (Abstract number: V-143), http://app2.capitalreach.com/esp1204/servlet/tc?c=10164&cn=retro&e=7000&s=20348&&display=0&day=28

2006 National STI&HIV Conference, Amsterdam, the Netherlands

poster presentations2009 16th Conference on Retroviruses and Opportunistic Infections, Montreal, Canada.

Attendance funded by a student travel award of the CROI (Abstract numbers: 817 and 1019)

2005 12th Conference on Retroviruses and Opportunistic Infections, Boston, USA. At-tendance funded by a student travel award of the CROI (Abstract number: 675)

2004 11th Conference on Retroviruses and Opportunistic Infections, San Francisco, USA Poster & Poster Discussion. Attendance funded by a student travel award of the CROI (Abstract number: 679)

2003 2nd IAS Conference on HIV Pathogenesis and Treatment, Paris, France (Abstract number: 782)

schools2002 Summer school in ‘Mathematical Biology’, Lisbon, Portugal. Attendance funded by

a student award of the organizing comity 2004 ‘Epidemiology of Infectious Diseases’ Imperial College London, UK2004 ‘Infectious Diseases’, Academic Medical Center, University of Amsterdam, the

Netherlands

Grants obtained2005 NWO – Department of infectious disease epidemiology, Imperial College Lon-

don, United Kingdom, Dr. C. Fraser2002 Preparing for PhD in evolutionary epidemiology at the Institute for Biodiversity

and Ecosystem Dynamics, University of Amsterdam, Prof. dr. M.W. Sabelis2000-2001 Socrates – Departemento de genetica, evolutiva, Instituto Cavanilles, University

of Valencia, Spain, Prof. dr. A. Moya1999 Erasmus – Departemento de genetica, evolutiva, Instituto Cavanilles, University

of Valencia, Spain, Prof. dr. A. Moya

currIculum vItae

117

degrees1999 Master’s Degree in biology obtained at Utrecht University, The Netherlands. Ful-

filled courses in Pattern Analysis, Bioinformatics, Prof. dr. P. Hogeweg, and Non-linear Differential Equations, Dr. A.V. Panvilov at the department of Theoretical Biology & Bioinformatics, and wrote final thesis on experimental evolution at the department of Evolution, Dr. G. de Jong.

1992 Secondary education at: De Drie Waarden in Schoonhoven (the Netherlands); Het Nieuwe Lyceum in Bilthoven (the Netherlands); Fagerborg Vidergående skole in Oslo (Norway)

employees October 2002 – 2006 GGD Amsterdam, the Netherlands, Prof. dr. R.A. Coutinho and Dr.

M. Prins, 3 years funded by grant 7014 from AIDS Fund Netherlands and 6 months by IBED/FNWI, UvA, Prof. dr. M.W. Sabelis

March 2006 – today Stichting HIV Monitoring, Amsterdam, the Netherlands, Prof. dr. F. de Wolf

1999 – 2000 Teaching science, mathematics and biology at several schools1998 Teaching assistant in ‘Introduction to Biology’ for first year biologists

at the University of Utrecht1996 Teaching assistant in ‘Mathematics’ (ODEs) for second year biologists

at the University of Utrecht

LanguagesDutch, Norwegian, English, Polish, Spanish, and notion of German and French

119

Acknowledgements

I thank everyone I collaborated with and all those who supported, inspired and motivated me to complete this thesis.

Hora est!

A resurgent HIV-1 epidemic among men who have sex with men in the era of potent antiretroviral therapy

Documents