WHO training, Pretoria, SA Jens D. Lundgren, MD, DMSc Director, Copenhagen HIV Programme (044) Hvidovre University Hospital, 2650 Hvidovre, Denmark ;

WHO training, Pretoria, SA Jens D. Lundgren, MD, DMSc Director, Copenhagen HIV Programme (044) Hvidovre University Hospital, 2650 Hvidovre, Denmark www.cphiv.dkwww.cphiv.dk; e-mail: [email protected]

Copenhagen HIV Programme Research unit at University of Copenhagen (located at Hvidovre University Hospital) Coordinating centre for: Randomized controlled trials (RCTs) COLATE, MaxCmin 1&2 NIH-sponsored: ESPRIT, SILCAAT, SMART Network: +300 clinics on 5 continents Cohort studies EuroSIDA (since 1994) Data collection of Adverse events of anti- HIV drugs (D:A:D) Network: +200 clinics on 3 continents

Agenda for 3 sessions ADR from ART examples The risk:benefit ratio of ART Methods to identify and understand AEs in addition to spontanous reporting Networking nationally and internationally Terminology ART=antiretroviral treatment ARVs=antiretrovirals (i.e. drugs used as part of ART)

Global effort to collect ADRs of ART WHO course, Pretoria, SA September 2004

Anti-HIV agents: 2004 Fusion inhibitors: T-20 (03) Integrase inhibitors: ? (03)

Main reasons of discontinuation of first HAART regimen within 1st year: ICONA ICO N A Italian Cohort Naive Antiretroviral Monforte et al. AIDS 1999

Side effects of anti-HIV drugs Early onset Varies by drug: GI, renal, CNS, rash, liver Late onset peripheral neuropathy osteopenia liver toxicity altered fat distribution elevated lactic acid levels diabetes mellitus lipid changes in blood (cardiovascular disease)

(AZT-nonAZT) difference (and 95% CI) of one-year change in haemoglobin Moyle et al, 4 th IWADRL, 2002 Differences in Haemoglobin (g/dl) at 1 year Oz-1 (n=105) Oz-2 (n=61) START I & II (n=301) BMS-148 (n=705) BMS-152 (n=491) Combined (n=1663)

Abacavir hypersensitivity reaction (HSR) Symptoms: Fever, rash, malaise Risk: From 1-8 (10) weeks from start. If HSR, exposure is associated with immediate death Presence of HSR and HLA-B*5701 status Mallal et al, Lancet 2002) : B*5701 pos: 14/18 (78%) B*5701 neg: 4/167 (2%) Reduction of prevalence of HSR by denying patients with HLA-B*5701, HLA-DR7, HLA- DQ3 abacavir: 9% to 2.5%

Time to initial grade 3 or 4 AE Proportion of subjects without a grade 3/4 AE Time (weeks) 0412243648 0.00 0.25 0.50 0.75 1.00 Saquinavir/r Indinavir/r Adverse events P = 0.0002 (log rank test) MaxCmin1: Dragsted et al, JID, 2003

Retinoid syndrome Nails deformation, hair loss, dry lips or skin, itchy skin, eczema or ulcers Assessment using a LDCD Study type questionnaire, i.e. both worsening and improvement of symptoms At Week 24 and 48 Patients and physicians assessment of improvements and worsening Cases defined at least moderate symptoms of retinoid worsening at one or more sites

Retinoid status at Week 48 N 242 Randomized Treatment Gr IDV/rtv SAQ/rtv (n=124) (n=120) p-value Cases (%) Non-cases (%) 98 (40%) 144 (60%) 76 (61%) 23 (19%) 48 (39%) 97 (81%) < 0.0001

The BEST Study: Treatment Arms TID group. Continue with: Indinavir 800 mg TID in combination with same 2 NRTIs BID group. Switch to: Indinavir 800 mg BID + Ritonavir 100 mg BID (liquid formulation) in combination with same 2 NRTIs Arnaiz et al, AIDS, 2003

Nephrolithiasis/haematuria: time to development

Lipoatrophy on arms Lipoatrophy on legs Increased abdominal fat (visceral) Mammae hypertrophy Lipoatrophy in face Buffalo hump Lipoatrophy in face Lipoatrophy on arms Lipoatrophy on legs Abnormal fat distribution

Both increased fasting and 2-hour insulin levels are evidence of insulin resistance in lipodystrophy P

Lactic acidemia lactic acidosisvenous lactate > 2 mmol/L +arterial ph 2 mmol/L grade oflactate acidosissymptoms mortality acidemia(mmol/L) (%) severe>10 often always80 moderate5 -10 rare usual 0 mild2 - 5 no sometimes 0 Terminology Risk & treatment 2-9 per 1,000 PY Stop ART time to clinical recovery 1-3 weeks (risk of relapse higher if restarting same drug combination)

Reversibility of symptomatic hyperlactatemia other NRTIs or NRTI sparing ? Symptomatic hyperlactatemia in TARHELL (d4T, n=16 to ZDV(4) or ABC (12) 1 At wk 48 (med.): -0.80 mmol/L Symptomatic hyperlactatemia at UCSD (d4T to ZDV or ABA, n=12) 2 At diagnosis: S-Lactate : 5.4 mM 1 relapse of symptomatic hyperlactatemia 2 discontinued due to unrelated reasons 9 remained asymptomatic after median 27 months S-lactate (med.) : 1.3 mM 1: Lonergan et al, 4 th IWADRL, 2002. Abs 21 2: Lonergan et al, 42 nd ICAAC, 2002. H-1080

Risk factors for femoral osteonecrosis (MRI): % of HIV+ patients with osteonecrosis PresentAbsentRR (95% CI) Lipodystrophy5%4%1.1 (0.4-2.9) Low testosterone12%4%3.2 (1.1-9.0) Syst. corticosteroids8%2%3.8 (1.3-11.0) Lipid lowering agents13%3%4.7 (1.8-11.9) Testosterone8%2%3.9 (1.3-11.6) Weights lifting7%2%3.3 (1.1-9.8) Prevalence: 15/339 (4.4%) in HIV+; 0/118 (0%) in HIV- (age, sex matched);p=0.02 Miller et al, AIM, 2002

The balance when assessing appropriate use of a treatment intervention Effect Toxicity GOOD BAD

AIDS rates EuroSIDA 1994 -2003 36 2.5 Mocroft et al, Lancet 2003

Changing population CD4 lymphocyte count in EuroSIDA CD4 count during period (/mm 3 ) Mocroft et al, Lancet, 2003

Risk of clinical disease progression by CD4 cell count at start of HAART Years from starting HAART 0123 0.75 0.80 0.85 0.90 0.95 1.00 0-99 100-199 200-349 >350 Egger et al, ART Cohort Collaboration, Lancet, 2002 Rate without AIDS or death

But this does NOT indicate that ART works less well in severely immunocompromised patients !!!

Predictive ability of pre-therapy CD4 cell count on risk of disease progression in ART-naive patients starting HAART # events 267 44 32 # w/CD4 count 237 29 23 Pre-therapy CD4 count (cells/L) Rate % SHCS, Frankfurt, EuroSIDA Phillips et al, JAMA, 2001 N=2742

1.5 1 2.0 0.67 0.5 >500 400- 300- 200- 100- < 100 499 399 299 199 Baseline CD4 count (per cubic millimeter) Relative hazard of viral load suppression < 500 c/mL within 32 weeks N=2742 SHCS, Frankfurt, EuroSIDA Phillips et al, JAMA, 2001

Weeks from viral load < 500 copies per milliliter Percent with viral load > 500 c/mL baseline CD4 count CD4 count > 350 CD4 count 200 - 349 CD4 count < 200 N.S. N=2346 SHCS, Frankfurt, EuroSIDA Phillips et al, JAMA, 2001

Differential diagnosis of clinical events developing in severely immunodeficient patients recently started on ART Further complications from pre-therapy impaired health status Still susceptible to opportunistic infections also after initiation Immune reconstitution syndrome Adverse events

Swiss HIV Cohort(*): Relative risk of different AIDS-defining events in 7/1997- 6/1998 versus 1992-4 *: 6.636 patients followed for 18.498 person- years Ledergerber et al, BMJ 1999;319:23

Systems complementary to spontaneous reporting

Enthusiasm for an agent as a function of time since first introduced Enthusiasm Time since initiation of phase I trials (years) CUREDOGREALISTIC Textbook in Pharmacology, 1960s

Enthusiasm for HAART as a function of time since first introduced Enthusiasm Time since initiation of phase I trials (years) 1996?

Toxicity - ways of detection Randomised trial: randomised phase open-label follow-up Passive surveillance Active survaillance: cohort studies

Why Randomization? Conscious and unconscious bias eliminated from treatment assignment Known and unknown confounders balanced on average Moderate treatment effects cannot be reliably established in the presence of moderate bias.

0.10.50.7511.251.51.75 Male health workers Social insurance, men Male chemical workers Hyperlipidaemic men Nursing home residents Social insurance, women Male physicians Male smokers (Ex)-smokers, asbestos workers Trials Cohorts Skin cancer patients USA Finland Switzerland USA USA Finland Finland USA USA USA Relative risk (95% CI) Egger et al. BMJ 1998 Beta-carotene intake and cardiovascular mortality

ONLY RANDOMISED TRIALS CAN RELIABLE DEFINE THE RISK:BENEFIT RATIO OF ART IN A GIVEN SETTING BUT IT IS NOT ALWAYS FEASIBLE TO DO THEM, OR THEY DO NOT ANSWER THE QUESTION !

Why are randomised trials not always able to provide the answers we are looking for ? Stopped when there is significant differences Ethically correct But, durability ? (ART has to continue for life) Use of laboratory endpoints (e.g. viral load) minimises duration and size of trial - result in rapid introduction of new drugs Snap-short of the entire duration of ART Not powered to detect differences in clinical meaningful outcomes related to benefit and risk from ART

0-2090.52(0.39 - 0.68) 1-3570.94(0.76 - 1.16) 2-4401.05(0.87 - 1.27) 3-3691.12(0.91 - 1.38) 4-3070.98(0.78 - 1.23) 5+2261.10(0.84 - 1.43) Pooled Analysis of Immediate vs. Deferred AZT No. AIDS/Death Events Hazard Ratio* Year of Follow-up *Immediate vs. deferred AZT

PI-HAART versus dual NRTI Therapy in Advanced Patients 0 - 61670.490.49 6 - 121410.330.41 12 - 181370.130.30 18 - 24940.150.26 24 - 30860.200.25 30 - 36540.160.24 No. AIDS/ Death Events Hazard Ratio* Interval of Follow-up (months) Interval Cum. *PI regimen vs. nRTIs adjusted for baseline CD4+

Toxicity - use of randomised trials BENEFITS Causal relationship can be evaluated Methodology ADR reporting well- developed PROBLEMS Size of population is relatively small - rare events Patient population is selected Randomised trials usually have a limited duration - long-term toxicity Assessment of drug under study - multiple combinations

Pivotal Phase 3 trial Hill et al, 4 th IWADRL

Toxicity missed in randomised trials Abnormal fat distribution 1995-97: Randomised trials evaluating efficacy/ toxicity of ART. Lipodystrophy not identified Feb. 98: First report, Carr et al. PIs is responsible 2002: ACTG 384 substudy: NRTIs responsible (PIs only play a minor role) Myocardial infarction 1998: Dyslipidaemia acknowledged 2002: Do not result in accelerated risk of myocardial infarction 2003: Do result in myocardial infarction

Lipodystrophy AIDS 1998, 12: F51-F58

Other options when RCT are not able to provide the relevant answer Expert opinion used in marked research Other sources of data: Case reports Passive surveillance Cohort studies

Relative importance: summary of experience in last 8 years personal perspective Randomised controlled trials Early onset, frequent adverse events Cohort studies Complemented findings in RCTs Rare early onset and late onset adverse events Spontanous reporting/passive surveillance Confusion Perscription studies None Case reports & expert opinion Confusion

Cohort group of patients: the number aint the only relevant characteristic Prospective or retrospective Enrolment criteria Which data are collected ? How are data collected ? Which quality control measures are utilized Power to detect the outcome being investigated

EuroSIDA - data collection Consecutive patients New cohorts added every 2 year - refreshment Routine outpatient clinic appointment Age > 16 (cohort I-III: CD4 < 500/mm 3 ) Every 6 months (June, December) Data collection form format adjustable Data check At site: check of computerised data (preprinted) At coordinating centre: data entry queries site visits

EuroSIDA Cohorts I-V Cohort I n = 3116 May 1994 Cohort II n = 1365 1996 Cohort III n = 2839 1997 Cohort IV n = 1225 1999 Cohort V n = 1257 2001 EuroSIDA n = 9802 72 hospitals in 24 European countries + Israel and Argentina Cohort VI started in November 2003 (additional 1,300 patients)

Surveillance of emerging adverse events outside of RCT Rare early and all late-onset adverse events Identification Case description of phenomenon Biological plausibility Cohort studies Requires open-ended questions Not feasible in larger cohort studies

Quality versus quantity Volume of questions/ work required Quality of data

In cohort studies it is not important to collect all sorts of information BUT rather focus on collecting the information required for the need of the cohort and ENSURE THAT THE QUALITY OF THE DATA IS GOOD (garbage in=garbage out)

Role of large prospective cohort studies for emerging adverse events Study a priori identified signals Although methods exist to use large cohorts to identify signals not suspected previously (discussed later) Assess association with drug classes or individual drugs Quantify risk in subgroups of patients

Inclusion: selection External validity extrapolation Active recruitment versus extraction from databases developed for other reasons Consecutive versus non-consecutive Retrospective studies ! Study of trends over time the addition of new patients

Risk of AE as function of time since starting the drug: More on selection bias !! Enrolment: Drug nave cohort Complete assessment of risk Drug experienced cohort AEs may be missed (if they occurs prior to time when patient enters cohort) Cohort still on drug can tolerate the drug Biological mechanism of how AE may develop may assist in making rational assessment of whether a cohort is suitable to assess risk of a certain AE

Identification of a potential toxicity with a late onset using cohort data Incidence of potential toxicity Time from initiation of therapy or follow-up Initiated therapy Not Initiated therapy

Incidence of adverse event Person-year of follow-up # Events

If adverse event is late onset If incidence is calculated On versus of drug % of patients on drug followed prior to biologically plausible onset of adverse event Time intervals since starting drug Define time lag Ability to detect adverse event Time lag versus total exposure time per patient

Event: what is possible and how collect Ascertainment (within a population, who developed the AE and who did not) Case definition Objectively documented Reliable picked up in the patient record notes Quality control source documentation Collected prospectively or retrospectively Prospectively: allows for training and proper work-up awareness high Retrospective: awareness variable Source verification Competing risks HIV-related (e.g. chest pain) Co-morbidities, eg CVD (next slide)

Power Risk of type I error (study detect a difference that is not there in reality) Risk of type II error (study did not detect a difference that is there in reality) Formulate hypothesis prior to launch study/analysis Stipulate what difference is acceptable to be missed

Co-morbidities as adverse events: noise or true problem ? Adverse event/background risk ratio ! Characteristics of the cohort followed Background risk low: unusual high rate, but requires many patients Background risk high: signal may be missed An independent effect associated with drugs Requires the collection of all important risk factors for the co-morbidity

Lost-to-follow-up Should be low ! Is health situation (for the parameter evaluated) for those lost better or poorer than for those remaining ? Emigration versus transferral to hospice Organisation of health system: Single - centralised Plural Private insurance organisations Government supported programs Ability to follow patients switching program

Principal for working: think outside and work within the box Einsteins definition of insanity: repeating the same experiment over and over, and expecting different results

Critical criteria for a successful observational study Quality of data (garbage in = garbage out) Limit the volume of data to be collected to critical important items Describe what you want to achieve prior to launch Allow for flexibility while is ongoing Standardized case record form (with the flexibility of additional items in the future) Reciprocal quality control: Data already in the database should be available for review clinical site staff On-site training of staff Dynamic & ongoing dialogue between clinicians and epidemiological and statistical functions to ensure Timely extract of clinical relevant information Optimise engagement by entire study team

Prognosis without HAART 3-year probability of AIDS in 1604 men enrolled in the Multicenter AIDS Cohort Study (MACS) 1984-1985 Viral load >60,00020 - 60,000 6 - 20,000 1 - 5,000

WHO training, Pretoria, SA Jens D. Lundgren, MD, DMSc Director, Copenhagen HIV Programme (044) Hvidovre University Hospital, 2650 Hvidovre, Denmark ;

Documents