UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)
UvA-DARE (Digital Academic Repository)
Ageing with HIVFrom pathogenesis to policyvan Zoest, R.A.
Link to publication
Creative Commons License (see https://creativecommons.org/use-remix/cc-licenses):Other
Citation for published version (APA):van Zoest, R. A. (2019). Ageing with HIV: From pathogenesis to policy.
General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, statingyour reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Askthe Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,The Netherlands. You will be contacted as soon as possible.
Download date: 14 Jun 2020
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 177PDF page: 177PDF page: 177PDF page: 177
CHAPTER 8CARDIOVASCULAR DISEASE PREVENTION POLICY IN HUMAN IMMUNODEFICIENCY VIRUS: RECOMMENDATIONS FROM A MODELING STUDY
Mikaela Smit*Rosan A van Zoest*Brooke E NicholsIlonca VaartjesColette SmitMarc van der ValkArd van SighemFerdinand W WitTimothy B HallettPeter Reiss
On behalf of the Netherlands AIDS Therapy Evaluation inThe Netherlands (ATHENA) Observational HIV Cohort
Clinical Infectious Diseases 2018; 66: 743-750.doi: 10.1093/cid/cix858.
* both authors contributed equally
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 178PDF page: 178PDF page: 178PDF page: 178
| Chapter 8170
ABSTRACT
Background
Cardiovascular disease (CVD) is expected to contribute a large non-communicable
disease burden among human immunodeficiency virus (HIV)-infected people. We
quantify the impact of prevention interventions on annual CVD burden and costs
among HIV-infected people in the Netherlands.
Methods
We constructed an individual-based model of CVD in HIV-infected people using
national ATHENA (AIDS Therapy Evaluation in the Netherlands) cohort data on
8,791 patients on combination antiretroviral therapy (cART). The model follows
patients as they age, develop CVD (by incorporating a CVD risk equation), and start
cardiovascular medication. Four prevention interventions were evaluated: (1) increasing
the rate of earlier HIV diagnosis and treatment; (2) avoiding use of cART with increased
CVD risk; (3) smoking cessation; and (4) intensified monitoring and drug treatment of
hypertension and dyslipidemia, quantifying annual number of averted CVDs and costs.
Results
The model predicts that annual CVD incidence and costs will increase by 55% and 36%
between 2015 and 2030. Traditional prevention interventions (ie, smoking cessation
and intensified monitoring and treatment of hypertension and dyslipidemia) will avert
the largest number of annual CVD cases (13.1% and 20.0%) compared with HIV-
related interventions—that is, earlier HIV diagnosis and treatment and avoiding cART
with increased CVD risk (0.8% and 3.7%, respectively)—as well as reduce cumulative
CVD-related costs. Targeting high-risk patients could avert the majority of events and
costs.
Conclusions
Traditional CVD prevention interventions can maximize cardiovascular health and
defray future costs, particularly if targeting high-risk patients. Quantifying additional
public health benefits, beyond CVD, is likely to provide further evidence for policy
development.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 179PDF page: 179PDF page: 179PDF page: 179
Cardiovascular Prevention Policy in HIV | 171
8
BACKGROUND
As life expectancy of people living with human immunodeficiency virus (PLHIV)
on effective combination antiretroviral therapy (cART) has improved (1), human
immunodeficiency virus (HIV) care focus is shifting from treatment and prevention of
opportunistic infections to that of noncommunicable diseases (NCDs). Cardiovascular
disease (CVD) is predicted to contribute one of the greatest NCD burdens among
PLHIV on cART (2). CVD prevalence and lifetime CVD risk appear to be higher in
PLHIV than in HIV-uninfected controls (3–5).
Although cardiovascular pathophysiology in the HIV context is not yet fully
understood, likely mechanisms include complex interactions between traditional (eg,
smoking), and HIV-related (eg, greater HIV-related inflammation linked to cART
initiation at low CD4 counts) risk factors (6,7). Late HIV diagnosis and thus low
CD4 count at cART initiation remains a problem in many European countries (8,9),
despite new HIV guidelines recommending immediate cART initiation (10). The
prospective observational Data Collection on Adverse Effects of Anti-HIV Drugs
(D:A:D) Study carried out one of the most detailed analysis of modifiable CVD risk
factors in >49,000 PLHIV from 11 cohorts across 212 clinics in Europe (including
the Netherlands), Argentina, Australia, and the United States, combining them into
a risk model predicting 5-year CVD risk (11). These factors include smoking, total
and high-density lipoprotein (HDL) cholesterol, systolic blood pressure, current CD4
count, current abacavir use, and cumulative exposure to protease inhibitors (PIs) and
nucleoside reverse transcriptase inhibitors, suggesting these could be key targets for
CVD prevention interventions in HIV care.
Prevention interventions will need to be firmly integrated into HIV care, and supported
by evidence-based studies to effectively mitigate the emerging CVD burden, mortality
risk, and impact on quality of life of PLHIV. Yet no studies have systematically
compared the impact of different CVD prevention interventions on CVD burden in
PLHIV, who are at CVD risk due to both traditional and HIV-related factors. We
adapt an existing model of aging PLHIV on cART (2) to explore the impact of CVD
prevention interventions and evaluate which one would maximize cardiovascular
health. Interventions include HIV-related interventions (earlier HIV diagnosis and
treatment and avoiding use of cART regimens with increased CVD risk) and traditional
interventions (smoking cessation and intensified monitoring and drug treatment of
hypertension and dyslipidemia).
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 180PDF page: 180PDF page: 180PDF page: 180
| Chapter 8172
METHODS
Model
We adapted an existing individual-based model of aging PLHIV in the Netherlands to
generate detailed predictions of CVD (2). Figure 8.1 shows the basic model structure,
with technical details in Supplement 8.1. The model follows PLHIV on cART from
2010 to 2030, and probabilistically simulates clinical events (hypertension, dyslipidemia,
diabetes, CVD, and death) and CVD treatment initiation. CVD includes myocardial
infarction, stroke, coronary artery angioplasty, coronary bypass, carotid endarterectomy,
and CVD-related death.
The model is populated with all 8,791 PLHIV on cART in care and registered with
the Dutch national ATHENA (AIDS Therapy Evaluation in the Netherlands) cohort
in 2010 aged ≥18 years, infected with HIV type 1 (HIV-1) only, cART naïve at entry,
who initiated cART from 1996 (baseline patients, Figure 8.1A) by directly importing
anonymized patient records into the model. Multiple imputation by chained equation
was used to deal with missing data (MICE package, R Foundation for Statistical
Computing, version 3.2.2), creating 20 imputed datasets. Values were assumed to be
randomly missing, and extensive checks were carried out (Supplement 8.1).
The majority of ATHENA patients were virologically suppressed Dutch men who have
sex with men, and median age was 44 years (Table 8.1). In 2010, 44.9% of moderate-
to high-risk (HR, ie, predicted 5-year CVD risk ≥5%) patients with hypertension
used antihypertensive medication, of whom 7.9% reached blood pressure levels of
<120/80 mmHg. In HR patients with dyslipidemia, 32.5% used lipid-lowering
medication, of whom 61.1% reached a total/HDL cholesterol ratio of <5 (Table 8.1).
All clinical definitions were based on European guidelines (10,12) and consultation
with cardiovascular specialists; hypertension was defined as systolic blood pressure ≥140
mmHg, diastolic blood pressure ≥90 mmHg, and/or use of antihypertensive medication;
dyslipidemia was defined as total cholesterol >6.5 mmol/L (>250 mg/dL), total/HDL
cholesterol ratio >5, and/or use of lipid-lowering medication, and CVD was based on
clinical patient records.
Each month, new simulated patients joined the modeled cohort representing cART
initiators (Figure 8.1A). New patients’ profiles were randomly selected from baseline
patients, accounting for predicted changes in age up to 2030, CD4 cell count at cART
initiation up to 2015, and changes in treatment guidelines in 2014 (use of integrase
inhibitors for first-line cART) (2,13,14).
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 181PDF page: 181PDF page: 181PDF page: 181
Cardiovascular Prevention Policy in HIV | 173
8
2010 2030
+ 1 month
Model check Interventions
2010-2015 2017-2030
ATHENA
cART
Icons made by Freepik from www.flaticon.com
D:A:D risk equation (Friis-Møller et al 2016)
DemographicAge
GenderClinicalDiabetes
Family historySmoking (current/former)
Total cholesterolHDL cholesterol
Systolic blood pressureHIV-related
Current CD4 countAbacavir use
Cumulative PI exposureCumulative NRTI exposure
DEMOGRAPHICS
CLINICAL FACTORS
HIV-RELATED FACTORS
CARDIOVASCULAR MEDICATION
A
B C
FIGURE 8.1. MODEL DESIGN. A, Illustrates the basic model structure. The model starts on 1 January 2010 with all human immuno-
deficiency virus (HIV)–infected individuals on combination antiretroviral therapy (cART) in care and
registered with the ATHENA cohort aged ≥18 years old, infected with HIV-1 only, cART naive prior
to entering the cohort, who had initiated cART after 1 January 1996 (depicted in red). Each month, the
model simulated calendar events (eg, aging), new events in a probabilistic manner (eg, deaths and new
cardiovascular events), and new patients are added as they start cART (depicted in blue). When the
model stops on 31 December 2030, the modeled patient cohort contains all patients from 2010 who are
still alive as well as new patients who have entered care between 1 January 2010 and 31 December 2030
and who are still alive. The period from 2010 to 2015 is used to carry out out-of-sample model validation,
and interventions targeting cardiovascular disease are simulated between 2017 and 2030. B, Shows the
risk factors included in the D:A:D risk equation. C, Illustrates the interactions simulated in the model
between demographic (red), HIV-related (green), cardiovascular treatment (yellow), and clinical factors
(blue), for example, how the risk of death is determined by age, sex, and clinical factors. The dashed lines
show interactions between individual factors and solid lines between a category and a specific risk factor.
Abbreviations: ART, antiretroviral therapy; ATHENA, AIDS Therapy Evaluation in The Netherlands;
cART, combination antiretroviral therapy; CVD, cardiovascular disease; D:A:D, Data Collection on Ad-
verse Effects of Anti-HIV Drugs; HDL, high-density lipoprotein; HIV, human immunodeficiency virus;
NRTI, nucleoside reverse transcriptase inhibitor; PI, protease inhibitor.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 182PDF page: 182PDF page: 182PDF page: 182
| Chapter 8174
TABLE 8.1. COHORT CHARACTERISTICS OF HUMAN IMMUNODEFICIENCY VIRUS–INFECTED INDIVIDUALS ON COMBINATION ANTIRETROVIRAL THERAPY, ALIVE AND ENROLLED IN THE ATHENA COHORT ON 1 JANUARY 2010
Characteristic
ATHENA
Cohort Populationa
(n = 8,791)
Demographic characteristics
Age, years, median (IQR) 43.8 (37.4-50.6)
Male sex 6,851 (77.9%)
HIV transmission risk categoryb
Men who have sex with men 4,619 (52.5%)
Heterosexual 3,183 (36.2%)
Injecting drug user 137 (1.6%)
Other or unknown 851 (9.7%)
Region of originb
Netherlands 4,901 (55.8%)
Sub-Saharan Africa 1,583 (18.0%)
Europe 676 (7.7%)
Other 1,630 (18.5%)
Smoking status b
Never smoker 1,302 (23.2%)
Ex-smoker 3,217 (57.4%)
Current smoker 1,090 (19.4%)
Diabetes prevalenceb 373 (4.2%)
Cardiovascular treatment in patients with moderate to high CVD risk
(predicted 5-y CVD risk ≥5%)
(n=2,895)
Hypertensionc 1,469 (50.7%)
Receiving antihypertensive medication 660 (44.9%)
Blood pressure < 120/80 mmHg among those receiving antihypertensive medication 52 (7.9%)
Dyslipidemiad 1,521 (52.5%)
Receiving lipid-lowering medication 494 (32.5%)
Total/HDL cholesterol ratio ≤5 among those receiving lipid-lowering medication 302 (61.1%)
HIV-related characterstics (n=8,791)
Time on combination antiretroviral therapy, years, median (IQR) 4.1 (1.6-7.5)
HIV RNA <200 copies/mL among cART-treated patients in the year prior to enrollmentb 7,477 (86.7%)
Prior diagnosis of AIDS 2,394 (27.2%)
Nadir CD4 cell countb, cells/μL, median (IQR) 180 (74-270)
Current CD4 cell count, cells/μL, median (IQR) 500 (362-660)
Data are presented as No. (%) unless otherwise indicated.Abbreviations: ATHENA, AIDS Therapy Evaluation in The Netherlands; cART, combination antiretroviral therapy; CVD, cardiovascular disease; HDL, high-density lipoprotein; HIV, human immunodeficiency virus; IQR, interquartile range.aThe model is populated with all 8,791 HIV-infected individuals on cART in care and registered with the ATHENA cohort aged ≥18 years old, infected with HIV-1 only, cART naive prior to entering the cohort, who had initiated cART after 1 January 1996.bData for some individuals were missing on 1 January 2010 and hence not included in this table.cHypertension was defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, and/or use of antihypertensive medication.dDyslipidemia was defined as total cholesterol >6.5 mmol/L (250 mg/dL), total/HDL cholesterol ratio >5, and/or use of lipid-lowering medication.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 183PDF page: 183PDF page: 183PDF page: 183
Cardiovascular Prevention Policy in HIV | 175
8
CVD was simulated using a CVD risk equation, developed in PLHIV by the D:A:D
study (11). The equation accounts for demographic factors, traditional risk factors, and
HIV-related risk factors to calculate 5-year CVD risk (Figure 8.1B). Constant risk
was assumed within the 5-year timeframe to convert to monthly risks. CVD type was
determined using the reported CVD distribution (11).
The model accounts for a large number of interactions between demographic (Figure
8.1C), HIV-related (Figure 8.1C), therapy-related (Figure 8.1C), and clinical factors
(Figure 8.1C) – for example, the impact of age, sex, and clinical factors on mortality, or
sex, age, use of lipid-lowering drugs, and cART regimen on cholesterol. For example,
the model includes sex-specific monthly increments in blood pressure and cholesterol
per 10-year age group based on the Dutch Doetinchem Cohort Study data (15),
cholesterol changes triggered by cART switches based on an extensive literature review,
and blood pressure and cholesterol changes triggered by treatment based on ATHENA
data (Supplement 8.1).
Model performance was evaluated via out-of-sample checks against 2010-2015
ATHENA data (Supplement 8.1). CVD burden was projected for 2010-2030,
assuming demographic, epidemic, and clinical trends remain constant while accounting
for predicted increasing CD4 count trends at cART initiation, and current levels in
treatment of hypertension and dyslipidemia (baseline scenario). The impact of CVD
prevention interventions was evaluated for 2017-2030, by comparing with baseline.
Model results were based on the average of 100 simulations, consisting of running each
of the 20 imputed dataset 5 times.
The model was used to evaluate the following interventions:
HIV-related interventions:
1. Increasing the rate of earlier HIV diagnosis and treatment (ie, initiating cART at
CD4 counts ≥500 cells/μL)
2. Avoiding use of cART regimens with increased CVD risk (as defined by the D:A:D
risk equation, ie, switching off/not initiating abacavir [ABC] and PIs)
Traditional CVD interventions:
1. Increasing smoking cessation
2. Intensified monitoring and treatment of hypertension and dyslipidemia with
antihypertensive and lipid-lowering drugs (to ensure patients achieve target blood
pressure [<120/80 mmHg] and cholesterol levels [total/HDL cholesterol ratio of
≤5]).
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 184PDF page: 184PDF page: 184PDF page: 184
| Chapter 8176
In addition, we evaluated the effect of implementing all interventions jointly. For each
intervention, we simulated the impact of reaching (1) all patients in care or HR patients
only (except reducing “late” presenters, which was not evaluated for HR patients
separately); or (2) 100% or 50% of patients successfully via the interventions.
For example, the use of cART regimens with no known increased CVD risk at
50% successful implementation switched 50% of patients off ABC and PIs, while
the remainder continued on their cART regimens, including ABC and PIs. For
the intensified monitoring and treatment of hypertension and dyslipidemia at 50%
successful implementation, all patients with hypertension and dyslipidemia were
prescribed treatment, but only 50% reached target blood pressure and cholesterol levels,
while the remainder reached blood pressure and cholesterol levels currently observed in
ATHENA among patients on antihypertensive and lipid-lowering medication for ≥6
months (ie, some patients do not reach target levels and some do).
Economic evaluation
An economic evaluation quantified the annual costs of CVD care in PLHIV between
2015 and 2030. In addition, annual CVD-related cost reductions were calculated for
each intervention in 2017-2030. Economic evaluations accounted for lifetime CVD-
related costs incurred; these comprised costs related to hospitalization, diagnostics,
surgical procedures, and outpatient visits. Costs relating to intervention costs of
medication were not included in the analysis. Cost sources included Dutch databases of
healthcare, and literature, and costs were discounted at 3% per year (Supplement 8.1).
RESULTS
Predicted CVD burden and effect of prevention interventions
Annual CVD incidence is predicted to increase from 6.0 to 9.3 events per 1,000 person
years between 2015 and 2030. Prevalence of ever having had CVD will increase from
2.7% to 11.4%, and the proportion of HR patients from 14.1% to 32.4%, between 2015
and 2030 (Figure 8.2).
All interventions tested are expected to reduce CVD burden by 2030. Traditional CVD
interventions are predicted to have greater impact, and targeting HR patients would
avert a majority of CVD cases (Figure 8.3). Earlier HIV diagnosis and treatment will
have the smallest effect on CVD burden, preventing on average 0.1%-0.8% of CVD
cases annually, assuming 50%-100% successful implementation. Avoiding the use of
cART regimens with increased CVD risk could avert an average of 1.7%-3.7% of CVD
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 185PDF page: 185PDF page: 185PDF page: 185
Cardiovascular Prevention Policy in HIV | 177
8
cases annually in all patients and 0.3%-1.5% of CVD cases in HR patients, assuming
50-100% successful implementation.
FIGURE 8.2. PREDICTED CARDIOVASCULAR DISEASE (CVD) RISK BETWEEN 2010 AND 2030. The proportion of human immunodeficiency virus–infected individuals on combination antiretroviral
therapy in low (<5%, green), moderate (5%–10%, yellow), and high (≥10%, red) predicted 5-year CVD
risk groups between 2010 and 2030 as simulated by the model using the D:A:D risk equation.
Smoking cessation could prevent an average of 6.0%-13.1% of CVD cases annually
in all and 5.1%-12.6% of CVD cases in HR patients, assuming 50%-100% successful
implementation. Intensified monitoring and treatment of hypertension and dyslipidemia
could have the greatest impact on future CVD burden, averting an annual average of
17.0%-20.0% of CVD cases in all and 11.9%-15.9% of CVD cases in HR patients,
assuming 50%-100% successful implementation.
Implementing all interventions simultaneously could avert an average of 20.2%-30.5%
of CVD cases annually in all, and 13.1%-20.4% of CVD cases in HR patients, assuming
50%-100% successful implementation. Targeting HR patients only is likely to reduce
the majority of CVD cases, particularly in smoking cessation, intensified monitoring
and treatment of hypertension and dyslipidemia, and a joint intervention program
(averting 96%, 80%, and 67%, respectively, compared with targeting all patients when
assuming 100% successful implementation).
Annual CVD-related care costs for PLHIV are estimated to increase by 36%, from
€471,000 in 2015 to €642,000 in 2030, in absence of additional CVD prevention
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 186PDF page: 186PDF page: 186PDF page: 186
| Chapter 8178
FIGURE 8.3. AVERAGE ANNUAL PERCENTAGE OF AVERTED CARDIOVASCULAR DISEASE (CVD) CASES. Averages per intervention are calculated based on annual reductions in CVD cases as simulated by the
model between 2017 and 2030 in all patients (A) and interventions limited to those at moderate to
high CVD risk only (B). Interventions include (1) reducing the number of “late” presenters (ie, initiate
combination antiretroviral therapy [cART] at CD4 counts >500 cells/μL); (2) use of cART with no
known increase in CVD risk; (3) smoking cessation; (4) intensified monitoring and drug treatment
of hypertension and dyslipidemia; (5) joint intervention, in which all interventions are implemented
simultaneously. Reduction of late presenters was only evaluated in all patients, not in high-risk patients
separately. The dark-colored area represents percentage averted assuming 50% successful implementation
and the light-colored area for 100% successful implementation. Abbreviations: cART, combination
antiretroviral therapy; CVD, cardiovascular disease; HIV, human immunodeficiency virus.
0
5
10
15
20
25
30
35
Earlier HIVdiagnosis and
treatment
Avoiding cARTwith
increased CVDrisk
Smokingcessa on
Monitoring/treatment ofhypertension
anddyslipidaemia
Jointinterven on
Ave
rage
ann
ual p
erce
ntag
e re
duc
on in
CVD
ca
ses
A
0
5
10
15
20
25
Avoiding cARTwith increased
CVD risk
Smokingcessa on
Monitoring/treatment ofhypertension
anddyslipidaemia
Jointinterven on
Aver
age
annu
al p
erce
ntag
e re
duc
on in
CVD
ca
ses
B
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 187PDF page: 187PDF page: 187PDF page: 187
Cardiovascular Prevention Policy in HIV | 179
8
interventions. This is despite HIV costs being predicted to decrease by 13% in the same
period. Cumulative CVD-related costs in 2017-2030 could decrease by €63,000 (0.8%)
with earlier HIV diagnosis and treatment, €308,000 (3.7%) by avoiding the use of
cART regimens with increased CVD risk, €1,093,000 (13.1%) with smoking cessation,
€1,670,000 (20.0%) with intensified monitoring and treatment of hypertension and
dyslipidemia, and €2,539,000 (30.5%) with a joint intervention, assuming all patients
are targeted 100% successfully. The reduction in CVD-related costs is largely driven
by HR patients. Cumulative CVD-related costs, for example, could be reduced by
€1,330,000 for intensified monitoring and treatment of hypertension and dyslipidemia,
or 80% of the costs reduction achievable by targeting all PLHIV.
DISCUSSION
The growing burden of CVD among PLHIV in Europe will necessitate evidence-based
prevention interventions to mitigate the impact on mortality, quality of life, clinical care,
and healthcare budgets. The interventions considered in this study could jointly avert
30.5% of CVD events. Based on the output of this work, we recommend that CVD
prevention policy in HIV care focus on traditional CVD prevention interventions,
particularly intensifying the monitoring and successful treatment of hypertension and
dyslipidemia. Such an intervention could most likely maximize cardiovascular health and
reduce CVD-related costs among the set of CVD prevention interventions evaluated
in this study. Targeting HR patients could avert the majority of CVD events and costs,
averting 80% of CVD events and costs between 2017 and 2030, while targeting only
14.1%-32.4% (2015-2030) of PLHIV. Greater impact could be achieved if rolling out
a program implementing all interventions simultaneously and targeting all PLHIV,
though likely at much higher expected programmatic costs.
Our findings are in line with preliminary work in PLHIV showing that dyslipidemia,
hypertension, and smoking had much larger population attributable fractions for
myocardial infarctions than HIV-related factors (16). Intensification of monitoring
and treatment of hypertension and dyslipidemia should be feasible, as measurement
and management of blood pressure and cholesterol are standard in HIV care in the
Netherlands. Current European and Dutch HIV guidelines recommend annual
CVD risk assessment and treatment of hypertension (in patients with high CVD
risk) and dyslipidemia (in patients with moderate to high CVD risk) (10,17). CVD
risk assessment, utilizing for example the D:A:D equation, provides a simple and
inexpensive tool to identify patients at elevated CVD risk. Yet, in 2010, 55%-68% of
Dutch PLHIV at moderate to high CVD risk remain untreated and the majority did
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 188PDF page: 188PDF page: 188PDF page: 188
| Chapter 8180
not reach optimal blood pressure or cholesterol levels (Table 8.1). Suboptimal CVD risk
management has been described in other settings (18,19). The proposed intervention
would therefore focus on strengthening implementation of current HIV guidelines,
utilizing existing frameworks, and addressing gaps including undiagnosed and/or
undertreated hypertension and dyslipidemia. Further research is needed to identify
innovative strategies to support patients and healthcare providers in addressing these
gaps, for which mobile phone- or web-based interventions might be promising tools.
Smoking cessation is also predicted to substantially impact CVD burden. However,
our assumption that 50%-100% of individuals will successfully cease to smoke is bold
considering previous smoking cessation studies (20–23). In sensitivity analyses of
more realistic estimates (ie, 10% smoking cessation) an average of 1.3% of CVD cases
annually were averted in all PLHIV (Supplement 8.1). The low success rates of smoking
cessation despite tailored interventions and the modest effect of a more realistic smoking
cessation scenario on CVD burden means smoking cessation might not be the most
effective and feasible CVD prevention intervention, despite its large theoretical impact.
Studies suggest that apart from their lipid-lowering properties, statins may also act as
inflammation modulators (24). In PLHIV, in whom immune activation/dysfunction
can persist despite viral suppression, studies have shown that statins can reduce certain
markers of inflammation and immune activation (25) as well as subclinical CVD (26).
The Randomized Trial to Prevent Vascular Events in HIV (REPRIEVE) trial (27) may
quantify the effectiveness of statins in preventing CVD in PLHIV, and inform further
modeling studies in the field.
The model captures key factors involved in CVD pathophysiology. Extensive
consultations with HIV and CVD physicians ensure the model incorporated key
interactions between factors influencing CVD, ensuring the model captured the natural
progression and aggregation of risk factors. The model created a direct representation
of patients on cART in the Netherlands by directly importing patient records from the
large national non-selective cohort, with parameter values based on large cohort studies
and in-depth literature reviews. Extensive out-of-sample model checks and sensitivity
analyses ensured that the model output was robust.
Despite these strengths, the study has a number of limitations. Some ATHENA data
(eg, family history, smoking status) are collected at entry and not systematically time-
updated. These and other variables (eg, cholesterol levels) can also be incomplete. We
used multiple imputation to handle missing data, assuming variables were missing
randomly; any heterogeneity could introduce bias of the results. Other interventions,
including interventions targeting lifestyle and diet, were not evaluated. These risk
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 189PDF page: 189PDF page: 189PDF page: 189
Cardiovascular Prevention Policy in HIV | 181
8
factors are historically poorly collected in HIV cohorts and, although their importance
is not disputed, consensus on their relative impact in the context of HIV is yet to be
reached. As information about their relative impact becomes available, the model can
be expanded to evaluate lifestyle interventions. In the absence of robust data in PLHIV,
the model assumes that changes in blood pressure and cholesterol with age are the same
in PLHIV and the general population. Although uncertainty around the superiority of
various CVD risk equations remains (28–31), the D:A:D equation is likely among the
superior tools for CVD risk prediction in PLHIV in Europe. It is the only one specifically
developed in PLHIV, taking into account HIV-related factors. Yet the D:A:D equation
does not distinguish between well and poorly controlled diabetes, nor between CVD
risk of individual PIs (cumulative exposure to all PIs predicted CVD risk better than
individual PIs, despite preliminary work describing that not all contemporarily used
PIs are independently associated with an increased CVD risk (32)). We were therefore
unable to evaluate interventions centred around diabetes, or individual PIs. In addition,
the D:A:D equation does not account for the long-term ‘fading’ effect of stopping PIs
or smoking on CVD risk, although we tested this in sensitivity analyses (Supplement
8.1), with results indicating that the model results remain robust. Additionally, we
have presented the amount of CVD-related costs that could be averted if a program
is 100% successful. It is important to note that these figures do not take into account
any intervention-related costs, as these costs are too uncertain to estimate reliably for
all interventions, as well as have impact beyond CVD. It is likely that many of these
interventions will actually incur costs once programmatic costs are taken into account.
In the model dyslipidemia was defined using total/HDL cholesterol ratio due to poorly
collected low-density lipoprotein cholesterol data. While this might not fully represent
management and recommendations in clinical care, we believe it to be a reasonable
proxy. While 100% successful implementation scenarios served to illustrate the
maximum intervention impact, even 50% successful implementation scenarios could be
overly optimistic – for example, for smoking cessation, with well-documented difficulty
of long-term smoking cessation in PLHIV (20–23).
While the focus of this study is on CVD prevention, it is likely that all interventions
investigated in this study could have wider public health benefits. For example, smoking
cessation could reduce the incidence of other diseases including lung cancer (33),
other malignancies (34), chronic obstructive pulmonary disease (35), and peripheral
atherosclerotic disease (36). Furthermore, reducing “late” presenters could prevent
AIDS, HIV transmission, and NCDs other than CVD (7). Systematically comparing
these interventions with a wide public health perspective, including a cost-effectiveness
analysis, could provide further evidence regarding policy development.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 190PDF page: 190PDF page: 190PDF page: 190
| Chapter 8182
We expect the results to be generalizable to other settings with a long history of access
to cART and a concentrated and mature HIV epidemic. It is not clear how the results
would translate to epidemics with larger proportions of female patients. In settings such
as sub-Saharan Africa, which is also seeing PLHIV aging (37), there is an urgent need
for country-specific estimates of future CVD and NCD burdens to identify effective
intervention and best use of resources, particularly in light of overstretched healthcare
systems.
The CVD burden among PLHIV on cART in Europe is increasing. Intensified
monitoring plus successful treatment of hypertension and dyslipidemia in PLHIV is
expected to be the most feasible intervention accompanied by the largest cardiovascular
health benefit and could safeguard the quality of life of HIV-infected people.
ACKNOWLEDGEMENTS
The authors acknowledge A. Blokstra, H.S.J. Picavet, and W.M.M. Verschuren for
providing access to Doetinchem Cohort Study data, and Luuk Gras for his advice in
analysing changes in CD4 count with time on cART.
FUNDING
The work was supported by the Stichting HIV Monitoring, which is funded by The
Netherlands Ministry of Health, Welfare and Sport through the Center for Infectious
Disease Control of the National Institute for Public Health and Environment. The
funder had no role in the analysis or the decision to publish.
PRESENTED IN PART
This study was presented in part at the 20th International Workshop on HIV and
Hepatitis Observational Databases, Budapest, Hungary, April 2016, and the Conference
on Retroviruses and Opportunistic Infections, Seattle, Washington, United States, 13-
16 February 2017.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 191PDF page: 191PDF page: 191PDF page: 191
Cardiovascular Prevention Policy in HIV | 183
8
AUTHORS’ CONTRIBUTIONS
MS formulated the research question, constructed the original model and made all
modifications to the adapted model, analyzed data for model parameterization,
generated the model output, interpreted the results, and co-authored the first draft of
the manuscript. RAvZ contributed to the research question and method development,
carried out data analyses for model modification and parameterization, carried out all
imputations, advised on the scientific and medical aspects of the model adaptation
and interpretation of the results, and co-authored the first draft of the manuscript.
BEN carried out the economic evaluations. CS prepared the ATHENA cohort data
for model population and parameterization. IV advised on the model imputation and
with MvdV, FWW, and PR advised on the medical aspects relating to the methods and
results interpretation. TBH and AvS provided technical help with the model design. All
authors contributed to the redrafting of the manuscript and approved the final draft.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 192PDF page: 192PDF page: 192PDF page: 192
| Chapter 8184
REFERENCES
1. May MT, Gompels M, Delpech V, Porter K, Orkin C, Kegg S, et al. Impact on life expectancy of
HIV-1 positive individuals of CD4+ cell count and viral load response to antiretroviral therapy.
AIDS. 2014 May 15;28(8):1193–202.
2. Smit M, Brinkman K, Geerlings S, Smit C, Thyagarajan K, Sighem A van, et al. Future challenges
for clinical care of an ageing population infected with HIV: a modelling study. Lancet Infect Dis.
2015 Jul;15(7):810–8.
3. Freiberg MS, Chang C-CH, Kuller LH, Skanderson M, Lowy E, Kraemer KL, et al. HIV Infection
and the Risk of Acute Myocardial Infarction. JAMA Internal Medicine. 2013 Apr 22;173(8):614.
4. Schouten J, Wit FW, Stolte IG, Kootstra NA, van der Valk M, Geerlings SE, et al. Cross-sectional
comparison of the prevalence of age-associated comorbidities and their risk factors between
HIV-infected and uninfected individuals: the AGEhIV cohort study. Clin Infect Dis. 2014 Dec
15;59(12):1787–97.
5. Losina E, Hyle EP, Borre ED, Linas BP, Sax PE, Weinstein MC, et al. Projecting 10-year, 20-year,
and Lifetime Risks of Cardiovascular Disease in Persons Living With Human Immunodeficiency
Virus in the United States. Clin Infect Dis. 2017 15;65(8):1266–71.
6. Zanni MV, Schouten J, Grinspoon SK, Reiss P. Risk of coronary heart disease in patients with HIV
infection. Nat Rev Cardiol. 2014 Dec;11(12):728–41.
7. Grund B, Baker JV, Deeks SG, Wolfson J, Wentworth D, Cozzi-Lepri A, et al. Relevance of
Interleukin-6 and D-Dimer for Serious Non-AIDS Morbidity and Death among HIV-Positive
Adults on Suppressive Antiretroviral Therapy. PLoS One. 2016 May 12;11(5):e0155100.
8. Late presenters working group in COHERE in EuroCoord, Mocroft A, Lundgren J, Antinori A,
Monforte A d’Arminio, Brännström J, et al. Late presentation for HIV care across Europe: update
from the Collaboration of Observational HIV Epidemiological Research Europe (COHERE)
study, 2010 to 2013. Euro Surveill. 2015;20(47).
9. Mocroft A, Lundgren JD, Sabin ML, Monforte A d’Arminio, Brockmeyer N, Casabona J, et al.
Risk factors and outcomes for late presentation for HIV-positive persons in Europe: results from the
Collaboration of Observational HIV Epidemiological Research Europe Study (COHERE). PLoS
Med. 2013;10(9):e1001510.
10. European AIDS Clinical Society (EACS) treatment guidelines, version 8.1, October 2016. Available
at http://www.eacsociety.org/files/guidelines_8.1-english.pdf. Accessed 14 December 2016.
11. Friis-Møller N, Ryom L, Smith C, Weber R, Reiss P, Dabis F, et al. An updated prediction model
of the global risk of cardiovascular disease in HIV-positive persons: The Data-collection on Adverse
Effects of Anti-HIV Drugs (D:A:D) study. Eur J Prev Cardiol. 2016 Jan;23(2):214–23.
12. Mancia G, Fagard R, Narkiewicz K, Redon J, Zanchetti A, Böhm M, et al. 2013 ESH/ESC Practice
Guidelines for the Management of Arterial Hypertension. Blood Press. 2014 Feb;23(1):3–16.
13. Gras L, Van Sighem A, Fraser C, Griffin J, Miedema F, Lange J, et al. Predictors for Changes in
CD4 Cell Count 7 Years after Starting HAART. Conference on Retroviruses and Opportunistic
Infections, Abstract Number 530, 5-8 February 2006, Denver, Colorado. Available at https://
www.hiv-monitoring.nl/nl/onderzoek-datagebruik/ons-onderzoek/recente-presentaties/
presentaties-2006. Accessed 14 December 2016.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 193PDF page: 193PDF page: 193PDF page: 193
Cardiovascular Prevention Policy in HIV | 185
8
14. van Sighem AI, Gras LAJ, Smit C, Stolte IG, Reiss P. Monitoring Report 2015. Human
Immunodeficiency Virus (HIV) Infection in the Netherlands. Amsterdam: Stichting HIV
Monitoring, 2015. Available at www.hiv-monitoring.nl. Accessed at 14 December 2016.
15. Verschuren WMM, Blokstra A, Picavet HSJ, Smit HA. Cohort profile: the Doetinchem Cohort
Study. Int J Epidemiol. 2008 Dec;37(6):1236–41.
16. Althoff KN, Palella F, Gebo KA, Gange SJ, Rabkin CS, Thorne JE et al. Impact of smoking,
hypertension, and cholesterol on myocardial infarction in HIV-infected adults. Conference on
Retroviruses and Opportunistic Infections, Abstract Number 130, 13-16 February 2017, Seattle,
Washington. Available at http://www.croiconference.org/sessions/impact-smoking-hypertension-
cholesterol-myocardial-infarction-hiv-adults. Accessed 20 April 2017.
17. Nederlandse Vereniging van Hiv Behandelaren (NVHB) treatment guidelines, July 2016. Available
at http://www.nvhb.nl/richtlijnhiv/index.php/Richtlijn_HIV. Accessed 14 December 2016.
18. Shahmanesh M, Schultze A, Burns F, Kirk O, Lundgren J, Mussini C, et al. The cardiovascular
risk management for people living with HIV in Europe: how well are we doing? AIDS. 2016 Oct
23;30(16):2505–18.
19. Myerson M, Poltavskiy E, Armstrong EJ, Kim S, Sharp V, Bang H. Prevalence, treatment, and
control of dyslipidemia and hypertension in 4278 HIV outpatients. J Acquir Immune Defic Syndr.
2014 Aug 1;66(4):370–7.
20. Saumoy M, Alonso-Villaverde C, Navarro A, Olmo M, Vila R, Maria Ramon J, et al. Randomized
trial of a multidisciplinary lifestyle intervention in HIV-infected patients with moderate-high
cardiovascular risk. Atherosclerosis. 2016 Mar;246:301–8.
21. Shuter J, Morales DA, Considine-Dunn SE, An LC, Stanton CA. Feasibility and preliminary
efficacy of a web-based smoking cessation intervention for HIV-infected smokers: a randomized
controlled trial. J Acquir Immune Defic Syndr. 2014 Sep 1;67(1):59–66.
22. Cahill K, Stevens S, Perera R, Lancaster T. Pharmacological interventions for smoking cessation: an
overview and network meta-analysis. Cochrane Database Syst Rev. 2013 May 31;(5):CD009329.
23. Moadel AB, Bernstein SL, Mermelstein RJ, Arnsten JH, Dolce EH, Shuter J. A randomized
controlled trial of a tailored group smoking cessation intervention for HIV-infected smokers. J
Acquir Immune Defic Syndr. 2012 Oct 1;61(2):208–15.
24. Ridker PM, Danielson E, Fonseca FAH, Genest J, Gotto AM, Kastelein JJP, et al. Rosuvastatin to
prevent vascular events in men and women with elevated C-reactive protein. N Engl J Med. 2008
Nov 20;359(21):2195–207.
25. Funderburg NT, Jiang Y, Debanne SM, Labbato D, Juchnowski S, Ferrari B, et al. Rosuvastatin
reduces vascular inflammation and T-cell and monocyte activation in HIV-infected subjects on
antiretroviral therapy. J Acquir Immune Defic Syndr. 2015 Apr 1;68(4):396–404.
26. Lo J, Lu MT, Ihenachor EJ, Wei J, Looby SE, Fitch KV, et al. Effects of statin therapy on
coronary artery plaque volume and high-risk plaque morphology in HIV-infected patients with
subclinical atherosclerosis: a randomised, double-blind, placebo-controlled trial. Lancet HIV. 2015
Feb;2(2):e52-63.
27. Gilbert JM, Fitch KV, Grinspoon SK. HIV-Related Cardiovascular Disease, Statins, and the
REPRIEVE Trial. Top Antivir Med. 2015 Nov;23(4):146–9.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 194PDF page: 194PDF page: 194PDF page: 194
| Chapter 8186
28. Serrano-Villar S, Estrada V, Gómez-Garre D, Ávila M, Fuentes-Ferrer M, San RJ, et al. Diagnosis
of subclinical atherosclerosis in HIV-infected patients: higher accuracy of the D:A:D risk equation
over Framingham and SCORE algorithms. Eur J Prev Cardiol. 2014 Jun;21(6):739–48.
29. Thompson-Paul AM, Lichtenstein KA, Armon C, Palella FJ, Skarbinski J, Chmiel JS, et al.
Cardiovascular disease risk prediction in the HIV Outpatient Study. Clin Infect Dis. 2016 Sep
9;63(11):1508–16.
30. Monroe AK, Haberlen SA, Post WS, Palella FJ, Kinsgley LA, Witt MD, et al. Cardiovascular
disease risk scores’ relationship to subclinical cardiovascular disease among HIV-infected and HIV-
uninfected men. AIDS. 2016 Aug 24;30(13):2075–84.
31. Feinstein MJ, Nance RM, Drozd DR, Ning H, Delaney JA, Heckbert SR, et al. Assessing and
Refining Myocardial Infarction Risk Estimation Among Patients With Human Immunodeficiency
Virus: A Study by the Centers for AIDS Research Network of Integrated Clinical Systems. JAMA
Cardiol. 2017 Feb 1;2(2):155–62.
32. Ryom L, Lundgren JD, El-Sadr WM, Reiss P, Philips A, Kirk O, et al. Association between
cardiovascular disease and contemporarily used protease inhibitors. Conference on Retroviruses
and Opportunistic Infections, Abstract Number 128LB, 13-16 February 2017, Seattle, Washington.
Available at http://www.croiconference.org/sessions/association-between-cardiovascular-disease-
contemporarily-used-protease-inhibitors. Accessed 20 April 2017.
33. Doll R, Peto R, Boreham J, Sutherland I. Mortality from cancer in relation to smoking: 50 years
observations on British doctors. Br J Cancer. 2005 Feb 14;92(3):426–9.
34. Clifford GM, Polesel J, Rickenbach M, Dal Maso L, Keiser O, Kofler A, et al. Cancer risk in the Swiss
HIV Cohort Study: associations with immunodeficiency, smoking, and highly active antiretroviral
therapy. J Natl Cancer Inst. 2005 Mar 16;97(6):425–32.
35. Franklin W, Lowell FC, Michelson AL, Schiller IW. Chronic obstructive pulmonary emphysema; a
disease of smokers. Ann Intern Med. 1956 Aug;45(2):268–74.
36. Lord JW. Cigarette smoking and peripheral atherosclerotic occlusive disease. JAMA. 1965 Jan
18;191:249–51.
37. Hontelez JAC, de Vlas SJ, Baltussen R, Newell M-L, Bakker R, Tanser F, et al. The impact of
antiretroviral treatment on the age composition of the HIV epidemic in sub-Saharan Africa. AIDS.
2012 Jul 31;26 Suppl 1:S19-30.
This is a pre-copy-editing, author-produced PDF of an article accepted for publication in Clinical
Infectious Diseases following peer review. The definitive publisher-authenticated version "Smit M*, van
Zoest RA*, Nichols BE, Vaartjes I, Smit C, van der Valk M, van Sighem A, Wit FW, Hallett TB, Reiss
P; Netherlands AIDS Therapy Evaluation in The Netherlands (ATHENA) Observational HIV Cohort.
Cardiovascular Disease Prevention Policy in Human Immunodeficiency Virus: Recommendations From
a Modeling Study. Clin Infect Dis 2018 Feb 10;66(5):743–750. doi: 10.1093/cid/cix858." is available
online at: https://dx.doi.org/10.1093/cid/cix858.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 195PDF page: 195PDF page: 195PDF page: 195
Cardiovascular Prevention Policy in HIV | 187
8
SUPPLEMENT 8.1: MODEL TECHNICAL DETAILS
This supplementary material provides technical details of the model design including
parameter values and results of model validation and sensitivity analyses. This model
is an expansion of a previous model looking at the ageing people living with HIV
(PLHIV) in The Netherlands (1), with this supplement explaining the major additions
and modification to the model structure. Additional details on the basic model
structure can be found in the supplement by Smit and colleagues (1). Briefly, the model
is an individual-based model of cardiovascular disease (CVD) in ageing PLHIV on
combination antiretroviral therapy (cART) in The Netherlands. Supplementary Figure
8.1 shows the basic model design. The model follows PLHIV on cART from 1st January
2010 until 31st December 2030 or death (Supplementary Figure 8.1A), and simulates
how patients age, experience clinical events (hypertension, dyslipidemia, diabetes and
CVD), and start cardiovascular medication.
CVD includes myocardial infarction, stroke, coronary artery angioplasty, coronary by-
pass, carotid endarterectomy, and deaths from other coronary heart disease and these are
simulated by incorporating the D:A:D risk equation, a CVD risk equation specifically
developed in PLHIV, into the model (Supplementary Figure 8.1B) (2). Every month
new patients are added to the model, joining the modelled patient cohort on cART.
The model ends on 31st December 2030 with all patients who were on cART on 1st
January 2010 and are still alive (Supplementary Figure 8.1A see patients in red), as well
as a cohort of new patients who started cART between 2010 and 2030 (Supplementary
Figure 8.1A see patients in blue). The model takes into account a large number of
interactions between demographic (Supplementary Figure 8.1C red), HIV-related
(Supplementary Figure 8.1C green), treatment-related (Supplementary Figure 8.1C
yellow) and clinical factors (Supplementary Figure 8.1C blue). For example, the model
accounts for the impact of age, sex and clinical factors on the risk of death, or how
cholesterol levels can vary with the use of lipid lowering drugs, as a patient ages or
switches cART regimen (see below for more details).
The period from 2010 to 2015 is used to carry out out-of-sample model validation
(see section below “Model checks”), and interventions targeting CVD are simulated
between 2017 and 2030. Interventions tested are HIV-specific interventions including
i) an outreach intervention targeting “late” presenters (i.e. initiating cART at CD4
counts ≥500 cells/mm3), and ii) the use of cART regimens with no known increased
CVD risk, and ‘traditional’ interventions including i) a smoking cessation program and
ii) the intensified monitoring and treatment of hypertension and dyslipidemia.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 196PDF page: 196PDF page: 196PDF page: 196
| Chapter 8188
SUPPLEMENTARY FIGURE 8.1 MODEL DESIGN. A. Illustrates the basic model structure. The model starts on 1st January 2010 with all PLHIV on cART in care
and registered with the ATHENA cohort aged ≥18 years old, infected with HIV-1 only, cART naïve prior to
entering the cohort, who had initiated cART after 1st January 1996 (depicted in red). Each month, the model
simulated calendar events (e.g. ageing), new events in a probabilistic manner (e.g. deaths and new CVD events),
and new patients are added as they start cART (depicted in blue). When the model stops on 31st December
2030, the modelled patient cohort contains all patients from 2010 who are still alive as well as new patients who
have entered care between 1st January 2010 and 31st December 2030. The period from 2010 to 2015 is used
to carry out out-of-sample model validation, and interventions targeting CVD are simulated between 2017
and 2030. B. Shows the risk factors included in the D:A:D risk equation (2). C. Interactions simulated in the
model between demographic (red), HIV-related (green), cardiovascular treatment (yellow) and clinical factors
(blue), for example how the risk of death is determined by age, sex and clinical factors. The dashed lines show
interactions between individual factors and solid lines between a category and a specific risk factor.
2010 2030
+ 1 month
Model check Interventions
2010-2015 2017-2030
ATHENA
cART
Icons made by Freepik from www.flaticon.com
D:A:D risk equation (Friis-Møller et al 2016)
DemographicAge
GenderClinicalDiabetes
Family historySmoking (current/former)
Total cholesterolHDL cholesterol
Systolic blood pressureHIV-related
Current CD4 countAbacavir use
Cumulative PI exposureCumulative NRTI exposure
DEMOGRAPHICS
CLINICAL FACTORS
HIV-RELATED FACTORS
CARDIOVASCULAR MEDICATION
A
B C
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 197PDF page: 197PDF page: 197PDF page: 197
Cardiovascular Prevention Policy in HIV | 189
8
ATHENA data, definitions and baseline population
The ATHENA cohort is a national observational cohort which collects biological,
immunological and clinical data from all PLHIV in care in The Netherlands at entry
and at each follow-up visit. The cohort was established in 1998, and contains data
retrospectively from 1996. Its design has been described previously (3). ATHENA
data was used to populate the model on 1st January 2010 and to calculate parameters,
including HIV incidence, increases in mean age, changes in cART regimens, and age-
and-sex specific diabetes incidence. All patients from the ATHENA cohort aged ≥18
years old, infected with HIV-1 only, cART naïve prior to entering the cohort, and who
had initiated cART after 1st January 1996 were included.
Diabetes mellitus was defined as use of antidiabetic medication, a “diabetes mellitus”
diagnosis in clinical records, elevated fasting plasma glucose (≥7.0 mmol/L (126 mg/dL)
at ≥2 consecutive moments in time), or HbA1c ≥48 mmol/mol (6.5%). Hypertension
was defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90
mmHg, and/or use of antihypertensive medication (4). Dyslipidaemia was defined as
total cholesterol >6.5 mmol/L (250 mg/dL), total/HDL cholesterol ratio >5, and/or use
of lipid lowering medication. CVDs were based on clinical patient records, and the vast
majority of CVDs were validated according to procedures developed by D:A:D (92% of
strokes, 93% of myocardial infarctions, and 84% of coronary/cardiovascular procedures).
Data up to and including 1st January 2010 were used for model parameterisation,
comprising 10,005 individuals, and patients from 1st January 2010 and up to 31st
December 2014 (validation population) were used for out-of-sample validation.
8,791 patients were still in care on 1st January 2010 and were used to populate the
model at the start by directly importing patient records, in order to capture the real-
life complexity and overlap of CVD risk factors. A number of people in 2010 had
missing data (Supplementary Table 8.1) regarding for example family history or
missing measurements for current CD4 counts, blood pressure or cholesterol. Prior to
importing the patient records directly into the model we used imputation to handle
these missing data. The imputation approach is described in detail below. The imputed
baseline population was imported directly into the model with all their characteristics
(e.g. sex, age, smoking status, blood pressure and cholesterol levels) via anonymised
patient records.
Imputation of missing data
Missing data in the baseline and validation population was handled by using multiple
imputation, assuming values were missing at random. R software (version 3.2.2; The R
Foundation for Statistical Computing) and the MICE package were used for multiple
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 198PDF page: 198PDF page: 198PDF page: 198
| Chapter 8190
imputation, using a chained equations approach. Continuous variables, binary variables
and unordered categorical variables were imputed using predictive mean matching,
logistic regression, and polytomous logistic regression, respectively. All variables listed in
Supplementary Table 8.1 were included in the imputation procedure as either predictors
or imputed variables. Twenty imputed datasets were generated during the imputation,
using twenty iterations during the imputation procedure. Quality and consistency over
the twenty imputed datasets was assessed by comparing distributions and proportions
of imputed variables over the twenty imputed datasets. Graphs illustrating the imputed
mean and standard deviation with increasing iterations were inspected by eye for
stable distributions. Comparison of the observed and imputed values are described in
Supplementary Table 8.1.
New patients
In order to simulate the addition of new patients into the modelled patient cohort
between 2010 and 2030, we needed to:
a) make projections of the number of people starting cART each year, and
b) ‘recycle’ existing patients from the cohort on 1st January 2010 to use as ‘new’ patients
while accounting for future changes in cART regimens, mean age and CD4 counts.
To construct reasonable projections of future number of PLHIV starting cART, we
applied the same method as used in our previous model of the ageing PLHIV in The
Netherlands (see supplement of Smit et al 2015 (1)). In summary, a compartmental
model of HIV incidence, disease progression and cART initiation was constructed
to explore the different trajectories HIV incidence could take in the future. The
compartmental model simulated the HIV cascade, including HIV incidence, disease
progression and cART initiation, and assumed a minimum, medium and maximum
HIV incidence scenario. The manuscript presents the results for the medium scenario
with additional results listed at the bottom of this supplement (see section “Sensitivity
analyses”). The model projects the number of people starting cART each year for each
incidence scenario (Supplementary Table 8.2), and the model assumes that the number
of patients starting cART each month is constant over the year.
In order to model the entry of new patients we ‘recycled’ patients from the cohort of
patients in care on 1st January 2010. At the start of every year the model randomly
selects a sample of patients from the 2010 baseline cohort to match the number of
people projected to start cART that year (from Supplementary Table 8.2). The model
accounts for the fact that these new patients are expected to i) increase in mean age
up to 2030, ii) have increased in CD4 count at cART initiation up to 2015, and iii)
start different first-line cART regimens from 2015 onwards as reflected by updated
treatment guidelines.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 199PDF page: 199PDF page: 199PDF page: 199
Cardiovascular Prevention Policy in HIV | 191
8
SUPPLEMENTARY TABLE 8.1. LIST OF PATIENT VARIABLES INCLUDED IN THE MODEL, THE NUMBER OF MISSING VALUES FOR EACH VARIABLE OF INTEREST, THE OBSERVED AND IMPUTED VALUES FOR EACH VARIABLE OF INTERESTS, AND REASONS FOR MISSING VALUES. (SOURCE: ATHENA COHORT DATA AND IMPUTED DATASET).
Variables
Number (%) of
missing values
Observed values per
imputed variable
Imputed values per
imputed variable
Reasons for missing
values
Demographic characteristics
Age 0 (0%) n/a n/a n/a
Sex 0 (0%) n/a n/a n/a
CVD 0 (0%) n/a n/a n/a
Pregnancy 0 (0%) n/a n/a n/a
Cardiovascular disease risk factors
Diabetes 0 (0%) n/a n/a n/a
Smoking status 3,182 (36.2%)ATHENA cohort data
collection is incomplete
regarding smoking
status and family
history of CVD
Never 1,302 (23.2%) 2,187 (24.9%)
Ever 3,217 (57.4%) 4,897 (55.7%)
Current 1,090 (19.4%) 1,707 (19.4%)
Family history of CVD 6,186 (70.4%) 255 (9.8%) 835 (9.5%)
Recent1 systolic blood pressure 3,087 (35.1%) 125 (115-136) 123 (115-135) Blood pressure/
cholesterol levels
were included only if
assessed between 1st
January 2009 and 1st
January 2010
Recent1 diastolic blood pressure 3,089 (35.1%) 80 (70-85) 80 (70-85)
Recent1 total cholesterol 2,400 (27.3%) 4.9 (4.2-5.6) 4.9 (4.2-5.6)
Recent1 HDL cholesterol 4,801 (54.6%) 1.2 (0.9-1.5) 1.2 (1.0-1.5)
HIV-related characteristics
Recent1 CD4 count 697 (7.9%) 500 (360-660) 500 (362-660) CD4 count levels
were included only if
assessed between 1st
January 2009 and 1st
January 2010
CD4 count at entry ATHENA
cohort
33 (0.4%) 280 (110-460) 280 (110-460)
Cumulative use of protease
inhibitors
1 (0.01%) 0.3 (0-2.4) 0.3 (0-2.4)
Cumulative use of nucleoside reverse
transcriptase inhibitors
1 (0.01%) 4.1 (1.6-7.5) 4.1 (1.6-7.5)
Current use of abacavir 1 (0.01%) 1,172 (13.3%) 1,172 (13.3%)
Current use of protease inhibitor 0 (0%) n/a n/a n/a
Current use of nucleoside reverse
transcriptase inhibitors
0 (0%) n/a n/a n/a
Current use of non-nucleoside
reverse transcriptase inhibitors
0 (0%) n/a n/a n/a
Current use of integrase inhibitors 0 (0%) n/a n/a n/a
Lipid lowering medication 38 (0.4%) 688 (7.9%) 697 (7.9%) The start/stop date
of medication was
missing for a number
of patients.
Antihypertensive medication 66 (0.8%) 981 (11.2%) 995 (11.3%)
Antidiabetic medication 13 (0.1%) 242 (2.8%) 251 (2.9%)
Data are presented as number (%) or median (interquartile range).1 Defined as <12 months prior to 1 January 2010.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 200PDF page: 200PDF page: 200PDF page: 200
| Chapter 8192
The parameters describing increases in mean age over time can be found in the
supplement of Smit et al 2015 (1). In summary, the mean age at cART initiation
has been increasing linearly over time since 1996, and the model assumes that this
increase will continue until 2030. When the model randomly selects a sample of ‘new’
patients from the baseline cohort, the selection of patients are matched to the estimated
predicted mean age between 2010 and 2030 (i.e. the mean age of the selected sample
of patients from baseline cohort matched the estimated predicted mean age for that
specific year). In this manner, the sample of ‘new’ patients age over time, with their
profiles (e.g. blood pressure and cholesterol measurements) matching this demographic
transition. In addition, upon entry into the model patients’ CD4 counts and cART
regimens are adjusted to match expected changes. The annual CD4 count increase
between 2010 and 2015 amongst patients starting cART was calculated based on recent
changes in CD4 counts at cART initiation in the total patient population reported by
SUPPLEMENTARY TABLE 8.2. PROJECTED NUMBER OF PLHIV STARTING CART AS PREDICTED BY THE DETERMINISTIC MODEL OF HIV-INFECTION USING THREE INCIDENCE SCENARIOS FOR THE EPIDEMIC; MINIMUM, MEDIUM AND MAXIMUM. (SOURCE: SMIT ET AL 2015 SUPPLEMENT (1)).
Number of new treatment initiations
Year Minimum scenario Medium scenario Maximum scenario
2010 1009 1009 1009
2011 897 897 897
2012 805 805 805
2013 730 730 734
2014 667 667 681
2015 612 612 642
2016 563 563 614
2017 518 518 594
2018 476 476 580
2019 436 441 570
2020 397 414 564
2021 359 394 559
2022 322 379 556
2023 285 369 554
2024 252 361 552
2025 225 356 551
2026 204 352 550
2027 189 349 550
2028 177 348 550
2029 169 346 549
2030 163 345 549
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 201PDF page: 201PDF page: 201PDF page: 201
Cardiovascular Prevention Policy in HIV | 193
8
the ATHENA cohort between 2005 and 2013 (see Supplementary Figure 8.2) (5).
We assume the linear increase observed in this period will continue up to 2015, after
which CD4 counts at cART initiation are assumed to remain at 2015 levels. From
2015 onwards, the model puts new PLHIV on cART regimens reflecting changes in
European AIDS Society (EACS) guidelines, which increasingly recommended the use
of integrase inhibitors over other combinations (6). Parameters describing changes in
first-line cART regimens were taken from the ATHENA cohort (Supplementary Table
8.3).
SUPPLEMENTARY FIGURE 8.2. CHANGES OVER TIME IN MEDIAN CD4 COUNTS AT THE START OF CART IN THE TOTAL POPULATION, MEN WHO HAVE SEX WITH MEN (MSM), AND HETEROSEXUAL MEN AND WOMEN BETWEEN 2005 AND 2013. (SOURCE: SHM REPORT 2014 (5)).
SUPPLEMENTARY TABLE 8.3. PROPORTIONS AND PARAMETERS DESCRIBING CHANGES IN FIRST-LINE CART REGIMEN IN 2015 COMPARED TO 2010-2014 AS PER ATHENA COHORT DATA.
PIs NNRTIs IIs Other
A. Data
2010-2014 (average) 32.0% 61.6% 4.3% 2.1%
2015 19.5% 33.1% 46.3% 1.1%
B. Parameters
Proportion starting IIs instead 39.1% 46.3%
A. Proportion of people starting Protease inhibitors (PIs), Non-Nucleoside Reverse Transcriptase Inhibitors
(NNRTIs), or Integrase inhibitors (IIs) in 2010 to 2014 (on average between those four years) compared to
2015 as per ATHENA data. B. Parameters describing proportion of ‘new’ patients initiating IIs instead of PIs
and NNRTIs from 2015 at cART initiation to reflect first-line cART guideline changes. (Source: ATHENA
cohort data).
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 202PDF page: 202PDF page: 202PDF page: 202
| Chapter 8194
Mean age, mortality, diabetes, blood pressure and cholesterol
Demographic projections of sex-specific increases in mean age at start of cART, sex-
and age-specific diabetes incidence, and age-, sex- and cause-specific mortality rates
have been described previously (see supplement of Smit et al (1)). The modified model
further accounts for the impact of age on total cholesterol, HDL cholesterol, and
systolic and diastolic blood pressure. Literature suggests that the risk of hypertension
and dyslipidaemia changes with age and that these changes may differ by sex (7,8).
To account for this in our model we integrated age-and-sex-specific annual changes
in blood pressure and lipid levels based on data from the Doetinchem Cohort Study.
The Doetinchem Cohort Study is an ongoing long-lasting general population-based
health study in the Netherlands city of Doetinchem among men and women aged
20-59 years, which was initiated in 1987 (9). The study collects information on
demographics, lifestyle, biological factors, and comorbidities, including CVD. Data up
to and including January 2013 were used, encompassing 6,390 individuals. We included
patients not on antihypertensive or lipid lowering medication to estimate annual
changes in blood pressure and cholesterol, respectively. For every 10-year age group we
included all measurements of individuals within this age group, and any measurements
that occurred 5 years prior to the lower age limit or after the upper age limit. Individuals
with ≥2 measurements available were used to obtain the annual slope per individual,
from which we calculated the mean slope per age group, stratified by sex. When no
statistically significant difference in slope was found between males and females, the
combined estimate was used. The parameters describing these age-related changes are
listed in Supplementary Table 8.4 below.
SUPPLEMENTARY TABLE 8.4. ANNUAL CHANGES IN BLOOD PRESSURE AND LIPID LEVELS PER 10-YEAR AGE GROUP AND SEX AMONG INDIVIDUALS NOT USING BLOOD PRESSURE OR LIPID LOWERING THERAPY. (SOURCE: DOETINCHEM COHORT STUDY (9)).
Age Group
(years)
Systolic blood
pressure (mmHg)
Diastolic blood
pressure (mmHg)
Total cholesterol
(mmol/L)
HDL cholesterol
(mmol/L)
Men Women Men Women Men Women Men Women
20 to 30 -0.03031 -0.03031 0.3971 0.3971 0.0474 -0.00293 0.0103 0.0173
30 to 40 0.151 0.584 0.4791 0.4791 0.0283 0.0158 0.00811 0.0173
40 to 50 0.682 1.036 0.444 0.530 0.0168 0.0503 0.00813 0.0151
50 to 60 1.084 1.119 0.3381 0.3381 0.00254 0.0522 0.009491 0.009491
60 to 70 1.289 1.140 0.2101 0.2101 -0.00784 0.0177 0.00645 0.000787
Over 70 1.0131 1.0131 0.1541 0.1541 -0.01861 -0.01861 -0.001761 -0.001761
1 Males and females were combined to calculate the annual change in blood pressure/lipid levels when no
significant sex-difference in slope was observed.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 203PDF page: 203PDF page: 203PDF page: 203
Cardiovascular Prevention Policy in HIV | 195
8
SUPPLEMENTARY TABLE 8.5. VARIABLES INCLUDED IN FULL AND REDUCED D:A:D RISK EQUATION USED TO PREDICT THE 5-YEAR CVD RISK. (SOURCE: FRIIS -MØLLER ET AL 2016 (2))
Variables Full D:A:D risk
equation1
Reduced D:A:D risk
equation2
Variable type
Demographic characteristics
Age X X continuous
Gender X X binary
Clinical characteristics
Diabetes X X binary
Family history of CVD X X binary
Current smoker X X binary
Past smoker X X binary
Total cholesterol X X continuous
HDL cholesterol X X continuous
Systolic blood pressure X X continuous
HIV-related characteristics
Current CD4 count X X continuous
Abacavir use (current) X binary
Cumulative PI exposure X continuous
Cumulative NRTI exposure X continuous
1 The full D:A:D risk equation was used for individuals with cumulative NRTI exposure up to 9 years, and
PI exposure up to 5.5 years. 2The reduced D:A:D risk equation was used for individuals who were exposed to
NRTIs for more than 9 years and/or to PIs for more than 5.5 years.
D:A:D risk equation
As suggested by Friis-Møller et al (2) the full risk equation was used for individuals
with cumulative nucleoside reverse transcriptase inhibitors (NRTIs) exposure up to
9 years and protease inhibitors (PIs) exposure up to 5.5 years, and the reduced risk
equation for those exposed to NRTIs or PIs for a longer time period (Supplementary
Table 8.5). The 5-year CVD risk was calculated for every patient in the model, and
converted to a monthly risk assuming constant risk within the 5-year timeframe. The
occurrence of CVD was based on the calculated CVD risk, and assigned to the patient
in a probabilistic manner. Once a patient was predicted to have CVD, the specific CVD
event was determined using the parameters described in Supplementary Table 8.6 (e.g.
a patient who developed CVD had 41% chance to have had a non-fatal myocardial
infarction).
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 204PDF page: 204PDF page: 204PDF page: 204
| Chapter 8196
SUPPLEMENTARY TABLE 8.6. DISTRIBUTION OF VARIOUS TYPES OF CVD WITHIN D:A:D COHORT STUDY. (SOURCE: FRIIS -MØLLER ET AL 2016 (2)).
Type of CVD Number of events (%)
Myocardial infarction (non-fatal) 413 (41%)
Myocardial infarction (fatal) 80 (8%)
Stroke (non-fatal) 255 (25%)
Stroke (fatal) 40 (4%)
Coronary artery angioplasty 129 (13%)
Coronary bypass 36 (4%)
Carotid endarterectomy 13 (1%)
Death from other coronary heart disease 44 (4%)
TOTAL 1,010
Co-medication
The model simulates treatment with lipid lowering medication and antihypertensive
medication. The model assumes that prescription behaviour remains stable at 2010
levels, with data suggesting that the probability of being prescribed antihypertensive
and lipid lowering medication depending on the type of CVD and diabetes status of the
patient. Parameters describing prescription behaviour are listed in Supplementary Table
8.7 and Supplementary Table 8.8 and treatment is being modelled probabilistically.
SUPPLEMENTARY TABLE 8.7. PARAMETERS DEFINING USE OF LIPID LOWERING MEDICATION BASED ON ATHENA DATA.
Proportion starting lipid
lowering drugs
CVD only 0.75
High cholesterol only 0.13
High cholesterol and diabetes (no CVD) 0.29
High cholesterol and CVD (irrespective of the presence of diabetes) 0.76
Very high cholesterol only 0.17
Very high cholesterol and diabetes (no CVD) 0.33
Very high cholesterol and CVD (irrespective of the presence of diabetes) 0.75
CVD includes myocardial infarction, stroke, coronary artery angioplasty, coronary bypass and carotid
endarterectomy; high cholesterol is defined as a total/HDL cholesterol ratio above 5 and very high cholesterol
as a total/HDL cholesterol ratio above 8. (Source: ATHENA cohort data).
Once patients are started on antihypertensive or lipid lowering medication, they are
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 205PDF page: 205PDF page: 205PDF page: 205
Cardiovascular Prevention Policy in HIV | 197
8
assumed to remain on treatment for life and their blood pressure and cholesterol is
adjusted. The new blood pressure and cholesterol levels are set to the levels observed
amongst patients in care in 2010 in ATHENA who have been on medication for longer
than 6 months on 1st January 2010, with some patients returning to normal target levels,
and some patients remaining at elevated level, reflecting levels observed in the cohort
(Supplementary Table 8.9). We assumed that blood pressure and cholesterol levels
remained stable once patients started on antihypertensive or lipid lowering medication.
We tested normal distribution, Weibull and gamma distributions for best fit describing
the distribution of systolic, diastolic blood pressure, HDL and total cholesterol on co-
medication amongst patients on co-medication in ATHENA, with parameters listed in
Supplementary Table 8.9.
SUPPLEMENTARY TABLE 8.9. PARAMETERS DESCRIBING TOTAL CHOLESTEROL, HDL, SYSTOLIC BLOOD PRESSURE AND DIASTOLIC BLOOD PRESSURE DISTRIBUTION AMONGST PLHIV ON LIPID LOWERING OR ANTIHYPERTENSIVE MEDICATION. (SOURCE: ATHENA COHORT DATA).
Parameters Function type
Total cholesterol(mmol/L)1 Mean = 5.17; SD = 1.31 Normal distribution
HDL (mmol/L)1 Alpha= 8.74; Beta = 0.137 Gamma distribution
Systolic (mmHg)2 Mean = 135.77; SD = 19.11 Normal distribution
Diastolic (mmHg) 2 Mean = 83.39; SD = 12.03 Normal distribution
SD=standard deviation1 Lipid levels distribution based on individuals using lipid lowering medication for ≥6 months on 1st January
2010.2 Blood pressure distribution based on individuals using antihypertensive drugs for ≥6 months on 1st January
SUPPLEMENTARY TABLE 8.8. PARAMETERS DEFINING USE OF ANTIHYPERTENSIVE MEDICATION BASED ON ATHENA DATA.
Proportion starting
antihypertensives
CVD only 0.58
High blood pressure only 0.25
High blood pressure and diabetes (no CVD) 0.56
High blood pressure and CVD (irrespective of the presence of diabetes) 0.80
CVD includes myocardial infarction, stroke, coronary artery angioplasty, coronary bypass and carotid
endarterectomy; high blood pressure as systolic blood pressure ≥140 mm Hg and/or diastolic blood pressure ≥90
mm Hg. (Source: ATHENA cohort data).
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 206PDF page: 206PDF page: 206PDF page: 206
| Chapter 8198
2010.
CD4 increases on cART
The model simulates increases in CD4 counts with time on cART, accounting for factors
such as age and time since cART initiation. Previous work by Gras et al showed that the
increase is steepest soon after cART initiation and levelling off as duration on cART
is prolonged, with recovery differing in patients under and over 50 years old (10). We
used data from the ATHENA cohort to calculate increases in CD4 count by age and
time on cART. The ATHENA data showed that CD4 recovery rate differed between
patients aged over and under 50 years old (except in those starting cART at CD4 count
above 500 cells/mm3) and by CD4 count at cART initiation. Recovery rates also differ
according to duration on cART, with steepest recovery observed in the first 6 months,
followed by 6 months to two years, levelling off after two years. It was assumed that
no further changes would occur after 10 years. Parameters describing monthly CD4
recovery are listed in Supplementary Table 8.10.
Impact of switching cART on cholesterol
We reviewed the literature to assign realistic estimates in cholesterol alterations as a
result of switching antiretroviral drugs. We included randomized clinical trials that
described the effect of switching i) from a protease inhibitor (PI) to non-nucleoside
reverse-transcriptase inhibitor (NNRTI), ii) from PI to integrase inhibitor (II), and
iii) from NNRTI to II. We included studies conducted in patients aged 18 years and
over on cART, in high-income settings, and with undetectable viral load. Exclusion
criteria were mono- or dual-therapy prior to cART initiation. A list of included studies
SUPPLEMENTARY TABLE 8.10. MONTHLY CD4 COUNT IN CELLS/MM3 RECOVERY BY AGE AND TIME ON CART. (SOURCE: ATHENA COHORT DATA)
CD4 count at cART initiation (cells/mm3)
Time on cART <200 200-350 350-500 ≥500
0 to 6 months
Under 50 years old 14.80 16.50 18.50 19.80
Over 50 years old 12.70 15.80 16.00 19.80
6 to 24 months
Under 50 years old 7.30 5.67 6.11 5.94
Over 50 years old 5.39 5.33 5.44 5.94
2 to 10 years
Under 50 years old (-3.40*t +51.88)/12 (-3.11*t +43.32)/12 (0.74*t +18.2)/12 (0.12*t +14.02)/12
Over 50 years old (-4.71*t +48.14)/12 (-2.69*t +41.24)/12 (-7.33*t +66.16)/12 (0.12*t +14.02)/12
t=time since start of cART (in years)
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 207PDF page: 207PDF page: 207PDF page: 207
Cardiovascular Prevention Policy in HIV | 199
8
is shown in Supplementary Table 8.11.The model included the impact of switching
from a PI to an NNRTI, a PI to an II, or from an NNRTI to an II on total cholesterol,
assumed to be reduced by 0.6 mmol/L, 0.8 mmol/L, or 0.2 mmol/L, respectively, with
no effect on HDL cholesterol.
List of incorporated interactions
The model takes into account a number of interaction between demographic, clinical,
HIV-related and co-medication factors (see Supplementary Figure 8.1C), including:
• Mortality: the impact of gender, age and co-morbidity on the risk of mortality (see
supplement of Smit et al 2015 for more details (1))
• CD4 count: the impact of cART and age on CD4 (see section “CD4 increases on
cART”). The model takes into account CD4 increases after cART initiation from
2010 to 2015 (see section “New patients”)
• Blood pressure: the impact of age (see section “Mean age, mortality, diabetes,
blood pressure and cholesterol”) and the use of antihypertensives (see section “Co-
medication”) on blood pressure levels
• Cholesterol: the impact of age (see section “Mean age, mortality, diabetes, blood
pressure and cholesterol”), the use of lipid lowering drugs (see section “Co-
medication”) and of cART regimen changes (see section “Impact of switching cART
on cholesterol”) on cholesterol levels
• Diabetes: the impact of age on diabetes (see supplement Smit et al 2015 (1))
• Co-medication: impact of developing a new clinical event (diabetes, hypertension or
dyslipidaemia and serious CVD event) on the probability of starting co-medication
(see supplement of Smit et al 2015 (1) for anti-diabetics and section “Co-medication”
for antihypertensive and lipid lowering drugs)
• cART: the impact of changes in EACS guidelines on first-line regimens (see section
“New patients”)
Economic evaluation
An economic evaluation of the mathematical modelling output was carried out.
Supplementary Table 8.12 shows costs (€) in The Netherlands in 2016. These were used
to calculate total costs for the baseline scenario and per intervention.
Model checks
A number of checks were carried out to ensure that the model adequately captures clinical
care in The Netherlands and could be used to reliably predict the future age-structure,
CD4 count trends, and burden of non-communicable disease in The Netherlands. These
checks include doing out-of-sample checks comparing 2010 to 2015 model output
with 2010 to 2015 observational ATHENA cohort data. The results of these model
validation checks show that the model consistently generates output of the right order
of magnitude, leading to the conclusion that the model provides projections of the right
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 208PDF page: 208PDF page: 208PDF page: 208
| Chapter 8200
SUPP
LEM
ENTA
RY T
ABL
E 8.
11. L
ITER
ATU
RE R
EVIE
W S
WIT
CH
ING
STU
DIE
S A
ND
EFF
ECT
ON
TO
TAL
AN
D H
DL
CH
OLE
STER
OL.
Stu
dy
Stu
dy
Per
iod
Po
pu
lati
on
Fo
llo
w-u
p
tim
e (w
eek
s)
Inte
rven
tio
nO
utc
om
e
NM
edia
n a
ge
(yea
rs)
Mal
e (%
)L
ipid
low
erin
g
dru
gs
(%)
Bas
elin
e A
RV
To
tal c
ho
lest
ero
l (m
mo
l/L
)H
DL
ch
ole
ster
ol (
mm
ol/
L)
Bas
elin
eF
oll
ow
-up
Mea
n c
han
ge
Bas
elin
eF
oll
ow
-
up
Mea
n c
han
ge
PI
to N
NR
TI
Ech
ever
ría,
2014
(11)
2009-1
143 (
I=22;
C=21)
I=47.4
C=46.2
I=63.6
%
C=76.2
%
I=18.2
C=9.5
AT
V/r
(I=
50.5
%; C
=57.1
%)
LP
V/r
(I=
31.8
%; C
=19.0
%)
fAP
V/r
(I=
13.6
%; C
=14.3
%)
SQ
V/r
(I=
4.5
%; C
=9.5
%)
48w
Sw
itch
PI
to E
TR
5.3
6 (
1.1
1)
4.9
3 (
1.0
3)1
n/a
1.3
7 (
0.4
7)
1.3
7
(0.5
3)
n/a
Pal
ella
, 2014 (
12)
2010-1
2476 (
I=317;
C=159)
I=42
C=43
I=86.1
%
C=90.6
%
n/a
AT
Z/r
(I=
38.5
%; C
=34.0
%)
LP
V/r
(I=
30.6
%; C
=36.5
%)
DR
V/r
(I=
19.9
%; C
=20.8
%)
fAP
V/r
(I=
7.9
%; C
=7.5
%)
SQ
V/r
(I=
1.9
%; C
=1.3
%)
AP
V/r
(I=
0.3
%; C
=0%
)
48w
Sw
itch
PI-
bas
ed
regi
men
to s
ingl
e-
table
t re
gim
en
(RP
V/
FT
C/T
DF
)
n/a
n/a
-0.6
21
n/a
n/a
- 0.0
5
PI
to I
I
Ero
n, 2
010 (
13)
2007-0
8702 (
I=350;
C=352)
I=43
C=42
I=81%
C=76%
0%
LP
V/r
24w
Sw
itch
LP
V/r
to
RA
L
5.5
7 (
1.5
6)
n/a
-12.6
%1
1.2
3 (
0.3
8)
n/a
-0.7
%
Sau
moy
, 2012
(14)
2008
81 (
I=40;
C=41)
I=44
C=48
I=85%
C=73.2
%
I=17.5
%
C=22.8
%
LP
V/r
(I=
42.5
%; C
=36.6
%)
AT
Z/r
(I=
37.5
%; C
=39.0
%)
fAP
V/r
(I=
10.0
%; C
=12.2
%)
SQ
V/r
(I=
2.5
%; C
=12.2
%)
48w
Sw
itch
PI/
r to
RA
L4.9
7 (
3.3
8-6
.63)
n/a
-0.8
1 (
=-
15.0
2%
)1
1.1
2 (
0.6
5-
1.9
4)
n/a
-0.1
1 (
-9.3
7%
)
Arr
ibas
, 2014 (
15)
2011-1
2433 (
I=293;
C=140)
I=41
C=40
I=85%
C=86%
n/a
AT
Z/r
(I=
42%
; C=37%
)
DR
V/r
(I=
39%
; C=43%
)
LP
V/r
(I=
17%
; C=16%
)
fAP
V/r
(I=
2%
; C=4%
)
SQ
V/r
(I=
1%
; C=0%
)
48w
Sw
itch
PI/
r, F
TC
,
TD
F t
o E
VG
,
cobic
ista
t, F
TC
,
TD
F
5.2
6 (
0.8
4)
5.1
0 (
0.9
5)
LP
V/r
:
-0.6
2 (
SD
0.7
0)1
,
oth
er g
roup
s N
S
1.3
2 (
0.3
2)
1.2
7
(0.3
7)
LP
V/r
: -0.0
5 (
SD
0.2
3)
NN
RT
I to
II
Pozn
iak, 2
014
(16)
2011-1
2433 (
I=290;
C=143)
I=43
C=39
I=92%
C=94%
n/a
EF
V (
I=80%
; C=74%
)
NV
P (
I=16%
; C=19%
)
RP
V (
I=3%
; C=7%
)
ET
R (
I=1%
; C=0%
)
48w
Sw
itch
NN
RT
I to
EV
G/c
obic
ista
t
4.9
5 (
0.9
2)
n/a
EF
V: -
0.1
8
(0.8
1)1
1.4
0 (
0.3
7)
n/a
-0.0
8 (
0.2
4)1
Gup
ta, 2
013 (
17)
2011-1
230 (
I=15;
C=15)
I=39
C=38
I=93%
C=87%
0%
EF
V24w
Sw
itch
EF
V t
o R
AL
3.9
8 (
0.9
5)
n/a
0.3
3 (
0.5
1)1
1.0
2 (
0.3
0)
n/a
0.0
08 (
0.1
6)
Ngu
yen
, 2011
(18)
2009-1
053 (
I=29;
C=24)
I=48
C=47
I=79.3
1%
C=66.6
7%
I=17.2
%
C=20.8
%
EF
V4w
Sw
itch
EF
V t
o R
AL
5.2
(5.0
-5.9
)4.7
-0.4
1.3
(1.1
0-
1.5
5)
1.2
-0.1
Abbre
viat
ion
s: A
PV
/r,
amp
ren
avir
/rit
on
avir
; A
RV
, an
tire
trovi
ral
ther
apy;
AT
V/r
, at
azan
avir
/rit
on
avir
; C
, co
ntr
ol
arm
; D
RV
/r,
dar
un
avir
/rit
on
avir
; E
TR
, et
ravi
rin
e; E
VG
, el
vite
grav
ir;
fAP
V/r
, fo
sam
pre
nav
ir/r
iton
avir
; F
TC
, em
tric
itab
ine;
I, i
nte
rven
tion
arm
; II,
in
tegr
ase
inh
ibit
or;
LP
V/r
, lop
inav
ir/r
iton
avir
; N, n
um
ber
; n/a
, not
avai
lable
; NN
RT
I, n
on
-nucl
eosi
de
reve
rse-
tran
scri
pta
se i
nh
ibit
ors
; NS
, not
stat
isti
call
y si
gnifi
can
t (P
≥0.0
5);
NV
P, n
evir
apin
e;
PI,
pro
teas
e in
hib
itor;
RA
L, r
alte
grav
ir; R
PV
, ril
piv
irin
e; S
D, s
tan
dar
d d
evia
tion
; SQ
V/r
, saq
uin
avir
/rit
on
avir
; TD
F, t
enofo
vir.
1 R
educt
ion
in
ch
ole
ster
ol
leve
l st
atis
tica
lly
sign
ifica
nt
(P<0.0
5)
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 209PDF page: 209PDF page: 209PDF page: 209
Cardiovascular Prevention Policy in HIV | 201
8
SUPPLEMENTARY TABLE 8.12. COSTS.
Lifetime disease cost components Cost (2016, €) Source
Myocardial Infarction1 €4,037 Open data van de Nederlandse Zorgautoriteit (19,20)
Stroke1 €2,498 Open data van de Nederlandse Zorgautoriteit (19,21)
Coronary artery angioplasty2 €17,062 Open data van de Nederlandse Zorgautoriteit (19,22)
Coronary by-pass2 €14,521 Open data van de Nederlandse Zorgautoriteit (19,22)
Carotid endarterectomy2 €7,670 Open data van de Nederlandse Zorgautoriteit (19)
Yearly HIV-related costs3
Antiretrovirals
ABC+3TC+PI €11,939 Zorginstituut Nederland Medicijn Kosten (23)
ABC+3TC+II €9,817 Zorginstituut Nederland Medicijn Kosten (23)
ABC+3TC+NNRTI €7,232 Zorginstituut Nederland Medicijn Kosten (23)
TDF/ZDV+FTC+PI €11,058 Zorginstituut Nederland Medicijn Kosten (23)
TDF/ZDV+FTC+II €11,861 Zorginstituut Nederland Medicijn Kosten (23)
TDF/ZDV+FTC+NNRTI €9,790 Zorginstituut Nederland Medicijn Kosten (23)
Patient monitoring4 €523 Nichols et al 2015 (24)
Abbreviations: 3TC, lamivudine, ABC, abacavir; FTC, emtricitabine; II, integrase inhibitor, NNRTI, non-
nucleoside reverse-transcriptase inhibitors; PI, protease inhibitor; TDF, tenofovir; ZDV, zidovudine. 1 Cost weighted by deaths by MI/Stroke2Including long-term recovery3Weighted by the percentage of patients on each potential regimen combination4Twice yearly visits, viral load monitoring, basic laboratory monitoring
direction and magnitude.
Population in care
Results of the model output for number of patients in care at the 31st December of
each validation year (2010-2014) were compared to out-of-sample observational
ATHENA cohort data, to check if the deterministic model of HIV incidence was
reliable in predicting the number of new patients starting cART and the number of
PLHIV on cART in care. The comparison between the number of new patients starting
cART annually in the model and the ATHENA data are presented in Supplementary
Figure 8.3A and the total number of PLHIV on cART annually in Supplementary
Figure 8.3B. The results show that the model is able to accurately recreate the number
of patients on cART in care annually up to 2015, despite a slight underestimation in the
number of patients starting in cART.
In addition, we compared sex-specific mean age of PLHIV in care in 2010 to 2014
between the model and ATHENA data, to check whether projected age-trends were
adequately capturing trends ongoing in the observation cohort (Supplementary Figure
8.4). The validation results show that the model is performing adequately in projecting
short-term trends in mean age amongst PLHIV in care in The Netherlands.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 210PDF page: 210PDF page: 210PDF page: 210
| Chapter 8202
Mortality
The full details of the validation checks for mortality can be found in the supplement
by Smit et al 2015 (1). Mortality was validated in three ways; i) by comparing the
standardized mortality ratio between the PLHIV and the general population, ii) by
comparing age-specific death rates amongst PLHIV and the general Dutch population,
and iii) by comparing annual percentage of deaths amongst PLHIV on cART between
the model and the observational ATHENA cohort data (Supplementary Figure 8.5).
Model results show that mortality patterns were accurately reconstructed.
Smoking status
In order to test our assumption that smoking status at entry will not change substantially
over time, we compared smoking status amongst new patients between observational
ATHENA cohort data and model output in 2010 to 2014, stratified by sex. The results
are presented in Supplementary Figure 8.6 and show that smoking status in both new
female and male patients starting cART are comparable between the model and the
SUPPLEMENTARY FIGURE 8.3. COMPARISON OF POPULATION IN CARE STRATIFIED BY SEX BETWEEN THE MODEL OUTPUT AND OBSERVATIONAL ATHENA COHORT DATA. A. Number of PLHIV newly starting cART and B. Total number of PLHIV on cART.
SUPPLEMENTARY FIGURE 8.4. MEAN AGE ACCORDING TO OBSERVATIONAL DATA FROM ATHENA COHORT AND MODEL OUTPUT FOR A. MEN AND B. WOMEN.
4041424344454647484950
2010 2011 2012 2013 2014 2015
Age
Year
Mean Age - Men
95% CIs Mean age - Data Mean age - Model
4041424344454647484950
2010 2011 2012 2013 2014 2015
Age
Year
Mean Age - Women
95% CIs Mean age - Data Mean age - Model
A B
0
500
1.000
1.500
Num
ber o
f peo
ple
Year
Number of New Entries
Men - Data Men - Model Women - Data Women - Model
A
0
5.000
10.000
15.000
Num
ber o
f peo
ple
Year
Total Popula on on ART
Men - Data Men - Model Women - Data Women - Model
B
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 211PDF page: 211PDF page: 211PDF page: 211
Cardiovascular Prevention Policy in HIV | 203
8
SUPPLEMENTARY FIGURE 8.5. ANNUAL PERCENTAGE OF DEATHS AMONGST PLHIV ON CART ACCORDING TO THE OBSERVATIONAL ATHENA COHORT DATA AND THE MODEL, STRATIFIED BY SEX.
012345
2010 2011 2012 2013 2014
Perc
enta
ge
Year
Percentage Deaths
Men -Data Men - Model Women - Data Women - Model
SUPPLEMENTARY FIGURE 8.6 . SMOKING STATUS AMONG PLHIV WHO ENTERED CARE AFTER 1 JANUARY 2010 ACCORDING TO OBSERVED DATA FROM THE ATHENA COHORT AND MODEL OUTPUT FOR A. MEN AND B. WOMEN.
0
10
20
30
40
50
60
70
80
90
100
Perc
enta
ge
Year
Smoking Status Amongst New Entries - Men
Never Smoker - Data Never Smoker - ModelEver Smoker - Data Ever Smoker - ModelCurrent Smoker - Data Current Smoker - Model
0
10
20
30
40
50
60
70
80
90
100
Perc
enta
ge
Smoking Status Amongst New Entries - Women
Never Smoker - Data Never Smoker - ModelEver Smoker - Data Ever Smoker - ModelCurrent Smoker - Data Current Smoker - Model
Year
A B
data.
Diabetes, blood pressure and cholesterol
The model simulates the development of diabetes based on age and sex. Changes in
blood pressure and cholesterol levels are simulated based on age and sex, use of CVD
co-medication, as well the impact of switches in cART regimens on cholesterol. In order
to check that the model is able to recreate new diagnosis of diabetes and sex-and-age-
specific blood pressure and cholesterol profiles as seen in the observational ATHENA
cohort data we compared model output with data in 2010 to 2014. The results for
diabetes (Supplementary Figure 8.7) show that the model is able to recreate trends in
new diagnosis of diabetes between 2010 and 2014. Age-and sex-specific cholesterol
(Supplementary Figure 8.8) and blood pressure (Supplementary Figure 8.9) profiles
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 212PDF page: 212PDF page: 212PDF page: 212
| Chapter 8204
0
1
2
3
4
5
2010 2011 2012 2013 2014
Perc
enta
ge
Year
New Diabetes Events - Percentage
Data Model
SUPPLEMENTARY FIGURE 8.8. CHOLESTEROL PROFILES ACCORDING TO OBSERVATIONAL ATHENA COHORT DATA AND MODEL OUTPUT FOR A. MEN UNDER 50 YEARS OLD, B. MEN OVER 50 YEARS OLD, C. WOMEN UNDER 50 YEARS OLD AND D. WOMEN OVER 50 YEARS OLD.
0102030405060708090
100
Perc
enta
ge
Year
Cholesterol - Men Under 50 Year Old
TC - Data TC - ModelTC/HDL 5-8 - Data TC/HDL 5-8 - ModelTC - Data TC - Model
0102030405060708090
100
Perc
enta
ge
Year
Cholesterol - Men Over 50 Year Old
TC - Data TC - ModelTC/HDL 5-8 - Data TC/HDL 5-8 - ModelTC - Data TC - Model
0102030405060708090
100
Perc
enta
ge
Year
Cholesterol - Women Under 50 Year Old
TC - Data TC - ModelTC/HDL 5-8 - Data TC/HDL 5-8 - ModelTC - Data TC - Model
0102030405060708090
100
Perc
enta
ge
Year
Cholesterol - Women Over 50 Year Old
TC - Data TC - ModelTC/HDL 5-8 - Data TC/HDL 5-8 - ModelTC - Data TC - Model
A B
C D
SUPPLEMENTARY FIGURE 8.7. PERCENTAGE OF NEWLY DIAGNOSED DIABETES CASES ACCORDING TO OBSERVATIONAL ATHENA COHORT DATA AND THE MODEL.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 213PDF page: 213PDF page: 213PDF page: 213
Cardiovascular Prevention Policy in HIV | 205
8
also show robust model predictions.
CVD incidence
The model simulated CVD incidence (including myocardial infarction, stroke, coronary
artery angioplasty, coronary by-pass, carotid endarterectomy, and deaths from other
coronary heart disease) by using the D:A:D risk equation. In order to check the
robustness of incorporating this equation into the model the percentage of incident
CVDs among individuals without a prior history of CVD was compared between
out-of-sample ATHENA data and the model output in 2010 to 2014. CVDs in the
ATHENA data base were based on clinical patient records and the vast majority of
CVDs were validated according to procedures developed by the D:A:D Study. The
results are presented in Supplementary Figure 8.10 and show that the model predicted
0
50
100
Perc
enta
ge
Year
Blood Pressure - Men Under 50 Year Old
Normal - Data Normal - ModelPrehypertension - Data Prehypertension - ModelStage 1 - Data Stage 1 - ModelStage 2 - Data Stage 2 - Model
0
50
100Pe
rcen
tage
Year
Blood Pressure - Men Over 50 Year Old
Normal - Data Normal - ModelPrehypertension - Data Prehypertension - ModelStage 1 - Data Stage 1 - ModelStage 2 - Data Stage 2 - Model
A B
A B
0
50
100
Perc
enta
ge
Year
Blood Pressure - Women Over 50 Year Old
Normal - Data Normal - ModelPrehypertension - Data Prehypertension - ModelStage 1 - Data Stage 1 - ModelStage 2 - Data Stage 2 - Model
0
50
100
Perc
enta
ge
Year
Blood Pressure - Women Under 50 Year Old
Normal - Data Normal - ModelPrehypertension - Data Prehypertension - ModelStage 1 - Data Stage 1 - ModelStage 2 - Data Stage 2 - Model
SUPPLEMENTARY FIGURE 8.9. BLOOD PRESSURE PROFILES ACCORDING TO OBSERVATIONAL ATHENA COHORT DATA AND MODEL OUTPUT FOR A. MEN UNDER 50 YEARS OLD, B. MEN OVER 50 YEARS OLD, C. WOMEN UNDER 50 YEARS OLD AND D. WOMEN OVER 50 YEARS OLD. Normal blood pressure: <120/80 mmHg; Prehypertension: 120/80 to 139/89 mmHg; Stage 1 hyperten-
sion: 140/90 to 159/99 mmHg; Stage 2 hypertension: 160/100 mmHg and above.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 214PDF page: 214PDF page: 214PDF page: 214
| Chapter 8206
the incidence of CVD robustly.
Co-medication
We further compared the percentage of people using co-medication (including
antidiabetics, antihypertensives and lipid lowering drugs) between the model and
observational ATHENA cohort data from 2010 and 2014. The results are presented in
Supplementary Figure 8.11A, B and C for anti-diabetics, antihypertensives and lipid
lowering drugs, respectively and show robustness of model predictions.
CD4 count
Finally, we compared the mean CD4 counts between 2010 and 2014 amongst new
entries and in the total population according to observational ATHENA cohort data
and model output. The results amongst new entries and the total population are shown
in Supplementary Figure 8.12A and B, respectively, and demonstrate that the model
was able to recreate CD4 count trends observed in the ATHENA data in the short-
term.
Additional results and sensitivity analyses
Sensitivity analyses
The below section shows additional results for sensitivity analyses we carried out to test
model output robustness and specificity. We tested alternative HIV incidence scenarios
(minimum and maximum), as well as the impact of fading-out the impact of cumulative
time on PIs once patients switch off PIs as well as the long-term impact of smoking
cessation. In addition, we evaluated the effect of reaching 10% of smokers successfully,
0,00
0,50
1,00
1,50
2,00
2010 2011 2012 2013 2014
Perc
enta
ge
Year
Serious CVD Events
Data Model
SUPPLEMENTARY FIGURE 8.10. COMPARISON OF PERCENTAGE OF NEW CVD ACCORDING TO OBSERVATIONAL ATHENA COHORT AND MODEL OUTPUT BETWEEN 2010 AND 2014.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 215PDF page: 215PDF page: 215PDF page: 215
Cardiovascular Prevention Policy in HIV | 207
8
SUPPLEMENTARY FIGURE 8.11. COMPARISON OF USE OF CO -MEDICATION ACCORDING TO OBSERVATIONAL ATHENA COHORT DATA AND MODEL OUTPUT BETWEEN 2010 AND 2014 FOR A. ANTI-DIABETICS, B. ANTIHYPERTENSIVE MEDICATION, AND C. LIPID LOWERING DRUGS.
A B
0
5
10
15
20
2010 2011 2012 2013 2014
Perc
enta
ge
Year
An -diabe cs
Data Model
0
5
10
15
20
2010 2011 2012 2013 2014
Perc
enta
ge
Year
An -hypertensives
Data Model
C
0
5
10
15
20
2010 2011 2012 2013 2014
Perc
enta
ge
Year
Lipid lowering drugs
Data Model
SUPPLEMENTARY FIGURE 8.12. COMPARISON OF MEAN CD4 COUNTS ACCORDING TO OBSERVATIONAL ATHENA COHORT DATA AND MODEL FROM 2010 TO 2014 AMONGST A. NEW ENTRIES AND B. TOTAL PATIENT POPULATION.
A B
200250300350400450500550600650700750
2010 2011 2012 2013 2014 2015
CD4
coun
t
Year
Mean CD4 Count - New Entries
95% CIs Mean CD4 count - Data Mean CD4 count - Model
200250300350400450500550600650700750
2010 2011 2012 2013 2014 2015
CD4
coun
t
Year
Mean CD4 Count - Total Popula on
95% CIs Mean CD4 count - Data Mean CD4 count - Model
since low rates of smoking cessation have been reported despite tailored interventions
(25–28).
1. Minimum and maximum HIV incidence scenario
In this scenario we assume that HIV incidence decreases more rapidly or less rapidly over
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 216PDF page: 216PDF page: 216PDF page: 216
| Chapter 8208
time (see section “Entry of new patients”). The results are presented in Supplementary
Figure 8.13 and show that the overall trends of CVD burden and risk do not change
according to the HIV incidence scenario. The results further show that if HIV incidence
continues to remain high, CVD burden will be greater, probably driven by the larger
patient population. However, CVD risk in the patient population will be lower if HIV
incidence remains high, probably due to the addition of new patients of younger age
keeping the risk at the population level lower.
SUPPLEMENTARY FIGURE 8.13. THE PROPORTION OF PATIENTS IN LOW (<5%, GREEN), MODERATE (5-10%, YELLOW), AND HIGH (≥10%, RED) PREDICTED 5-YEAR CVD RISK GROUPS BETWEEN 2010 AND 2030 AS SIMULATED BY THE MODEL USING THE D:A:D RISK EQUATION ASSUMING A. MINIMUM HIV INCIDENCE AND B. MAXIMUM HIV INCIDENCE.
2. Decrease in cumulative time on PI effect
We tested a scenario where the effect of cumulative time on PI ‘fades-out’ over time
once patients switch off PIs, by assuming that five years after patients were switched off
PIs the effect of cumulative time on PIs is zero. In the D:A:D risk equation, any time on
PIs contributes toward CVD risk with no reduction in its effect even after long periods
when patients no longer use PIs. We compared the CVD burden at baseline assuming
no fading effect and fading after five years, as any fading impact of cumulative time
on PIs after switching would impact baseline and all interventions tested. The results
showed that the cumulative CVD burden between 2010 and 2030 would be 0.5% lower
if the effect of cumulative PIs was to fade out completely five years after switching off
PIs.
3. Decrease in ever smoking status effect
We tested a scenario where patients who had stopped smoking five or more years ago no
longer have a CVD risk attributed to being ‘ex-smoker’ as per the risk equation. In the
A B10%
5-10%
<5%
10%
5-10%
<5%
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 217PDF page: 217PDF page: 217PDF page: 217
Cardiovascular Prevention Policy in HIV | 209
8
D:A:D risk equation the same risk is assumed for ex-smokers, irrespective of duration
since smoking cessation. Literature suggests that the effect of being a smoker fades over
time after cessation (29). We compared the CVD burden at baseline assuming no fading
effect and fading after five years, as any fading impact of smoking after cessation would
impact baseline and all interventions tested. The results showed that the cumulative
CVD burden between 2010 and 2030 would be 7.4% lower if the effect of being an ex-
smoker was to fade out completely five years after smoking cessation.
4. Smoking cessation intervention - reaching 10% of patients successfully
We tested a more feasible smoking cessation scenario of 10% (25–28) (as opposed to
the scenarios of 50-100% successful implementation described in the manuscript).
Smoking cessation could prevent an average of 1.3% of CVD cases annually in all and
HR patients if implemented 10% successfully.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 218PDF page: 218PDF page: 218PDF page: 218
| Chapter 8210
REFERENCES
1. Smit M, Brinkman K, Geerlings S, Smit C, Thyagarajan K, Sighem A van, et al. Future challenges
for clinical care of an ageing population infected with HIV: a modelling study. Lancet Infect Dis.
2015 Jul;15(7):810–8.
2. Friis-Møller N, Ryom L, Smith C, Weber R, Reiss P, Dabis F, et al. An updated prediction model
of the global risk of cardiovascular disease in HIV-positive persons: The Data-collection on Adverse
Effects of Anti-HIV Drugs (D:A:D) study. Eur J Prev Cardiol. 2016 Jan;23(2):214–23.
3. van Sighem AI, van de Wiel MA, Ghani AC, Jambroes M, Reiss P, Gyssens IC, et al. Mortality and
progression to AIDS after starting highly active antiretroviral therapy. AIDS. 2003 Oct;17(15):2227–
36.
4. Weber MA, Schiffrin EL, White WB, Mann S, Lindholm LH, Kenerson JG, et al. Clinical
Practice Guidelines for the Management of Hypertension in the Community. J Clin Hypertens.
2014;16(1):14–26.
5. van Sighem AI, Gras LAJ, Smit C, Stolte IG, Reiss P. Monitoring Report 2014. Human
Immunodeficiency Virus (HIV) Infection in the Netherlands. Amsterdam: Stichting HIV
Monitoring, 2014. Available at www.hiv-monitoring.nl. Accessed at 14 December 2016
6. European AIDS Clinical Society (EACS) treatment guidelines, version 7.1, November 2014.
Available at http://www.eacsociety.org/files/guidelines_english_71_141204.pdf. Accessed 14
December 2016.
7. Wilson PW, Anderson KM, Harris T, Kannel WB, Castelli WP. Determinants of change in total
cholesterol and HDL-C with age: the Framingham Study. J Gerontol. 1994 Nov;49(6):M252–7.
8. Landahl S, Bengtsson C, Sigurdsson JA, Svanborg A, Svärdsudd K. Age-related changes in blood
pressure. Hypertension. 1986 Nov;8(11):1044–9.
9. Verschuren WMM, Blokstra A, Picavet HSJ, Smit HA. Cohort profile: the Doetinchem Cohort
Study. Int J Epidemiol. 2008 Dec;37(6):1236–41.
10. Gras L, Van Sighem A, Fraser C, Griffin J, Miedema F, Lange J, et al. Predictors for Changes in
CD4 Cell Count 7 Years after Starting HAART. Conference on Retroviruses and Opportunistic
Infections, Abstract Number 530, 5-8 February 2006, Denver, Colorado. Available at https://
www.hiv-monitoring.nl/nl/onderzoek-datagebruik/ons-onderzoek/recente-presentaties/
presentaties-2006. Accessed 14 December 2016.
11. Echeverría P, Bonjoch A, Puig J, Moltó J, Paredes R, Sirera G, et al. Randomised study to assess the
efficacy and safety of once-daily etravirine-based regimen as a switching strategy in HIV-infected
patients receiving a protease inhibitor-containing regimen. Etraswitch study. PloS One. 2014
Feb;9(2): e84676.
12. Palella FJ, Fisher M, Tebas P, Gazzard B, Ruane P, Van Lunzen J, et al. Simplification to rilpivirine/
emtricitabine/tenofovir disoproxil fumarate from ritonavir-boosted protease inhibitor antiretroviral
therapy in a randomized trial of HIV-1 RNA-suppressed participants. AIDS. 2014 Jan;28(3):335–
44.
13. Eron JJ, Young B, Cooper DA, Youle M, Dejesus E, Andrade-Villanueva J, et al. Switch to a
raltegravir-based regimen versus continuation of a lopinavir-ritonavir-based regimen in stable HIV-
infected patients with suppressed viraemia (SWITCHMRK 1 and 2): two multicentre, double-
blind, randomised controlled trials. Lancet. 2010 Jan;375(9712):396–407.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 219PDF page: 219PDF page: 219PDF page: 219
Cardiovascular Prevention Policy in HIV | 211
8
14. Saumoy M, Sánchez-Quesada JL, Martínez E, Llibre JM, Ribera E, Knobel H, et al. LDL subclasses
and lipoprotein-phospholipase A2 activity in suppressed HIV-infected patients switching to
raltegravir: Spiral substudy. Atherosclerosis. 2012 Nov;225(1):200–7.
15. Arribas JR, Pialoux G, Gathe J, Di Perri G, Reynes J, Tebas P, et al. Simplification to coformulated
elvitegravir, cobicistat, emtricitabine, and tenofovir versus continuation of ritonavir-boosted
protease inhibitor with emtricitabine and tenofovir in adults with virologically suppressed HIV
(STRATEGY-PI): 48 week results of a randomised, open-label, phase 3b, non-inferiority trial.
Lancet Infect Dis. 2014 Jul;14(7):581–9.
16. Pozniak A, Markowitz M, Mills A, Stellbrink H-J, Antela A, Domingo P, et al. Switching to
coformulated elvitegravir, cobicistat, emtricitabine, and tenofovir versus continuation of non-
nucleoside reverse transcriptase inhibitor with emtricitabine and tenofovir in virologically suppressed
adults with HIV (STRATEGY-NNRTI): 48 week results of a randomised, open-label, phase 3b
non-inferiority trial. Lancet Infect Dis. 2014 Jul;14(7):590–9.
17. Gupta SK, Mi D, Moe SM, Dubé MP, Liu Z. Effects of switching from efavirenz to raltegravir on
endothelial function, bone mineral metabolism, inflammation, and renal function: a randomized,
controlled trial. J Acquir Immune Defic Syndr. 2013 Nov;64(3):279–83.
18. Nguyen A, Calmy A, Delhumeau C, Mercier I, Cavassini M, Mello AF, et al. A randomized cross-
over study to compare raltegravir and efavirenz (SWITCH-ER study). AIDS. 2011 Jul;25(12):1481–
7.
19. Open data van de Nederlandse Zorgautoriteit. Available at: http://www.opendisdata.nl/. Accessed 1
June 2016.
20. Soekhlal RR, Burgers LT, Redekop WK, Tan SS. Treatment costs of acute myocardial infarction in
the Netherlands. Neth Heart J. 2013 May;21(5): 230–5.
21. Buisman LR, Tan SS, Nederkoorn PJ, Koudstaal PJ, Redekop WK. Hospital costs of ischemic stroke
and TIA in the Netherlands. Neurology. 2015 Jun;84(22):2208–15.
22. Osnabrugge RL, Magnuson EA, Serruys PW, Campos CM, Wang K, van Klaveren D, et al. Cost-
effectiveness of percutaneous coronary intervention versus bypass surgery from a Dutch perspective.
Heart. 2015 Dec;101(24):1980–8.
23. Zorginstituut Nederland. Medicijn Kosten 2016. Available at: https://www.medicijnkosten.nl
Accessed 1 June 2016.
24. Nichols BE, Götz HM, van Gorp ECM, Verbon A, Rokx C, Boucher CAB, et al. Partner Notification
for Reduction of HIV-1 Transmission and Related Costs among Men Who Have Sex with Men: A
Mathematical Modeling Study. PloS One. 2015 Nov;10(11):e0142576.
25. Saumoy M, Alonso-Villaverde C, Navarro A, Olmo M, Vila R, Ramon JM, et al. Randomized
trial of a multidisciplinary lifestyle intervention in HIV-infected patients with moderate-high
cardiovascular risk. Atherosclerosis. 2016 Mar;246:301–8.
26. Shuter J, Morales DA, Considine-Dunn SE, An LC, Stanton CA. Feasibility and preliminary
efficacy of a web-based smoking cessation intervention for HIV-infected smokers: a randomized
controlled trial. J Acquir Immune Defic Syndr. 2014 Sep;67(1):59–66.
27. Cahill K, Stevens S, Perera R, Lancaster T. Pharmacological interventions for smoking cessation: an
overview and network meta-analysis. Cochrane Database Syst Rev. 2013 May;(5):CD009329.
533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-Zoest533703-L-bw-ZoestProcessed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019Processed on: 17-9-2019 PDF page: 220PDF page: 220PDF page: 220PDF page: 220
| Chapter 8212
28. Moadel AB, Bernstein SL, Mermelstein RJ, Arnsten JH, Dolce EH, Shuter J. A randomized
controlled trial of a tailored group smoking cessation intervention for HIV-infected smokers. J
Acquir Immune Defic Syndr. 2012 Oct;61(2):208–15.
29. Mons U, Müezzinler A, Gellert C, Schöttker B, Abnet CC, Bobak M, et al. Impact of smoking
and smoking cessation on cardiovascular events and mortality among older adults: meta-analysis of
individual participant data from prospective cohort studies of the CHANCES consortium. BMJ.
2015 Apr;350: h1551.