In this analysis of 5056 HIV-1 infected individuals initiating randomized antiretroviral treatment in clinical trials, abacavircontaining regimens did not appear associated with increased risk of myocardial infarction (MI). Classic cardiovascular disease risk factors were the strongest predictors of MI.
Abstract
Background. Observational and retrospective clinical trial cohorts have reported conflicting results for the association of abacavir use with risk of myocardial infarction (MI), possibly related to issues that may bias estimation of treatment effects, such as time-varying confounders, informative dropout, and cohort loss due to competing events.
Methods. We analyzed data from 5056 individuals initiating randomized antiretroviral treatment (ART) in AIDS Clinical Trials Group studies; 1704 started abacavir therapy. An intent-to-treat analysis adjusted for pretreatment covariates and weighting for informative censoring was used to estimate the hazard ratio (HR) of MIs after initiation of a regimen with or without abacavir.
Results. Through 6 years after ART initiation, 36 MI events were observed in 17,404 person-years of follow-up. No evidence of an increased hazard of MI in subjects using abacavir versus no abacavir was seen (over a 1-year period: P = .50; HR, 0.7 [95% confidence interval {CI}, 0.2-2.4]); over a 6-year period: P = .24; HR, 0.6 [95% CI, 0.3-1.4]); these results were robust over as-treated and sensitivity analyses. Although the risk of MI decreased over time, there was no evidence to suggest a time-dependent abacavir effect. Classic cardiovascular disease (CVD) risk factors were the strongest predictors of MI.
Conclusion. We find no evidence to suggest that initial ART containing abacavir increases MI risk over short-term and long-term periods in this population with relatively low MI risk. Traditional CVD risk factors should be the main focus in assessing CVD risk in individuals with human immunodeficiency virus infection.
Recommended first-line antiretroviral therapy (ART) for human immunodeficiency virus (HIV) infection includes the use of two nucleoside or nucleotide reverse-transcriptase inhibitors (NRTIs) combined with one or more agents from other antiretroviral drug classes (nonnucleoside reverse-transcriptase inhibitors [NNRTIs], protease inhibitors [PIs], or integrase inhibitors) [1, 2]. Lifelong treatment is required to maintain HIV suppression, improve immune function, and minimize morbidity and mortality associated with HIV disease progression. Therefore, understanding long-term risks associated with ART is paramount.
The observational Data Collection on Adverse events of Anti-HIV Drugs (D:A:D) cohort reported that current or recent use of abacavir was associated with increased risk of myocardial infarction (MI) [3]; data from other cohort studies and clinical trials of abacavir have provided conflicting results [4–11]. For example, although analyses of the Strategies for Management of Antiretroviral Therapy (SMART) study found more than a 4-fold increase in risk [4], a GlaxoSmithKline meta-analysis of randomized trials found no increased risk [5] and an association detected in the French Agence Nationale de Recherches sur le SIDA (ANRS) cohort was not apparent after cocaine and intravenous drug users were excluded from the analysis cohort [10]. These differences may be explained by confounding and population differences, as well as differences in analytical approach. Because abacavir was considered an attractive ART for subjects with cardiovascular disease (CVD) risk [6,12–14], confounding is particularly problematic [15–17]. Recent treatment guidelines now caution against the use of abacavir in subjects with high CVD risk [1, 2], which will further confound analyses addressing this issue.
Previous analyses have focused on the increased relative risk of MI associated with abacavir. When an event is rare—as is MI in this context—even very large relative risks may have limited population-level relevance, making estimation of absolute risk difference essential.
We ascertained population level CVD risk associated with initiating ART with abacavir-containing regimens using a cohort of ART-naive persons randomly assigned ART through prospective clinical trials. Our unique cohort provides a means to address confounding and channeling bias issues of traditional observational cohorts. Our analytical approach and estimation of absolute risk places the CVD risk associated with abacavir for the treatment of ART-naive populations into a broader clinical context.
METHODS
Study Population
The AIDS Clinical Trials Group (ACTG) Longitudinal Linked Randomized Trials (ALLRT) cohort is an observational cohort that enrolls HIV-infected persons who have been prospectively randomized to receive ART within ACTG clinical trials [18]. Patients are observed longitudinally from the start of randomized ART to determine long-term clinical, virological, and immunological outcomes. Clinical trials approved for co-enrollment (parent protocols) are those providing randomized ART and at least 24 weeks of follow-up. ALLRT and its parent protocols are approved by Institutional Review Boards at all sites; all participants provide written informed consent.
This analysis includes all ART-naive individuals (defined as those with <30 days ART exposure before study entry) initiating ART in ALLRT parent protocols from June 1998 through November 2007 (Figure 1) [19–26]; subjects who did not start treatment were excluded. Patients were followed-up on their parent protocol with visits at least every 12 weeks. Subsequent to parent study completion, follow-up continued at 16 week intervals in ALLRT. The ALLRT database is cleaned and frozen annually; June 2009 datasets were used for our analyses.
Figure 1.
Study population. For each parent study, shaded follow-up cells reflect the potential for participant follow-up in the parent study follow-up; unshaded cells reflect that participant follow-up is only available for participants enrolled in the AIDS Clinical Trials Group (ACTG) Longitudinally Linked Randomized Trials cohort. Missing cells reflect that, based on study enrollment, no subjects had the potential for having follow-up data available at the time that the data were frozen for analysis. Truncated at a maximum of 6 years for each individual, a total of 17,579 person-years of follow-up data were available for the cohort as a whole. aAll studies randomized patients equally to each of study arm. ddI, didanosine; d4T, stavudine; EFV, efavirenz; NFV, nelfinavir; 3TC/ZDV, lamivudine/zidovudine (Combivir); IDV, indinavir; LPV/r, Kaletra; ABC/3TC/ZDV, abacavir/3TC/ZDV (Trizivir); NRTI, nucleoside reverse-transcriptase inhibitor (provider choice of ZDV, d4T, or TDF made prior to randomization); ATV/r, atazanavir pharmacologically boosted with ritonavir; TDF/FTC, tenofovir disoproxil fumarate/emtricitabine. bBecause A5202 did not close to follow-up until November 2009, in the present analysis, some subjects are administratively censored prior to 96 weeks. cCategory totals do not sum to the total exclusions because subjects may have been excluded for multiple reasons. dSubjects with a record of starting and stopping randomized antiretroviral treatment (ART) on the same day are considered to have not started randomized ART.
Study Outcomes
The primary end point was MI. MIs were prospectively reported in each parent study and in ALLRT in the ACTG database and, during parent study follow-up, via the Division of AIDS serious adverse events (SAE) reporting system. Additionally, standardized reporting criteria that required confirmation via electrocardiogram or elevation in serum myocardial injury enzymes were added to the reporting system in December 2000. For the present analysis, all MI diagnoses identified from the ACTG and SAE databases underwent independent review by two study investigators (C.B. and J.S.) and were classified as definite, possible, or unclassifiable [27]. The investigators were provided with supporting documentation for each MI as well as with details of adverse events occurring within 6 months of the putative MI; as necessary, additional supporting documentation was reviewed. A final determination of MIs classified differently by the initial reviewers was made by two additional investigators (A.C. and S.K.). Analyses included definite and possible MIs; for subjects with multiple qualifying MIs, only the first event was included.
Statistical Considerations
The primary objective of the analysis was to estimate the short-term and long-term relative hazard (HR) of MI between initial ART regimens with and without abacavir. Analyses were performed as intent-to-treat, including all available follow-up regardless of changes in ART. Observed follow-up was divided into monthly intervals and censored at the earliest of 6 months after the last clinic visit or 30 June 2009. For subjects who were not part of the ALLRT cohort, follow-up ended at the end of their parent study (or earlier for subjects prematurely lost to follow-up); because A5202 did not complete follow-up until November 2009, all A5202 follow-up is through that parent study. For an assessment of short-term and long-term risk, data were truncated at 1 and 6 years, respectively. Event times were defined from start of ART until first MI or non-MI death (which was considered a competing event) [28].
Cause-specific HR of MI associated with abacavir use was estimated using Cox proportional hazards regression analysis adjusted for a priori selected pretreatment patient characteristics and CVD risk factors; for a more parsimonious model, a single covariate for CVD risk was created (Table 1). Interactions of abacavir use with age and pretreatment CVD risk factors were evaluated; in long-term analyses, an interaction with abacavir use and time evaluated the model's proportional hazards assumption. Sensitivity analyses included as-treated analysis that censored follow-up at discontinuation of randomized abacavir strategy; including only studies with randomized abacavir (A5014, A5095, and A5202); including only contemporary regimens (2 or 3 NRTIs with an NNRTI or a PI pharmacologically boosted with ritonavir [boosted-PI]); excluding regimens with agents that have also been associated with increased CVD risk (specifically any PI or didanosine); and including all MI events regardless of investigator review.
Table 1.
Definition of Model Covariates
| Characteristic | Description of covariate definitiona | Included as a time-updated covariate in the estimation of censoring weight |
| HIV disease status | ||
| CD4+ cell count | Categorized as <200; 200–349;350–499; >500/mm3. | Yes |
| HIV-1 RNA | Categorized as <2.6; 2.6–4; 4–5; and >5 log10 copies/mL. | Yes (latest 2 measurements) |
| CVD risk factors | ||
| MDRD GFR | Estimated creatinine clearance based on a 4-variable MDRD formula [29]; categorized as </>60 mL/min/1.73 m2. | Nob |
| Diabetes | Prior diagnosis of diabetes or fasting (nonfasting) glucose level >126 (200) mg/dL. | Yes |
| History of hypertension | Prior evidence of diagnosis of hypertension or increased BP or prior or current receipt of antihypertensive agents; when available for ALLRT participants, systolic BP >140 mmHg or diastolic BP >90 mmHg. | Yes |
| History of dyslipidemiac | Possible or confirmed as follows: possible: total (non-fasting) cholesterol>6.2 mmol/L (241.8 mg/dL) or non-fasting triglycerides >2.3 mmol/L (204.7 mg/dL); Confirmed: total (fasting) cholesterol or (fasting) triglycerides greater than thresholds given above, or current or prior receipt of a lipid lowering agent. Treated as a binary covariate for modeling: no dyslipidemia versus possible/confirmed. | Yes |
| History of CVD | Defined as serious or minor based on any prior diagnosis of the following:serious, acute coronary syndrome, acute myocardial infarction, angina pectoris,aortic aneurysm, arterial occlusive disease, arteriosclerosis coronary artery,cardiac failure, cardiac failure congestive, cardiac failure high output,cerebral infarction, cerebrovascular accident, coronary artery disease,coronary artery dissection, iliac artery occlusion,left ventricular failure, myocardial infarction, pericarditis,pulmonary embolism,silent myocardial infarction, transient ischaemic attack, or ventricular failure;minor, any prior diagnosis of any non-serious CVD condition. | Yes |
| Evaluable only in ALLRT participants | ||
| Ever smokedd | Based on participant self-report of current or a prior history of smoking on their first evaluated smoking record on ALLRT.Although ALLRT entry may occur after start of treatment,because smoking status as defined (ever smoked versus never smoked)in this cohort is expected to remain stable, evaluations afterthe start of treatment were treated as pretreatment covariates. | No |
| Family history of CVDd | Based on self-reported family history of premature CVD reported at any time during follow-up but treated at all times as a pretreatment characteristic. | No |
NOTE. ALLRT, AIDS Clinical Trials Group (ACTG) Longitudinally Linked Randomized Trials cohort; BP, blood pressure; CVD, cardiovascular disease; HDL, high-density lipoprotein; HIV, human immunodeficiency virus; MDRD GFR, modification of diet in renal disease glomerular filtration rate.
All modeling also adjusts for pretreatment age, sex, self-identified race/ethnicity, and body mass index.
Only abnormal (grade 3 or higher) serum creatinine was routinely captured in the clinical database.
Because HDL cholesterol was not consistently collected across studies, it was not included in dyslipidemia definition.
Because pretreatment smoking status and family history of CVD were evaluated only in the ALLRT cohort, these variables could not be included in analyses for weight estimation because of non-identifiability.
Cox proportional hazards analysis assumes that censoring is not informative of the outcome of interest. Causes of censoring in our cohort that have the potential to violate this assumption include premature parent study discontinuation and decisions not to enroll into ALLRT. Stabilized inverse probability of censoring weights (IPCWs) were incorporated in the long-term analyses to account for this potential bias [30]. The weights were estimated using pooled longitudinal logistic regression to represent the inverse of the probability of remaining in observation for each month given pretreatment and past time-updated covariates. For each month, these were defined according to observed data up to the end of the previous month and included the year of follow-up, status on randomized ART, markers of HIV-1 disease status, and CVD risk factors (Table 1). If an evaluation was not available for a given month, the covariate value defined for the prior month was carried forward. Covariate values defined based on any given clinical visit were generally carried forward over 2–5 successive monthly periods. To maximize the cohort size, categories for missing covariate information were created. Censoring weights were also used in as-treated analyses to account for censoring at discontinuation of abacavir strategy; they were not used to adjust for end of follow-up for non-MI deaths, because this would create an unrealistic scenario in which deaths from other causes do not occur.
To place these relative risk estimates of abacavir use into the context of absolute risk, the estimated parameters of a Cox proportional hazards model (adjusted for all statistically significant risk factors from our main analyses, P < .05) were transformed to provide estimated cumulative incidence of MI over time for a hypothetical set of individuals [31]; 95% confidence intervals [CIs] were obtained from 1000 bootstrapped samples.
All analyses used SAS, version 9.1 (SAS Institute); graphical representations used Splus, version 6 (MathSoft).
RESULTS
A total of 5056 eligible ART-naive individuals initiated randomized ART in one of the 6 ALLRT parent studies; of these, 3381 (67%) enrolled into ALLRT (Figure 1). As of 30 June 2009, 4640 (92%) and 1122 (22%) had 1 and 6 years follow-up, respectively; of these, 4290 (92%) and 619 (55%) remained on their randomized ART regimen. Median study follow-up was 3.1 years; median follow-up on initial ART was 2.8 years. Participants were relatively young (77% were <45 years old), predominantly male (82%), and racially and ethnically diverse (Table 2).
Table 2.
Pretreatment Characteristics and Cardiovascular Risk Factors of the Study Population
|
Parent study (Sample size) |
Overall(5056) |
Enrolled in ALLRT |
||||||||
| Characteristics | 384 (887) | 388 (404) | A5014 (55) | A5095 (1131) | A5142 (731) | A5202(1848) | Yes (3381) | No (1675) | ||
| Demographic characteristic | ||||||||||
| Age, years | Median (P10, P90) | 36 (25, 50) | 37 (28, 50) | 35 (24, 50) | 37 (26, 50) | 38 (26, 51) | 38 (25, 51) | 37 (26, 51) | 37 (27, 50) | 36 (25, 49) |
| ≥45 | 179 (20) | 92 (23) | 12 (22) | 230 (20) | 174 (24) | 495 (27) | 1182 (23) | 815 (24) | 367 (22) | |
| Sex | Female | 157 (18) | 72 (18) | 9 (16) | 215 (19) | 137 (19) | 319 (17) | 910 (18) | 612 (18) | 298 (18) |
| Race/ethnicitya | Non-Hispanic white | 376 (42) | 173 (43) | 15 (27) | 458 (40) | 273 (37) | 743 (40) | 2037 (40) | 1479 (44) | 558 (33) |
| Non-Hispanic black | 329 (37) | 142 (35) | 17 (31) | 405 (36) | 293 (40) | 613 (33) | 1799 (36) | 1073 (32) | 726 (43) | |
| Hispanic | 159 (18) | 77 (19) | 22 (40) | 244 (22) | 146 (20) | 426 (23) | 1075 (21) | 728 (22) | 347 (21) | |
| Prior IVDU | 85 (10) | 49 (12) | 5 (9) | 120 (11) | 74 (10) | 158 (9) | 491 (10) | 305 (9) | 186 (11) | |
| Disease status | ||||||||||
| CD4+ cell countb, cells/mm3 | Median (P10, P90) | 271 (22, 598) | 91 (10, 398) | 370 (86, 528) | 209 (21, 483) | 191 (13, 424) | 229 (23, 434) | 215 (19, 472) | 218 (20, 475) | 211 (18, 466) |
| <200 | 342 (39) | 298 (74) | 10 (20) | 544 (48) | 378 (52) | 797 (43) | 2369 (47) | 1566 (46) | 803 (48) | |
| ≥350 | 331 (37) | 50 (12) | 26 (53) | 258 (23) | 138 (19) | 396 (21) | 1199 (24) | 796 (24) | 403 (24) | |
| HIV-1 RNAb, log10 copies/mL | Median (P10, P90) | 5.0 (3.8, 6.0) | 5.5 (4.8, 6.1) | 4.3 (3.7, 5.1) | 4.8 (4.1, 5.9) | 4.8 (4.0, 5.9) | 4.7 (3.9, 5.7) | 4.8 (4.0, 5.9) | 4.8 (4.0, 5.9) | 4.8 (4.0, 5.9) |
| ≥100,000 | 429 (48) | 332 (82) | 5 (10) | 476 (42) | 265 (36) | 474 (26) | 1963 (39) | 1283 (38) | 680 (41) | |
| Prior AIDS diagnosis | 124 (14) | 125 (31) | 4 (7) | 217 (19) | 145 (20) | 359 (19) | 974 (19) | 606 (18) | 368 (22) | |
| CVD risk factors | ||||||||||
| BMI (kg/m2)b | <18.5 | 25 (3) | 22 (6) | 1 (2) | 43 (4) | 27 (4) | 68 (4) | 186 (4) | 123 (4) | 63 (4) |
| 25–29.9 | 269 (31) | 88 (24) | 16 (29) | 351 (32) | 222 (31) | 598 (32) | 1545 (31) | 1069 (32) | 476 (28) | |
| ≥30 | 128 (15) | 36 (10) | 10 (18) | 152 (14) | 119 (17) | 734 (15) | 475 (14) | 259 (15) | 152 (16) | |
| MDRD GFRb,c, mL/min/1.73 m2 | <60 | 11 (1) | 8 (2) | 2 (4) | 11 (1) | 21 (3) | 21 (1) | 74 (1) | 46 (1) | 28 (2) |
| H× of hypertension | 86 (10) | 35 (9) | 12 (22) | 158 (14) | 181 (25) | 442 (24) | 926 (18) | 602 (18) | 324 (19) | |
| H× of dyslipidemia | Possible | 215 (24) | 114 (28) | 23 (42) | 81 (7) | 17 (2) | 33 (2) | 482 (10) | 352 (10) | 130 (8) |
| Confirmed | 13 (1) | 3 (1) | 1 (2) | 126 (11) | 135 (18) | 349 (19) | 627 (12) | 433 (13) | 194 (12) | |
| H× of diabetes | 14 (2) | 15 (4) | 1 (2) | 46 (4) | 25 (3) | 95 (5) | 196 (4) | 122 (4) | 74 (4) | |
| H× of CVD | Serious | 15 (2) | 3 (1) | … | 12 (1) | 14 (2) | 25 (1) | 80 (2) | 61 (2) | 19 (1) |
| Minor | 24 (3) | 21 (5) | 1 (2) | 32 (3) | 19 (3) | 78 (4) | 179 (4) | 119 (4) | 60 (4) | |
| CVD risk factorsd | 2 or more | 113 (13) | 45 (11) | 12 (22) | 136 (12) | 128 (18) | 320 (17) | 763 (15) | 509 (15) | 254 (15) |
| Ever smokede | 353 (62) | 121 (57) | 23 (52) | 464 (60) | 306 (55) | 663 (36) | 1930 (38) | 1930 (57) | … | |
| Family hx of CVDe | 113 (20) | 34 (16) | 9 (20) | 146 (19) | 93 (17) | 222 (12) | 633 (13) | 633 (19) | … | |
| Ten-year CVD risk score (%)f | ≥10 | … | … | … | 47 (11) | 49 (9) | 170 (9) | 271 (5)) | 271 (8)) | … |
NOTE. Data are no. (%) of patients, unless otherwise indicated. ALLRT, AIDS Clinical Trials Group (ACTG) Longitudinally Linked Randomized Clinical Trials cohort; BMI, body mass index calculated as the weight in kilograms divided by the square of height in meters; CVD, cardiovascular disease; hx, history; IVDU, intravenous drug use; MDRDGFR, modification of diet in renal disease glomerular filtration rate; (P10, P90), 10th and 90th percentiles.
The study population also includes 87 participants of Asian origin; 30 participants of native American/Alaskan origin, and 22 with other self-identified race/ethnicity.
Percentages are calculated for participants with data; 12 participants are missing pretreatment CD4+ cell count data; 6 participants are missing pretreatment HIV-1 RNA data; 3 participants are missing pretreatment serum creatinine data; 92 participants are missing pretreatment weight or height data.
GFR based on a 4-variable MDRD formula [29].
Number of CVD risk factors from: extreme BMI, MDRD GFR <60 mL/min/1.73 m2, history of hypertension, dyslipidemia, diabetes, or CVD.
Smoking status, family history of CVD, and 10-year CVD risk score were only evaluated for subjects enrolled into ALLRT. Percentages are calculated for subjects with available data; 1678 participants did not have smoking evaluation data, 1680 participants did not have family history of CVD data, and 2887 participants did not have 10-year CVD risk score data.
Based on Framingham hard coronary heart disease equation [32].
Consistent with designs of the parent studies, approximately 44% of subjects initiated an NRTI plus NNRTI regimen, 20% initiated a regimen containing an unboosted PI, and 29% initiated a regimen containing a boosted PI; 34% initiated an abacavir-containing regimen (54% in protocol A5202, 45% in protocol A5095, and 1% in protocol A5014), 9% initiated a didanosine-containing regimen, and 21% initiated a regimen containing tenofovir disoproxil fumarate (85% in A5202 and 15% in A5142). Of 1704 subjects who initiated an abacavir-containing regimen, 74% and 44% of those still in follow-up continued to receive abacavir at 1 and 6 years, respectively; of 3352 subjects who initiated a non–abacavir-containing regimen, 2% had initiated abacavir by 1 year (8% by 6 years) (Table 3).
Table 3.
. Person-years of Follow-up and Observed Myocardial Infarction (MI) Events
| Abacavir strategy | No. of patients initiating randomized ART | Follow-up period | All follow-up |
Off strategy follow-upa |
||||||
| PYb | No. of MIs | Timing of MI, yearsc | Incidence, events/1000 PY (95% CI) | No. of patients discontinuing strategy (% of total) | PY (% of total) | No. of off-strategy MIsd | Timing of MI, yearse | |||
| Abacavir | 1704 | Up to 1 year | 1586 | 3 | 0.9 | 1.9 (0.0–4.0) | 442 (26) | 358 (23) | 1 | 0.9 |
| Up to 6 years | 5452 | 8 | 1.0 | 1.5 (0.5–2.5) | 915 (54) | 1414 (26) | 2 | 2.7 | ||
| No abacavir | 3352 | Up to 1 year | 3097 | 9 | 0.5 | 2.9 (1.0–4.8) | 52 (2) | 11 (<1) | 0 | - |
| Up to 6 years | 11,952 | 28 | 1.8 | 2.3 (1.5–3.2) | 271 (8) | 451 (4) | 3 | 1.2 | ||
NOTE. ART, antiretroviral treatment; CI, confidence interval; MI, myocardial infarction; PY, person-years.
For subjects randomized to an abacavir-containing regimen, strategy discontinuation is defined as discontinuation of abacavir; for subjects randomized to a non–abacavir-containing regimen, it is defined as initiation of an abacavir-containing regimen.
Follow-up is included from start of ART until first event, death, or last evaluation plus 6 months.
The median follow-up time (in years) at MI event across subjects with MI.
These events are a subset of those included in the overall total number of events.
The median years of off-strategy follow-up before MI (across subjects with MI after discontinuation of their randomized abacavir strategy).
Over the first year of ART, 12 MIs were observed (3 in the abacavir group and 9 in the non-abacavir group, over 1586 and 3097 person-years (PYs) of follow-up, respectively) (Table 3). Twenty-four additional MIs were observed over a subsequent follow-up to 6 years (5 in the abacavir group and 19 in the non-abacavir group). The estimated cumulative MI incidence over 6 years was 1.0% and 1.2% for the abacavir and non-abacavir groups, respectively.
In unadjusted and adjusted analyses, associations between abacavir use as part of initial ART and the hazard of MI were not detected either in the short term or in the long term (unadjusted short-term analysis, P = .52; unadjusted long-term analysis, P = .21; adjusted short-term analysis, P = .60; adjusted long-term analysis, P = .25; Table 4). Factors associated with an increased hazard of MI in the cohort included increased age (short-term analysis, P = .08; long-term analysis, P < .001), a pretreatment history of 2 or more CVD risk factors (P = .005; P = .008), and current or prior smoking history (P = .16; P = .05). There was no evidence of interactions between abacavir use as part of initial treatment and age (P = .46; P = .98) or the presence of 2 or more CVD risk factors (P = .46; P = .10). No evidence of violation of proportional hazard assumption was found (P = .89). Lower age, male sex, minority race/ethnicity, a history of injection drug use, documented pretreatment diabetes, detectable HIV-1 RNA level, the absence of dyslipidemia, and continuing to receive an initial ART regimen were associated with a higher probability of being censored from our cohort (data not shown). However, IPCW did not substantially change our model estimates or inference. Consistent results were observed across all of our sensitivity analyses (Table 5). The most extreme results came from analyses that included only contemporary regimens. This included 1677 subjects with 8 MIs in the abacavir group and 1819 subjects with 5 MIs in the non-abacavir group (hazard ratio, 1.7; P = .33).
Table 4.
Cox Proportional Hazards Model
| 1 Year |
0ver 6 years (unweighted) |
Over 6 years (weighteda) |
|||||
| Covariate | Hazard ratio (95% CI) | P | Hazard ratio (95% CI) | P | Hazard ratio (95% CI) | P | |
| Abacavir in the initial regimen (vs no abacavir) | |||||||
| Unadjusted | 0.7 (0.2–2.4) | .52 | 0.6 (0.3–1.3) | .21 | 0.7 (0.3–1.5) | .35 | |
| Adjustedb | 0.7 (0.2–2.6) | .60 | 0.6 (0.3–1.4) | .25 | 0.7 (0.3–1.6) | .40 | |
| Demographic characteristic | |||||||
| Age | per 10 years | 1.5 (1.0–2.2) | .08 | 1.7 (1.3–2.3) | <.001 | 1.7 (1.3–2.3) | <.001 |
| Sex (vs Male) | Female | 0.4 (0.1–2.8) | .34 | 0.4 (0.1–1.4) | .16 | 0.4 (0.1–1.3) | .11 |
| Race/ethnicity(vs non-Hispanic white) | Non-Hispanic black | 0.9 (0.3–2.9) | .84 | 0.6 (0.3–1.2) | .14 | 0.6 (0.3–1.2) | .14 |
| Hispanic | 0.4 (0.0–3.3) | .40 | 0.5 (0.2–1.4) | .16 | 0.6 (0.2–1.8) | .35 | |
| Other | 3.3 (0.4–27.1) | .27 | 0.9 (0.1–6.7) | .89 | 0.9 (0.1–6.8) | .91 | |
| CVD risk factors | |||||||
| CVD risk factors (vs <2) | 2 or more | 4.6 (1.6–13.4) | .005 | 2.6 (1.3–5.1) | .008 | 2.8 (1.3–5.6) | .006 |
| Ever smoked (vs Never) | 4.5 (0.5–36.6) | .16 | 2.6 (1.0–6.9) | .048 | 3.1 (1.2–8.3) | .023 | |
| Family history of CVD (vs no history) | Yes | 1.7 (0.4–6.9) | .43 | 1.8 (0.9–3.9) | .12 | 1.9 (0.9–4.2) | .10 |
NOTE. CI, confidence interval; CVD, cardiovascular disease.
a Weighted using inverse probability of censoring weights for which lower age, male sex, minority race/ethnicity, a history of IV drug use, documented pretreatment diabetes, detectable HIV-1 RNA level, the absence of dyslipidemia and remaining on initial ART regimen were all associated with a higher probability censoring.
b Adjusted for other covariates listed.
Table 5.
Sensitivity Analyses for the Estimated Adjusted Hazard Ratio (HR) for Initial Abacavir Use
| MI |
Serious CVD events |
|||||||||
| ABC strategy | No. of subjects | 1-Year period |
6-Year period |
1-Year period |
6-Year period |
|||||
| Analysis | No. of events/PY | HR (95% CI) | No. of events/PY | HR (95% CI) |
No. of events/PY | HR (95% CI) |
No. of events/PY | HR (95% CI) |
||
| Full cohort | ABC No ABC |
1704 3352 |
3/1586 9/3097 |
0.7 (0.2–2.6) |
8/5452 28/11952 |
0.6 (0.3–1.4) |
13/1583 24/3089 |
1.1 (0.5–2.1) |
27/5426 70/11857 |
0.9 (0.5–1.3) |
| As-treateda | ABC No ABC |
1704 3352 |
2/1228 9/3086 |
0.6 (0.1–2.6) |
7/4038 25/11501 |
0.7 (0.3–1.9) |
10/1226 23/3078 |
1.1 (0.5–2.2) |
17/4022 62/11416 |
0.8 (0.4–1.3) |
| Including only studies with randomized abacavir | ABC No ABC |
1704 1330 |
3/1586 2/1245 |
1.3 (0.2–8.5) |
8/5452 5/3966 |
1.3 (0.5–3.9) |
13/1583 5/1244 |
2.2 (0.8–6.2) |
27/5426 18/3945 |
0.7 (0.3–1.6) |
| Including only subjects randomized to contemporary regimensb | ABC No ABC |
1677 1819 |
3/1562 3/1699 |
1.3 (0.2–6.8) |
8/5344 6/5899 |
1.7 (0.6–5.1) |
13/1559 8/1696 |
2.1 (0.9–5.1) |
27/5319 23/5860 |
1.2 (0.7–2.1) |
| Excluding subjects randomized to didanosine or a PI | ABC No ABC |
1244 1219 |
3/1155 3/1136 |
1.2 (0.2–6.7) |
8/4323 7/4267 |
1.2 (0.4–3.2) |
9/1152 7/1133 |
1.5 (0.5–4.0) |
23/4299 23/4238 |
1.0 (0.6–1.8) |
| Including all MIs regardless of investigator review | ABC No ABC |
1704 3352 |
5/1585 9/3097 |
1.2 (0.4–3.6) |
10/5450 30/11952 |
0.8 (0.4–1.7) |
||||
NOTE. All estimates are adjusted with covariates as listed in Table 4; 6-year estimates are based on inverse probability of censoring weights (IPCW) analysis. ABC, abacavir; CI, confidence interval; CVD, cardiovascular disease; HR, hazard ratio; MI, myocardial infarction; PI, protease inhibitor; PY, person-years.
Model fitting for as-treated analyses include IPCW for discontinuation of abacavir strategy.
Contemporary regimens are defined as those containing either a nonnucleoside reverse-transcriptase inhibitor (NNRTI) or ritonavir boosted PI with 2 or 3 NRTIs not including stavudine.
Figure 2 shows model-based estimates of cumulative incidence of MI over time. The results are displayed for the full cohort and most extreme sensitivity analysis. They highlight the very low absolute MI risk in this population and demonstrate that, with the exception of the very high risk group (older smokers with 2 or more CVD risk factors), a small absolute risk difference is associated with abacavir use, even when estimated from our most extreme sensitivity analysis.
Figure 2.
Estimated cumulative incidence of myocardial infarction (MI) from 6-year inverse probability of censoring weights (IPCW) model for a hypothetical set of individuals defined by cardiovascular disease (CVD) risk factors: A, 55-year-old smoker with ≥2 CVD risk factors; B, 55-year-old smoker with <2 CVD risk factors; C, 55-year-old nonsmoker with ≥2 CVD risk factors. D, 55-year-old nonsmoker with <2 CVD risk factors; E–H, 35-year-old individuals with smoking and CVD risk factors as described for A–D. Cumulative incidence is based on IPCW Cox proportional hazards model including abacavir use and significant covariates (P < .05) from Table 4; 95% confidence intervals at 1 and 6 years are from 1000 bootstrapped samples. Solid lines are based on results from the full cohort; dotted lines are based on the sensitivity analyses including only contemporary regimens. Gray lines with circles indicate the cohort randomized to an abacavir regimen; black lines with squares indicate a non–abacavir-containing regimen.
To enhance the power of our study to analyze the adverse CVD effect associated with abacavir, the event of interest was expanded to include additional pre-specified serious CVD events (Table 1). A total of 37 events were observed during the first year of ART; 97 events (including the 36 MIs) were observed over a maximum of 6 years of follow-up (17,283 PY). The findings from these analyses gave results that were generally consistent with those for MI events (Table 5).
DISCUSSION
This analysis of 5056 individuals who initiated ART in 6 randomized clinical trials aims to contribute to clinical decision-making with respect to risk for CVD events and choice of initial ART regimens, specifically with respect to abacavir. Overall MI incidence in our cohort was 2.1 events per 1000 PY (95% CI, 1.8–3.7), compared with between 3.1–5.0 events per 1000 PY from prior studies in this area [3, 5–7], demonstrating a low absolute risk of MI in this typical HIV-1 cohort of individuals initiating ART. We did not find evidence of increased short-term or long-term risk of MI or of serious CVD events associated with abacavir use as part of initial ART. This result was robust in as-treated analysis and a range of sensitivity analyses. An increased risk of MI events was detected for persons of older age and with classic CVD risk factors, such as smoking and prior CVD history.
Our analysis was unique in including a long-term cohort of HIV-1–infected persons initiating ART in randomized clinical trials, and our intent-to-treat analysis with adjustment only for pretreatment covariates provides a broad population-level perspective of the risk of MI associated with abacavir-containing initial regimens. This analytic framework contrasts with use of time-updated covariates for drug exposure and/or risk factors [4, 6, 7, 10] or censoring follow-up at discontinuation of abacavir [5] that have been used in other studies. Although such studies are attractive for their simplicity, in the presence of related and temporally confounded effects, adjustment for time-updated effects is likely to induce or attenuate the estimated risk associated with the exposure of interest [16, 33]. Although the as-treated censoring approach is also often preferred because it is seen to more clearly isolate the direct effect of the exposure of interest (in this case, abacavir use), not only does this have the potential to introduce an informative censoring bias, it is important to note that such censoring changes the question being addressed to one that is a purely hypothetical scenario: estimating the effect of abacavir use under the condition that, once treatment with abacavir is started, it is never discontinued. Because toxicity issues, individual ART fatigue, and regimen failure dictate that most regimens will at some time be discontinued or modified, the resulting parameter estimates do not relate to real-life situations. For completeness, we performed an as-treated analysis as a sensitivity analysis and found consistent conclusions.
A criticism of the intent-to-treat approach is that, although it is suited for the analysis of randomized trials because of cross-over between strategies, it is not relevant in the context of understanding the effect of specific drugs in cohort data. On the contrary, such inference remains critically important in understanding population-level drug effects that are implicitly adjusted for prescribing biases and drug-taking behavior of the population under study—in our case a relatively young (median age, 37 years) cohort with well-controlled HIV disease over the course of follow-up—overall, 83% of the total follow-up period was spent with HIV-1 RNA levels <400 copies/mL, and 63% was spent with CD4+ cell counts >350 cells/mm3.
IPCW methodology was used in our long-term analysis to investigate the influence of informative censoring bias due to premature study discontinuation and individuals not enrolling into ALLRT. Noninformative censoring is a fundamental assumption of any survival analysis that is often not fully scrutinized. Assuming that subjects with a given covariate history who are no longer in follow-up can be represented by similar subjects who remain in follow-up, IPCWs adjust for potential informative censoring by weighting uncensored data to create a pseudo-population in which censoring does not exist [33, 34]. IPCW methodology did not dramatically impact our results. This could mean that informative censoring is not a concern for this analysis, that informative censoring exists but differential effects cancel each other out, or that our model for informative censoring has failed to adjust for the key variables that drive the informative censoring. This latter scenario reflects the failure of the key assumption of no unmeasured confounding that underpins all IPCW methodology. Of note, almost half of the censoring in our analysis was administrative, which can reasonably be assumed to be noninformative.
One study limitation concerns the power to detect a difference between abacavir-containing and non–abacavir-containing initial regimens, because only 36 MI events were observed during 6 years (17,404 PY) of follow-up. Furthermore, some of the comparator groups contained ART agents that are no longer commonly used and other agents that may have been implicated with adverse lipid profiles or increased CVD risk [1, 35, 36]. However, studies of similar size to our own have demonstrated significant results [4, 11], and our sensitivity analyses that restricted contemporary regimens (those including 2 or 3 NRTIs with NNRTI or boosted-PI) and excluded regimens that have also been implicated with increased CVD risk provided consistent conclusions to our main result.
The strength of our cohort is randomized ART assignment that lessens the potential for confounding and channeling bias that can complicate interpretation of treatment associations from observational data. These are particularly problematic for investigating MI risk with abacavir, because past prescribing practices favored abacavir regimens for individuals at high CVD risk [6, 12–14]. Because only 3 of the studies included in our analysis involved a randomized abacavir strategy, we were not able to estimate the direct effect of the abacavir randomization in a classic meta-analysis sense [37]. Rather, our approach relies on the assumption that the relative effect of interest is consistent across included studies. That is, the effect of abacavir in non-abacavir studies—had they included randomized abacavir—is assumed to be the same as that in the studies with randomized abacavir. Such an assumption is reasonable. All of the studies were conducted at the same clinical sites, had similar entry criteria, and, with the exception of a brief period early in follow-up, used the same mechanism for MI identification. Furthermore, sensitivity analyses restricting the cohort only to studies including randomized abacavir were consistent with findings of the entire cohort.
In conclusion, we studied individuals who initiated ART with and without abacavir as part of randomized clinical trials. Over a period of up to 6 years from ART initiation, we observed a very low absolute risk of MI and serious CVD events. We found no evidence of an increased risk of MI or serious CVD associated with the use of abacavir as part of initial treatment over the first year of ART and the longer term that was consistent in as-treated and sensitivity analyses. Classic CVD risk factors were the strongest predictors of MI and serious CVD events and should be the main focus in assessing CVD risk among HIV-1 infected individuals.
Acknowledgments
We thank the study volunteers who participate in ALLRT (A5001), and participated in ACTG 384, ACTG 388, A5014, A5095, A5142, and A5202; all the ACTG clinical units that enrolled patients in these ART naive trials and continue to enroll and follow patients in A5001, and the ACTG; Frontier Science Foundation for data management; and other members of the A5001 protocol team. We gratefully acknowledge the work of the contributing protocol teams, in particular the study chairs: G. Robbins and R. Shafer for ACTG 384, M. Fischl for ACTG 388, A. Landay and M. Lederman for ACTG A5014, R. Gulick for A5095, R. Haubrich and S. Riddler for A5142, and P. Sax and E. Daar for A5202. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Allergy and Infectious Diseases or the National Institutes of Health.
Financial support. The AIDS Clinical Trials Group, funded by the National Institute of Allergy and Infectious Diseases (NIAID) AI 68636, AI 38858, AI 68634, AI 38855, AI069474, AI069434. The following NIAID funded institutions (NIAID grant support given in parentheses) participated in the contributing AIDS Clinical Trials Group studies: Partners/Harvard AIDS CTU (AI69472), Johns Hopkins Adult AIDS CTU (AI69465), NYU/Bellevue/NYU HIV/AIDS CTU (AI69532), Stanford Univ. AIDS CTU (AI69556), UCLA AIDS Prevention & Treatment CTU (AI69424), UCSD Antiviral Research Ctr. CTU (AI69432), UCSF Adult AIDS CTU (AI69502), Univ. of Miami HIV/AIDS CTU (AI69477), Univ. of Pittsburgh CTU (AI69494), Univ. of Rochester CTU (AI69511), University of Southern California CTU (AI69428), University of Washington Clinical HIV Research Program (CHIRP) CTU (AI69434), University of Minnesota CTU (AI27661), Duke Univ. Med. Ctr. HIV/AIDS CTU (AI69484), Tulane-LSU AIDS CTU (AI38844), Mt. Sinai Medical Center CTU (AI46370), Wash. U AIDS CTU (AI069495, AI25903), The Ohio State Univ. AIDS CTU (AI69474), University of Cincinnati ACTU (AI69513), Case AIDS CTU (AI69501), Indiana University CTU (AI25859), Northwestern Univ. CTU (AI69471, AI25915), Beth Israel Medical Center CTU (AI46370), Miriam Hospital CTU (AI46381), UNC AIDS CTU (AI69423), Vanderbilt HIV CTU (AI69439), University of Texas Southwestern Medical Center CTU (AI46376-05S4), University of California, Davis ACTU (AI38858); Inst. of Human Virology (HIV Baltimore Treatment CRS) CTU (AI69447), University of Hawaii at Manoa, Leahi Hospital CTU (AI34853), Howard University CTU (AI34835), Univ. of Puerto Rico CTU (AI69415), Alabama CTU (AI69452), Emory HIV/AIDS CTU (AI69418), Colorado ACTU (AI69450), Children's Hosp of Philadelphia International Maternal Ped. Adolescent AIDS CTU (AI69467), University of Pennsylvania CTU (AI32783), University of Texas, Galveston CTU (Al32782), Cornell CTU (AI69419), Istituto Superiore di Sanita (A7901) (Grant: HIV Clinical Research Programme, Istituto Superiore di Sanità 1999–2002), Georgetown University Hospital (A9327), Emory Hemo Comp Evaluation Clinic (A9413), Louisiana Comp. Hemo. Care Ctr - Tulane (A9426), Columbia Collaborative HIV/AIDS CTU (AI69470), Terry Beirn Community Programs for Clinical Research on AIDS CTU (AI69503), Durban International CTU (AI69426), Moses H Cone Memorial Hospital (AI69423), Rush University Medical Center (AI69471), Hennepin County Medical Center (AI72626), Peabody Health Center (AI69471), University of Nebraska Medical Center (AI27661), IHV Baltimore Treatment CRS (AI69447), SUNY- Buffalo (A1027658), San Mateo County AIDS Program (AI27666), University of Iowa Healthcare (AI27661, AI58740), Wake County HHS (AI25868), Harlem ACTG CRS (AI69470), and McCree McCuller Wellness Center (AI69511).
Potential conflicts of interest. C.A.B. has served as a scientific advisor to Merck, GlaxoSmithKline, and ViiV; has received research contract support from Gilead; and has served as the chair of a Data Safety and Monitoring Board for Achillion and Johnson & Johnson, Australia. A.C.C. has served as a scientific advisor to GlaxoSmithKline, Merck, and Pfizer; has received research contract support from Gilead Sciences, Johnson & Johnson, Merck, and Schering-Plough; and is a member of a Data Safety and Monitoring Board for Merck. R.J.B. has served as a Scientific Advisor for Tibotec. All other authors: no conflicts.
References
- 1.Guidelines for the use of antiretroviral agents in HIV-1-infected adults and adolescents [homepage on the Internet] 2009. US Department of Health and Human Services (DHHS). Available from: http://www.aidsinfo.nih.gov.ezp-prod1.hul.harvard.edu/ContentFiles/AdultandAdolescentGL.pdf. Accessed 1 December 2009. [Google Scholar]
- 2.Clinical management and treatment of HIV infected adults in Europe. 2009. Available from: http://www.europeanaidsclinicalsociety.org/Guidelines2009/G2.htm. Accessed 8 August 2010. [PubMed] [Google Scholar]
- 3.D:A:D Study Group; Sabin CA, Worm SW, Weber R, et al. Use of nucleoside reverse transcriptase inhibitors and risk of myocardial infarction in HIV-infected patients enrolled in the D:A:D study: a multi-cohort collaboration. Lancet. 2008;371:1417–1426. doi: 10.1016/S0140-6736(08)60423-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Strategies for Management of Anti-Retroviral Therapy/INSIGHT, DAD Study Groups. Use of nucleoside reverse transcriptase inhibitors and risk of myocardial infarction in HIV-infected patients. AIDS. 2008;22:F17–F24. doi: 10.1097/QAD.0b013e32830fe35e. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Brothers CH, Hernandez JE, Cutrell AG, et al. Risk of myocardial infarction and abacavir therapy: no increased risk across 52 GlaxoSmithKline-sponsored clinical trials in adult subjects. J Acquir Immune Defic Syndr. 2009;51:20–28. doi: 10.1097/QAI.0b013e31819ff0e6. [DOI] [PubMed] [Google Scholar]
- 6.Bedimo R, Westfall A, Dreschler H, Tebas P. [abstract MOAB202]. In: Program and abstracts of the 5th IAS Conference on HIV Pathogenesis, Treatment and Prevention, July 19–22. Cape Town, South Africa: 2009. Abacavir use and risk of acute myocardial infarction and cerebrovascular disease in the HAART era. [Google Scholar]
- 7.Durand M, Sheehy O, Baril JG, Lelorier J, Tremblay C. 5th IAS Conference on HIV Pathogenesis, Treatment and Prevention, July 19–22. Cape Town, South Africa: 2009. Relation between use of nucleoside reverse transcriptase inhibitors (NRTI) and risk of myocardial infarction (MI): a nested case control study using quebec's public health insurance database (QPHID). [abstract TUPEB175] [Google Scholar]
- 8.Martinez E, Larrousse M, Perez I, et al. [abstract MOAB203]. 5th IAS Conference on HIV Pathogenesis, Treatment and Prevention, July 19–22. Cape Town, South Africa: 2009. No evidence for recent abacavir/lamivudine use in promoting inflammation, endothelial dysfunction, hypercoagulability, or insulin resistance in virologically suppressed HIV-infected patients: a substudy of the BICOMBO randomized clinical trial. [Google Scholar]
- 9.Martin A, Bloch M, Amin J, et al. Simplification of antiretroviral therapy with tenofovir-emtricitabine or abacavir-lamivudine: a randomized, 96-week trial. Clin Infect Dis. 2009;49:1591–1601. doi: 10.1086/644769. [DOI] [PubMed] [Google Scholar]
- 10.Lang S, Mary-Krause M, Cotte L, et al. Impact of individual antiretroviral drugs on the risk of myocardial infarction in human immunodeficiency virus-infected patients: a case-control study nested within the French hospital database on HIV ANRS cohort CO4. Arch Intern Med. 2010;170:1228–1238. doi: 10.1001/archinternmed.2010.197. [DOI] [PubMed] [Google Scholar]
- 11.Obel N, Farkas DK, Kronborg G, et al. Abacavir and risk of myocardial infarction in HIV-infected patients on highly active antiretroviral therapy: a population-based nationwide cohort study. HIV Med. 2010;11:130–136. doi: 10.1111/j.1468-1293.2009.00751.x. [DOI] [PubMed] [Google Scholar]
- 12.Post FA, Campbell LJ. Abacavir and increased risk of myocardial infarction. Lancet. 2008;372:803. doi: 10.1016/S0140-6736(08)61330-6. author reply 804–805. [DOI] [PubMed] [Google Scholar]
- 13.Keiser PH, Sension MG, DeJesus E, et al. Substituting abacavir for hyperlipidemia-associated protease inhibitors in HAART regimens improves fasting lipid profiles, maintains virologic suppression, and simplifies treatment. BMC Infect Dis. 2005;5:2. doi: 10.1186/1471-2334-5-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Schambelan M, Benson CA, Carr A, et al. Management of metabolic complications associated with antiretroviral therapy for HIV-1 infection: recommendations of an international AIDS Society–USA panel. J Acquir Immune Defic Syndr. 2002;31:257–275. doi: 10.1097/00126334-200211010-00001. [DOI] [PubMed] [Google Scholar]
- 15.Robins JM, Blevins D, Ritter G, Wulfsohn M. G-estimation of the effect of prophylaxis therapy for Pneumocystis carinii pneumonia on the survival of AIDS patients. Epidemiology. 1992;3:319–336. doi: 10.1097/00001648-199207000-00007. [DOI] [PubMed] [Google Scholar]
- 16.Sterne JA, Hernan MA, Ledergerber B, et al. Long-term effectiveness of potent antiretroviral therapy in preventing AIDS and death: a prospective cohort study. Lancet. 2005;366:378–384. doi: 10.1016/S0140-6736(05)67022-5. [DOI] [PubMed] [Google Scholar]
- 17.Hughes MD, Williams PL. Challenges in using observational studies to evaluate adverse effects of treatment. N Engl J Med. 2007;356:1705–1707. doi: 10.1056/NEJMp078038. [DOI] [PubMed] [Google Scholar]
- 18.Smurzynski M, Collier AC, Koletar SL, et al. AIDS clinical trials group longitudinal linked randomized trials (ALLRT): rationale, design, and baseline characteristics. HIV Clin Trials. 2008;9:269–282. doi: 10.1310/hct0904-269. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Robbins GK, De Gruttola V, Shafer RW, et al. Comparison of sequential three-drug regimens as initial therapy for HIV-1 infection. N Engl J Med. 2003;349:2293–2303. doi: 10.1056/NEJMoa030264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Shafer RW, Smeaton LM, Robbins GK, et al. Comparison of four-drug regimens and pairs of sequential three-drug regimens as initial therapy for HIV-1 infection. N Engl J Med. 2003;349:2304–2315. doi: 10.1056/NEJMoa030265. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Fischl MA, Ribaudo HJ, Collier AC, et al. A randomized trial of 2 different 4-drug antiretroviral regimens versus a 3-drug regimen, in advanced human immunodeficiency virus disease. J Infect Dis. 2003;188:625–634. doi: 10.1086/377311. [DOI] [PubMed] [Google Scholar]
- 22.Landay AL, Spritzler J, Kessler H, et al. Immune reconstitution is comparable in antiretroviral-naive subjects after 1 year of successful therapy with a nucleoside reverse-transcriptase inhibitor- or protease inhibitor-containing antiretroviral regimen. J Infect Dis. 2003;188:1444–1454. doi: 10.1086/379041. [DOI] [PubMed] [Google Scholar]
- 23.Gulick RM, Ribaudo HJ, Shikuma CM, et al. Triple-nucleoside regimens versus efavirenz-containing regimens for the initial treatment of HIV-1 infection. N Engl J Med. 2004;350:1850–1861. doi: 10.1056/NEJMoa031772. [DOI] [PubMed] [Google Scholar]
- 24.Gulick RM, Ribaudo HJ, Shikuma CM, et al. Three- vs four-drug antiretroviral regimens for the initial treatment of HIV-1 infection: a randomized controlled trial. JAMA. 2006;296:769–781. doi: 10.1001/jama.296.7.769. [DOI] [PubMed] [Google Scholar]
- 25.Riddler SA, Haubrich R, DiRienzo AG, et al. Class-sparing regimens for initial treatment of HIV-1 infection. N Engl J Med. 2008;358:2095–2106. doi: 10.1056/NEJMoa074609. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Sax PE, Tierney C, Collier AC, et al. Abacavir-lamivudine versus tenofovir-emtricitabine for initial HIV-1 therapy. N Engl J Med. 2009;361:2230–2240. doi: 10.1056/NEJMoa0906768. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Tunstall-Pedoe H, Kuulasmaa K, Amouyel P, Arveiler D, Rajakangas AM, Pajak A. Myocardial infarction and coronary deaths in the world health organization MONICA project. Registration procedures, event rates, and case-fatality rates in 38 populations from 21 countries in four continents. Circulation. 1994;90:583–612. doi: 10.1161/01.cir.90.1.583. [DOI] [PubMed] [Google Scholar]
- 28.Satagopan JM, Ben-Porat L, Berwick M, Robson M, Kutler D, Auerbach AD. A note on competing risks in survival data analysis. Br J Cancer. 2004;91:1229–1235. doi: 10.1038/sj.bjc.6602102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D. A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. modification of diet in renal disease study group. Ann Intern Med. 1999;130:461–470. doi: 10.7326/0003-4819-130-6-199903160-00002. [DOI] [PubMed] [Google Scholar]
- 30.Hernan MA, Brumback B, Robins JM. Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology. 2000;11:561–570. doi: 10.1097/00001648-200009000-00012. [DOI] [PubMed] [Google Scholar]
- 31.Putter H, Fiocco M, Geskus RB. Tutorial in biostatistics: competing risks and multi-state models. Stat Med. 2007;26:2389–2430. doi: 10.1002/sim.2712. [DOI] [PubMed] [Google Scholar]
- 32.US Department of Health and Human Services (DHHS) Third report of the expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (adult treatment panel III) US Department of Health and Human Services (DHHS), National Institute of Heart, Lung, and Blood (NIHLB). Available from: http://www.nhlbi.nih.gov/guidelines/cholesterol/index.htm. Accessed April 30, 2010. [Google Scholar]
- 33.Hernan MA, Hernandez-Diaz S, Robins JM. A structural approach to selection bias. Epidemiology. 2004;15:615–625. doi: 10.1097/01.ede.0000135174.63482.43. [DOI] [PubMed] [Google Scholar]
- 34.Hernan MA. A definition of causal effect for epidemiological research. J Epidemiol Community Health. 2004;58:265–271. doi: 10.1136/jech.2002.006361. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.DAD Study Group, Friis-Moller N, Reiss P, et al. Class of antiretroviral drugs and the risk of myocardial infarction. N Engl J Med. 2007;356:1723–1735. doi: 10.1056/NEJMoa062744. [DOI] [PubMed] [Google Scholar]
- 36.Hill A, Sawyer W, Gazzard B. Effects of first-line use of nucleoside analogues, efavirenz, and ritonavir-boosted protease inhibitors on lipid levels. HIV Clin Trials. 2009;10:1–12. doi: 10.1310/hct1001-001. [DOI] [PubMed] [Google Scholar]
- 37.Chou R, Fu R, Huffman LH, Korthuis PT. Initial highly-active antiretroviral therapy with a protease inhibitor versus a non-nucleoside reverse transcriptase inhibitor: discrepancies between direct and indirect meta-analyses. Lancet. 2006;368:1503–1515. doi: 10.1016/S0140-6736(06)69638-4. [DOI] [PubMed] [Google Scholar]