Abstract
Objective(s):
Heart failure (HF) risk is elevated in people with HIV (PWH). We investigated whether initial antiretroviral therapy (ART) regimens influenced HF risk.
Design:
Cohort study
Methods:
PWH who initiated an ART regimen between 2000–2016 were identified from three integrated healthcare systems. We evaluated HF risk by protease inhibitor (PI), non-nucleoside reverse transcriptase inhibitors (NNRTI), and integrase strand transfer inhibitor (INSTI)-based ART, and comparing two common nucleotide reverse transcriptase inhibitors: tenofovir disoproxil fumarate (tenofovir) and abacavir. Follow-up for each pairwise comparison varied (i.e., 7 years for PI vs. NNRTI; 5 years for tenofovir vs. abacavir; 2 years for INSTIs vs. PIs or NNRTIs). Hazard ratios (HRs) were from working logistic marginal structural models, fitted with inverse probability weighting to adjust for demographics, and traditional cardiovascular risk factors.
Results:
13,634 PWH were included (88% men, median 40 years of age; 34% non-Hispanic white, 24% non-Hispanic black, and 24% Hispanic). The HR (95% CI) were: 2.5 (1.5–4.3) for PI vs. NNRTI-based ART (reference); 0.5 (0.2–1.8) for PI vs. INSTI-based ART (reference); 0.1 (0.1–0.8) for NNRTI vs. INSTI-based ART (reference); and 1.7 (0.5–5.7) for tenofovir vs. abacavir (reference). In more complex models of cumulative incidence that accounted for possible non-proportional hazards over time, the only remaining finding was evidence of a higher risk of HF for PI compared with NNRTI-based regimens (1.8% vs. 0.8%; P=0.002).
Conclusions:
PWH initiating PIs may be at higher risk of HF compared with those initiating NNRTIs. Future studies with longer follow-up with INSTI-based and other specific ART are warranted.
Keywords: Heart failure, HIV, antiretroviral therapy, causal inference, epidemiology
Introduction
Heart failure is a large and growing public health problem in the general population affecting ~6 million Americans.[1] Persons with HIV (PWH) are known to experience a high burden of cardiovascular disease (CVD) compared with persons without HIV (PWoH)[2, 3], although HF has received very limited attention given the large sample sizes needed for the relatively rare outcome in PWH. Nevertheless, a few studies have noted an increased risk for HF in PWH, including the Veterans Aging Cohort Study (VACS),[4] noting an 81% higher relative risk of HF among Veterans with and without HIV. We also recently reported on a multicenter analysis in Kaiser Permanente (KP) showing a 73% higher adjusted risk of HF in PWH not explained by differences in clinical cardiovascular risk factors.[5] A follow-up study demonstrated that the risk of HF may also be mediated by HIV clinical factors including low CD4 cell counts.[6]
While traditional risk factors (e.g., tobacco, alcohol) in PWH may be contributing to the higher burden of CVD, toxicity from antiretroviral drugs may also be accelerating the CVD epidemic in these patients. For example, the excess risk of CVD has been observed with use of protease inhibitors (PI)[7], and non-nucleoside reverse transcriptase inhibitors (NNRTIs) (e.g., efavirenz[8]), but not integrase strand transfer inhibitors (INSTI).[9, 10] Conflicting data exist for nucleoside reverse transcriptase inhibitors (NRTIs), especially abacavir, with some studies[11], including our KP study[12], suggesting a higher risk of CVD with abacavir, while others have reported no significant association.[13] Importantly, few studies included HF as part of composite CVD endpoints,[14, 15] while only one study in US Veterans has evaluated the association of antiretroviral therapy (ART) specifically with HF and reported a 50% higher risk of both CVD and HF, but the association was not significant for HF.[16] The same study also noted that tenofovir disoproxil fumarate (tenofovir) use, was associated with an 80% higher risk of HF, but not other CVD.[16]
Thus, a comprehensive evaluation is needed to determine whether the initial choice of specific ART may increase or decrease HF risk independent of known CVD risk factors. Results may help inform initial ART selection and ongoing care for PWH who have already initiated ART by identifying those at greatest risk for developing HF in the future. The limited existing studies on this topic used standard statistical methods which are generally not adequate to account for confounding and selection bias from time-dependent covariates affected by prior treatment. To address these knowledge gaps, we analyzed data from three large integrated healthcare systems using a causal inference modeling approach to emulate inferences from randomized trials.[17]
Methods
Study Design, Setting, and Data Sources
We conducted a retrospective cohort study of members of KP in Northern and Southern California (KPNC and KPSC, respectively), and KP Mid-Atlantic States (KPMAS; Maryland, Virginia, and Washington, DC), which are integrated healthcare systems providing comprehensive medical services to >10 million members tracked through electronic health records (EHR). Eligible individuals included adult (aged ≥21 years) members with HIV who initiated their first ART regimen between 2000–2016.
Primary data sources include HIV Registries and associated EHR data. The HIV registries maintain current lists of all HIV patients, HIV risk factors, dates of known infection, AIDS diagnoses, complete HIV-related lab and pharmacy data, with manual confirmation of HIV status. The HIV registries included all known cases of HIV since the early 1980s for KPNC, since 2000 for KPSC, and since 2004 for KPMAS. Vital status was assessed comprehensively from member proxy reporting and deaths identified during a hospitalization from EHR and billing claims data, Social Security Administration vital status files and state death certificates.[18]
Study Population
The study population includes adult PWH without prevalent HF who initiated a first-line ART regimen (i.e., PI-based, NNRTI-based, and INSTI-based ART) between 2000–2016, with baseline defined as the date of ART initiation. Patients were excluded if evidence of prevalent HF or if not ART naïve based on the following criteria: (1) any prior ART use that was not first-line; (2) prior HIV RNA<200 copies/ml; (3) other evidence of prior ART use (i.e., first-line ART <6 months after study entry).
Study outcome and Follow-up
The primary study outcome was an incident HF diagnosis identified in the EHR using International Classification of Diseases, Ninth or Tenth Edition (ICD-9/10) diagnosis codes: 398.91, 402.x1, 404.x1, 404.x3, 428.x/I09.81, I11.0, I13.10, I12.0, I50, based on either 1 primary discharge diagnosis of HF, or, at least 3 ambulatory visits with a diagnosis of HF, with at least 1 from a cardiologist. This definition has been shown to have high (>95%) positive predictive value compared with manual chart review[19–23]. Patients were followed from ART initiation until the earliest of an incident HF diagnosis or until censoring at death, health plan disenrollment, or December 31, 2016.
Study exposure
We evaluated HF risk by the three major first-line ART classes during the study period, namely PI-based, NNRTI-based, and INSTI-based ART (i.e., three two-way comparisons), and by the two common nucleotide reverse transcriptase inhibitors (NRTI): tenofovir and abacavir.
Covariates
We ascertained information on risk factors for HF using ICD-9/10 and Current Procedural Terminology (CPT) diagnostic or procedure codes, laboratory results, vital signs or specific therapies received based on validated algorithms and approaches,[19] as well as from regional cancer and diabetes registries. Measures included demographics (age, gender as recorded in administrative records, self-reported race/ethnicity), socioeconomic factors (lower education: census block with more than 25% less than high school education; lower income: census block with median household income <$35,000), cardiovascular risk factors (tobacco use, alcohol dependence, substance dependence, overweight/obesity), cardiovascular and other medical history (acute myocardial infarction, atrial fibrillation, chronic liver disease, chronic renal disease, coronary revascularization, diabetes mellitus, depression, dyslipidemia, hypertension, hyperthyroidism, hypothyroidism, mitral or aortic valvular heart disease, peripheral artery disease, proteinuria, unstable angina), and dispensed outpatient pharmacy data (cardiac-related medications, antihypertensive medications, lipid-lowering therapy, diabetes therapy, non-steroidal anti-inflammatory drugs). All covariates were measured at baseline and time-updated, except for age, gender, census-based income and education, smoking status, KP site, HIV transmission risk factor, years known HIV-infected, and calendar year of ART initiation, which were only baseline measures.
The institutional review board at each participating institution approved this study with waivers of written informed consent.
Statistical methods
Our objective was to emulate intention-to-treat and per-protocol analyses[24] of conceptual trials focused on pairwise head-to-head comparisons of the effects of initial ART on HF risk: (1) PI vs NNRTI (reference); (2) PI vs INSTI (reference); (3) NNRTI vs. INSTI (reference); and (4) tenofovir vs. abacavir (reference). Unlike a randomized trial, treatment assignment in clinical practice is not random. Thus, our primary approach involves a causal inference analysis based on inverse probability weighting (IPW) [25, 26] and marginal structural models (MSM), in order to appropriately account for time-dependent sources of confounding, and also selection bias (i.e., informative right censoring).[25, 26] Specifically, we estimated the propensity to 1) initiate each treatment as a function of baseline characteristics, 2) continue the initial treatment as a function of both baseline and time-varying covariates, and 3) to experience each of the three censoring events, corresponding with death, disenrollment, and administrative end of study as a function of baseline and time-updated variables. Treatment and censoring propensity scores were estimated using super learning,[27] a machine learning algorithm that avoids reliance on a priori specified models for confounding and selection bias.[28] Super learning[27] is a data-adaptive estimation algorithm that combines predicted values from a library of various candidate estimators (i.e., learners) through a weighted average.
Crude and adjusted HRs were defined by working[29] logistic MSMs for discrete-time counterfactual hazards that emulate standard Cox Proportional Hazard models. Inverse probability weights were stabilized and truncated following recommendations in the causal inference field.[30] Here, IP weights were truncated at value 20 which was always above the 99th percentile of the IP weights in each analysis, thus demonstrating minor practical violations of the positivity assumption. HRs were estimated through the minimum of the third quartile of follow-up in each pairwise comparison, since precise effect estimation is not possible after this point. We used the missing indicator approach[31] for 8 covariates with missing data, which is a recommended approach to account for missingness in the causal inference field.[31–33] The method relies on the assumption that unmeasured covariates do not result in unmeasured confounding, since information unknown to clinicians cannot directly influence their treatment decisions. Here, covariates such as CD4 were missing in the EHR, and therefore not likely known to the treating providers or patients.
For each contrast, we present HRs for 5 different analyses (presented here for PI-based or NNRTI-based ART, but similar models were fit for all pairwise comparisons).
Crude intent-to-treat (ITT). This analysis evaluates HF risk from ART initiation until censoring, even if a patient switches regimens or discontinues all therapy after baseline and does not include inverse probability weights for confounding and selection bias adjustment.
Adjusted ITT. Same as crude ITT, but with incorporation of inverse probability weights.
Adjusted per-protocol (PP). Here we emulate a randomized trial where patients initiate and continuously adhere to PI-based or NNRTI-based ART. HF risk is assessed from ART initiation until censoring or evidence of ART switches or interruptions.
Adjusted shortened ITT. Next we evaluated the HR in ITT analyses where follow-up time was truncated to match that of the PP analyses. This was needed to directly compare HR estimates in PP and ITT analyses, because the definition of the HR estimand from working MSM is a function of the maximum follow-up time when the proportionality assumption does not hold.[34]
Adjusted hybrid PP/ITT. To maximize follow-up, while ensuring at least 1 year exposure to PI-based or NNRTI-based ART, we emulated a randomized trial that aimed to study legacy treatment effects. In this approach, for patients who switch or discontinue therapy during the first 1 year of ART, their follow-up will be censored at the time of the change. However, any change in treatment after 1 year is ignored as in ITT analyses.
Finally, survival curves were estimated for each of the adjusted ITT analysis for each of the 4 pairwise comparisons by fitting saturated logistic MSMs which included interaction terms between time and the exposure[35], and thus do not rely on parametric assumptions such as constant HR over time (i.e., proportionality of hazards assumption). P-values were computed for tests that the differences between the areas under the two survival curves compared on each plot are 0.
Results
A total of 39,000 PWH aged ≥21 years were initially identified (Figure 1). Of these, we made the following exclusions: did not initiate first-line ART (n=7,453); evidence of prior ART (n=7,791); prior HIV RNA<200 copies/ml (n=7,901); first ART <6 months after study entry (n=2,142); prevalent HF (n=79). The final study population included 13,634 PWH.
Figure 1. Flow diagram of study exclusions for final study population of people with HIV who initiated a first-line antiretroviral therapy regimen between 2000 and 2016.
Depicts study exclusions for participants from full source population to obtain the final study population. The full source population includes all adults with HIV during 2000 to 2016 from Kaiser Permanente Northern California (KPNC), Kaiser Permanente Southern California (KPSC), and Kaiser Permanente Mid-Atlantic States (KPMAS). The final study population includes all adults with HIV with a first-line antiretroviral therapy (ART) regimen among people previously naïve to ART. Study exclusions include (1) no evidence of first-line ART use during follow-up; (2) evidence that participant was not ART naïve at first-line ART initiation; and (3) prevalent heart failure at ART initiation.
PWH were predominantly men (88%), with a median age at initiation of ART of 40 years (Table 1). PWH were 34% non-Hispanic white, 24% non-Hispanic black, 24% Hispanic, 5% Asian/Pacific Islander, and 13% other/unknown race/ethnicities. Fifty-two percent were men who have sex with men, 5% had a history of injection drug use, and the remainder had HIV risk factors of heterosexual sex (18%), other (1%) or unknown status (23%). PWH had a median CD4 at study entry of 263 cells/μl (IQR 119, 412), median log10 HIV RNA levels of 4.7 (IQR 4.2, 5.2) and 15% had a prior diagnosis of clinical AIDS. Differences in baseline characteristics including cardiovascular and other medical history and medications by ART categories are also shown in Table 1.
Table 1.
Baseline characteristics of adults with HIV initiating ART
| Antiretroviral therapy class | NRTI exposure | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| All PWH N=13,634 |
PI N=4108 |
NNRTI N=6648 |
INSTI N=2350 |
Tenofovir N=9455 |
Abacavir N=1089 |
|||||||
| Sociodemographics | ||||||||||||
| Age (years), Median (IQR) | 41 | (33, 48) | 41 | (34,48) | 41 | (33,48) | 40 | (30,50) | 41 | (32,49) | 42 | (34,51) |
| Men, N (%) | 11,982 | (87.9) | 3437 | (83.7) | 6011 | (90.4) | 2057 | (87.5) | 8369 | (88.5) | 945 | (86.8) |
| Race/ethnicity | ||||||||||||
| Non-Hispanic White | 4,693 | (34.4) | 1366 | (33.3) | 2383 | (35.9) | 744 | (31.7) | 3219 | (34.1) | 366 | (33.6) |
| Non-Hispanic Black | 3,259 | (23.9) | 1008 | (24.5) | 1500 | (22.6) | 661 | (28.1) | 2381 | (25.2) | 286 | (26.3) |
| Hispanic | 3,308 | (24.3) | 979 | (23.8) | 1595 | (24) | 603 | (25.7) | 2358 | (24.9) | 242 | (22.2) |
| Asian or Pacific Islander | 650 | (4.8) | 162 | (3.9) | 314 | (4.7) | 148 | (6.3) | 498 | (5.3) | 41 | (3.8) |
| Other | 191 | (1.4) | 53 | (1.3) | 86 | (1.3) | 47 | (2.0) | 143 | (1.5) | 15 | (1.4) |
| Unknown | 1533 | (11.2) | 540 | (13.2) | 770 | (11.6) | 147 | (6.3) | 856 | (9.1) | 139 | (12.8) |
| Health care system, N (%) | ||||||||||||
| KPMAS | 2,134 | (15.7) | 719 | (17.5) | 886 | (13.3) | 486 | (20.7) | 1703 | (18.0) | 202 | (18.6) |
| KPNC | 4,760 | (34.9) | 1427 | (34.7) | 2341 | (35.2) | 794 | (33.8) | 3127 | (33.1) | 414 | (38.0) |
| KPSC | 6,740 | (49.4) | 1962 | (47.8) | 3421 | (51.5) | 1070 | (45.5) | 4625 | (48.9) | 473 | (43.4) |
| Low census education, N (%) | 2,955 | (27.0) | 842 | 26.8 | 1455 | 27.2 | 554 | 26.4 | 2100 | 22.2 | 225 | 20.7 |
| (2,682 missing) | (969 missing) | (1291 missing) | (251 missing) | (1542 missing) | (224 missing) | |||||||
| Low census income, N (%) | 1,638 | (15.0) | 489 | 15.6 | 771 | 14.4 | 315 | 15.0 | 1139 | 12.1 | 148 | 13.6 |
| (2,690 missing) | (972 missing) | (1295 missing) | (252 missing) | (1546 missing) | (224 missing) | |||||||
| HIV characteristics | ||||||||||||
| HIV risk factor, N (%) | ||||||||||||
| Men who have sex with men | 7,041 | (51.6) | 2001 | (48.7) | 3806 | (57.3) | 938 | (39.9) | 5020 | (53.1) | 486 | (44.6) |
| Injection drug use | 734 | (5.4) | 266 | (6.5) | 320 | (4.8) | 124 | (5.3) | 507 | (5.4) | 68 | (6.2) |
| Heterosexual | 2447 | (18.0) | 875 | (21.3) | 1165 | (17.5) | 335 | (14.3) | 1685 | (17.8) | 168 | (15.4) |
| Other | 145 | (1.1) | 65 | (1.6) | 69 | (1.0) | 5 | (0.2) | 86 | (0.9) | 16 | (1.5) |
| Unknown | 3267 | (23.4) | 901 | (21.9) | 1288 | (19.4) | 948 | (40.3) | 2157 | (22.8) | 351 | (32.2) |
| ART start year, median (IQR) | 2009 | (05, 13) | 2007 | (04, 11) | 2008 | (04, 12) | 2015 | (13, 16) | 2011 | (08, 14) | 2009 | (03, 15) |
| Years known HIV, Median (IQR) | 0.2 | (0.1, 1.9) | 0.2 | (0.1, 2.7) | 0.2 | (0.1, 1.9) | 0.1 | (0.0, 0.4) | 0.2 | (0.1, 1.7) | 0.2 | (0.1, 2.1) |
| CD4 cells/μl | 264 | (120, 415) | 189 | (59, 336) | 279 | (159, 416) | 354 | (187, 525) | 283 | (140, 430) | 281 | (143, 454) |
| (1,864 missing) | (579 missing) | (838 missing) | (355 missing) | (1212 missing) | (218 missing) | |||||||
| Nadir CD4 cells/μl, median (IQR) | 250 | (112, 390) | 176 | (54, 315) | 263 | (148, 387) | 342 | (182, 504) | 267 | (132, 407) | 268 | (132, 428) |
| (1,804 missing) | (562 missing) | (818 missing | (337 missing) | (1173 missing) | (209 missing) | |||||||
| HIV RNA log10 cp/ml, median (IQR) | 4.7 | (4.1, 5.1) | 4.8 | (4.2, 5.2) | 4.6 | (4.1, 5.0) | 4.7 | (4.2, 5.2) | 4.7 | (4.1, 5.1) | 4.6 | (4.1, 5.1) |
| (2,571 missing) | (930 missing) | (1118 missing) | (392 missing) | (1585 missing) | (269 missing) | |||||||
| CDC AIDS, N (%) | 2,079 | (15.3) | 867 | (21.1) | 887 | (13.3) | 245 | (10.4) | 1316 | (13.9) | 147 | (13.5) |
| Cardiovascular risk factors, N (%) | ||||||||||||
| Ever smoked | 6,009 | (44.1) | 1847 | (45.0) | 2870 | (43.2) | 1077 | (45.8) | 4368 | (46.2) | 484 | (44.4) |
| Ever alcohol dependence | 1,609 | (11.8) | 557 | (13.6) | 768 | (11.6) | 218 | (9.3) | 1069 | (11.3) | 114 | (10.5) |
| Ever substance dependence | 2,218 | (16.3) | 770 | (18.7) | 1029 | (15.5) | 328 | (14) | 1532 | (16.2) | 168 | (15.4) |
| Overweight/obese | 4360 | 53.1 | 1052 | 53.0 | 2104 | 53.4 | 1150 | 53.2 | 3857 | 52.5 | 310 | 57.8 |
| (5421 missing) | (2122 missing) | (2706 missing) | (190 missing) | (2108 missing) | (553 missing) | |||||||
| Cardiovascular and other medical history, N (%) | ||||||||||||
| Acute myocardial infarction | 34 | (0.3) | 14 | (0.3) | 16 | (0.2) | 3 | (0.1) | 18 | (0.2) | 5 | (0.5) |
| Atrial fibrillation | 50 | (0.4) | 18 | (0.4) | 23 | (0.4) | 9 | (0.4) | 40 | (0.4) | 3 | (0.3) |
| Chronic liver disease | 958 | (7.0) | 323 | (7.9) | 454 | (6.8) | 142 | (6.0) | 701 | (7.4) | 83 | (7.6) |
| Chronic renal disease* | 265 | (1.9) | 94 | (2.3) | 89 | (1.3) | 74 | (3.2) | 135 | (1.4) | 63 | (5.8) |
| Coronary revascularization | 13 | (0.1) | 3 | (0.1) | 7 | (0.1) | 2 | (0.1) | 7 | (0.1) | 3 | (0.3) |
| Diabetes mellitus | 615 | (4.5) | 189 | (4.6) | 271 | (4.1) | 132 | (5.6) | 425 | (4.5) | 67 | (6.2) |
| Diagnosed depression | 1,940 | (14.2) | 641 | (15.6) | 888 | (13.4) | 342 | (14.6) | 1464 | (15.5) | 147 | (13.5) |
| Dyslipidemia | 1,896 | (13.9) | 503 | (12.2) | 938 | (14.1) | 386 | (16.4) | 1435 | (15.2) | 182 | (16.7) |
| Hypertension | 1,821 | (13.4) | 534 | (13.0) | 890 | (13.4) | 337 | (14.3) | 1344 | (14.2) | 184 | (16.9) |
| Hyperthyroidism | 67 | (0.5) | 21 | (0.5) | 32 | (0.5) | 12 | (0.5) | 49 | (0.5) | 7 | (0.6) |
| Hypothyroidism | 237 | (1.7) | 67 | (1.6) | 119 | (1.8) | 41 | (1.7) | 173 | (1.8) | 19 | (1.7) |
| Mitral or aortic valvular disease | 61 | (0.5) | 18 | (0.4) | 33 | (0.5) | 9 | (0.4) | 49 | (0.5) | 7 | (0.6) |
| Peripheral artery disease | 20 | (0.2) | 5 | (0.1) | 10 | (0.2) | 5 | (0.2) | 15 | (0.2) | 1 | (0.1) |
| Proteinuria | 1,483 | (10.9) | 464 | (11.3) | 626 | (9.4) | 361 | (15.4) | 1158 | (12.3) | 150 | (13.8) |
| Unstable angina | 9 | (0.1) | 2 | (0.1) | 4 | (0.1) | 2 | (0.1) | 5 | (0.1) | 1 | (0.1) |
| Medication history | ||||||||||||
| ACE inhibitor | 784 | (5.8) | 221 | (5.4) | 404 | (6.1) | 140 | (6.0) | 582 | (6.2) | 79 | (7.3) |
| Aldosterone receptor antagonist | 32 | (0.2) | 10 | (0.2) | 14 | (0.2) | 5 | (0.2) | 20 | (0.02) | 7 | (0.6) |
| Alpha blocker | 154 | (1.1) | 50 | (1.2) | 65 | (1) | 28 | (1.2) | 93 | (1.0) | 18 | (1.7) |
| Angiotensin II receptor blocker | 125 | (0.9) | 20 | (0.5) | 61 | (0.9) | 39 | (1.7) | 101 | (1.2) | 12 | (1.1) |
| Anticoagulant | 82 | (0.6) | 26 | (0.6) | 44 | (0.6) | 12 | (0.5) | 62 | (0.7) | 6 | (0.6) |
| Beta blocker | 531 | (3.9) | 148 | (3.6) | 264 | (4) | 94 | (4.0) | 360 | (3.8) | 60 | (5.5) |
| Calcium channel blocker | 352 | (2.6) | 96 | (2.3) | 168 | (2.5) | 80 | (3.4) | 237 | (2.5) | 52 | (4.8) |
| Diabetic therapy | 438 | (3.2) | 126 | (3.1) | 189 | (2.8) | 107 | (4.6) | 309 | (3.3) | 44 | (4.0) |
| Diuretic | 479 | (3.5) | 142 | (3.5) | 259 | (3.9) | 63 | (2.7) | 323 | (3.4) | 51 | (4.7) |
| Non-steroidal anti-inflammatory drug | 1,494 | (11.0) | 464 | (11.3) | 739 | (11.1) | 250 | (10.6) | 1,049 | (11.1) | 93 | (8.5) |
| Other lipid-lowering agent | 104 | (0.8) | 28 | (0.7) | 57 | (0.9) | 14 | (0.6) | 73 | (0.8) | 7 | (0.6) |
| Statin | 541 | (4.0) | 142 | (3.5) | 265 | (4) | 121 | (5.2) | 422 | (4.5) | 51 | (4.7) |
ART, antiretroviral therapy; IQR, interquartile range; PI, protease inhibitor; (N)NRTI, (non-) nucleoside reverse transcriptase inhibitors; INSTI, integrase strand transfer inhibitor
estimated glomerular filtration rate below 60 mL/min/1.73m2
A total of 4,108 PWH started a PI-based ART regimen, 6,648 started an NNRTI-based regimen and 2,350 started an INSTI-based regimen (Table 2). In addition, 9,455 PWH initiated ART with TDF and 1,089 initiated ART with abacavir. Across regimens, the most common reasons for censoring included disenrollment from the respective health plan (ranged from 47% to 57%) and administrative end of study (ranged from 35% to a high of 67% for INSTI-based regimens). The outcome of HF was uncommon (1% or lower across groups). Overall, HF with reduced ejection fraction was more common than HF with preserved ejection fraction, but there were differences by ART categories (Table 2).
Table 2.
Follow-up and outcomes by ART use at baseline
| Antiretroviral therapy class | NRTI exposure | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| PI | NNRTI | INSTI | Tenofovir | Abacavir | ||||||
| N | 4,108 | 6,648 | 2,350 | 9,455 | 1,089 | |||||
| Person-years | 19,285 | 31,228 | 4,020 | 34,294 | 3,945 | |||||
| Censored, N (%) | ||||||||||
| Disenroll | 2,352 | (57%) | 3,657 | (55%) | 742 | (32%) | 4,581 | (48%) | 512 | (47%) |
| End of study | 1,442 | (35%) | 2,698 | (41%) | 1,576 | (67%) | 4,536 | (48%) | 508 | (47%) |
| Death | 259 | (6%) | 256 | (4%) | 25 | (1%) | 285 | (3%) | 63 | (6%) |
| Switch/discontinue regimen, N (%) | 2,876 | (71%) | 3,630 | (56%) | 743 | (31%) | 4,305 | (45%) | 598 | (55%) |
| HF cases, N | 55 | 37 | 7 | 53 | 6 | |||||
| HF Subtype, N (%) | ||||||||||
| HFpEF | 15 | (27%) | 7 | (19%) | 2 | (29%) | 15 | (28%) | 3 | (50%) |
| HFrEF | 13 | (13%) | 14 | (38%) | 1 | (14%) | 12 | (23%) | 2 | (33%) |
| HFmrEF | 11 | (20%) | 4 | (11%) | 1 | (14%) | 13 | (25%) | 0 | (0%) |
| Unknown EF | 16 | (29%) | 12 | (32%) | 3 | (43%) | 13 | (25%) | 1 | (17%) |
Table 3 presents HRs by ART class and by NRTI use. In adjusted ITT analysis, PI-based compared with NNRTI-based ART was associated with a higher risk of HF over 86 months of follow-up, with an HR of 2.5 (95% CI 1.5–4.3). The PP analysis also indicated a higher risk of HF with PI-based compared with NNRTI-based ART with an adjusted HR of 4.9 (95% CI 1.2–20.4) over 25 months of follow-up. Inferences were unchanged and the magnitude of association was similar for ITT with follow-up to 25 months and for the ITT/PP hybrid analysis. There was no evidence of a difference in HF risk comparing PI-based with INSTI-based ART in both ITT analyses with an HR of 0.5 (95% CI 0.2–1.8) over 28 months and PP analyses with an HR of 0.5 (95% CI 0.2–1.8) over 20 months; similar results were seen for ITT over 20 months and for the ITT/PP hybrid analysis. There was evidence of lower risk of HF for NNRTI-based compared with INSTI-based ART in both ITT analyses with an HR of 0.1 (95% CI 0.0–0.8) over 28 months and PP analyses with an HR of 0.1 (95% CI 0.0–0.7) of 20 months; similar results were seen for ITT over 20 months and for the ITT/PP hybrid analysis. No evidence of differences in HF risk were observed between those with tenofovir versus abacavir exposure (reference), with an HR of 1.7 (95% CI 0.5–5.7) for ITT over 63 months and an HR of 0.9 (95% CI 0.2–3.7) for PP over 17 months.
Table 3.
Hazard ratios (95% confidence intervals) for risk of heart failure by ART regimen
| Antiretroviral therapy class | NRTI exposure | |||||||
|---|---|---|---|---|---|---|---|---|
| PI vs NNRTI (ref) | PI vs INSTI (ref) | NNRTI vs. INSTI (ref) | Tenofovir vs. Abacavir (ref) | |||||
| Follow-up, months1 | Hazard Ratio (95% CI) | Follow-up, months1 | Hazard Ratio (95% CI) | Follow-up, months1 | Hazard Ratio (95% CI) | Follow-up, months1 | Hazard Ratio (95% CI) | |
| Unadjusted2 results | ||||||||
| ITT | 86 | 3.1 (1.9–5.1) | 28 | 1.1 (0.5–2.7) | 28 | 0.3 (0.1–0.8) | 63 | 0.8 (0.3–2.3) |
| Adjusted2 results | ||||||||
| ITT | 86 | 2.5 (1.5–4.3) | 28 | 0.5 (0.2–1.8) | 28 | 0.1 (0.0–0.8) | 63 | 1.7 (0.5–5.7) |
| PP | 25 | 4.9 (1.2–20.4) | 20 | 0.4 (0.1–1.8) | 20 | 0.1 (0.0–0.7) | 17 | 0.9 (0.2–3.7) |
| Shortened3 ITT | 25 | 3.3 (1.3–8.2) | 20 | 0.5 (0.1–1.8) | 20 | 0.1 (0.0–0.8) | 17 | 1.3 (0.3–5.4) |
| Hybrid PP/ITT4 | 53 | 5.0 (1.7–14.5) | 22 | 0.4 (0.1–1.8) | 22 | 0.1 (0.0–0.7) | 26 | 0.9 (0.2–3.7) |
ITT, Intention-to-treat; PP, per protocol
Maximum months post baseline where hazard functions were assessed
Unadjusted results from logistic marginal structural models that do not include inverse probability weights for confounding and selection bias adjustment. Adjusted results incorporate weights for confounding and selection bias adjustment with weights informed by covariates listed in Table 1.
Follow-up time was truncated to match that of the PP analyses
PP for one year (i.e., remain on treatment for 1st year) and ITT after 1 year
Survival curves (Figure 2) based on adjusted ITT analyses indicated evidence of a higher cumulative risk of HF for PI versus NNRTI-based ART over 86 months (Figure 2a; 1.8% vs. 0.8%; p=0.002). No evidence of different cumulative risks were noted for other pair-wise comparisons of survival curves including INSTI-based versus PI-based ART over 28 months (Figure 2b; 0.8% vs. 0.6%; p=0.42), INSTI-based ART vs. NNRTI-based ART over 28 months (Figure 2c; 0.9% vs. 0.1%; p=0.26), and tenofovir versus abacavir over 63 months (Figure 2d; 0.6% vs. 0.3%; p=0.42).
Figure 2. Cumulative incidence of heart failure by ART regimens.
Cumulative incidence curves for heart failure for each of two regimens in a pair for the time scale of months since antiretroviral therapy (ART) initiation. The cumulative incidence for each regimen at the end of follow-up is displayed at the end of each curve, with the regimen with the highest cumulative incidence depicted with a red dashed curve. The pairwise comparisons displayed included: (a) PI (dashed red) vs. NNRTI (solid black); (b) INSTI (dashed red) vs. PI (solid black); (c) INSTI (dashed red) vs. NNRTI (solid black); and (d) tenofovir (dashed red) vs. abacavir (solid black). Cumulative incidence curves were mapped from adjusted estimates of the hazards from intention-to-treat analysis. P-values on plot are for tests that the differences between the areas under the two curves are 0.
Discussion
In this large cohort study of PWH from three large integrated healthcare delivery systems, we observed a 2.5-fold elevated risk of HF for PI-based ART compared with NNRTI-based ART over 7 years after ART initiation, and no evidence of differences comparing regimens containing tenofovir or abacavir over 5 years after ART initiation. With more limited follow-up of 2 years after ART initiation we noted evidence of a higher risk of HF for INSTIs compared with NNRTIs that was not consistent across analyses, and no evidence of a difference in risk of HF comparing INSTIs and PIs. Given the inconsistent results for INSTIs and the limited follow-up, further research is needed. Of note, despite ART regimen associations observed, the outcome was rare resulting in absolute differences in HF risk of 1% or less between ART categories over follow-up.
The higher risk of HF in HIV patients is well established based on data from KP,[5, 6] a large Commercial and Medicare database study,[36] the Veteran’s Administration,[4, 37] Taiwanese cohort studies,[38, 39], and a university-based cohort.[40] Studies have noted that PWH treated with ART are less likely to get HF than PWH not yet on ART[39], and that lower HF rates are observed for those with longer ART durations[38] or those with better ART treatment responses (i.e., higher CD4 and lower HIV RNA levels).[37, 40] However, an emerging clinical question is whether certain ART regimens can minimize the risk of adverse events such as HF without compromising HIV viral suppression or immune reconstitution.
Limited research exists about whether specific ART influences HF risk in PWH, which could inform ART treatment choices for at-risk PWH. It was recognized early in the HIV epidemic that certain ART medications may be associated with cardiomyopathy as early generation PIs were associated with dyslipidemia and older NRTIs (i.e., stavudine, zidovudine) induced mitochondrial damage.[41] Risks of CVD also varied by use of PIs and certain NRTIs, including abacavir.[7, 11, 12] However, conflicting data exist about ART regimens or individual therapies and HF risk.[16, 38, 42, 43] A large Taiwanese study observed no associations of the major ART classes with HF, but there were few HF events and limited follow-up in those treated with INSTIs.[38] Among 394 PWH with HF in a tertiary care hospital in New York, reduced systolic function and HF readmissions were increased in PWH receiving PI-based ART compared with other ART.[42] Regarding NRTIs, among 10,931 ART-naïve patients in a Veteran’s Administration (VA) study, PWH who initiated tenofovir (adjusted HR 1.82) but not abacavir (adjusted HR 1.45), had a significantly increased risk of HF compared with PWH initiating other ART regimens.[16] In contrast, another VA study of 21,435 PWH reported that current tenofovir use was associated with a 32% lower risk of HF compared with never users, with no association with prior tenofovir use.[43] Furthermore, current tenofovir use, but not abacavir use, was associated with a lower risk of HF when modeled simultaneously.
Our study is the first to date to evaluate the three major ART classes on HF risk, or risk comparing two commonly used NRTIs which have been implicated previously with HF.[16, 43] In the absence of randomized trial data, observational studies evaluating the safety of therapies for adverse effects requires careful control of confounding and selection bias. Here we employed a causal inference framework to emulate a randomized trial to evaluate whether the risk of HF varies by initial ART among PWH. While one study did employ similar methods,[43] the exposures reported were current, past and never ART use, and did not map directly onto treatment arms of a trial for evaluating specific drug classes. Finally, our study evaluated INSTI use, currently the preferred initial ART regimen, with a larger number of treated patients and HF outcomes than previous studies.[38]
Given the established high risks of certain ART such as PIs and abacavir for MIs and other CVD outcomes, it is important to understand which therapies may also increase HF risk, in order to guide ongoing ART treatment in high risk PWH. Our study noted a higher risk of HF for those treated with PIs compared with NNRTIs. Several mechanisms may contribute to the higher HF risk in PWH on ART including myocardial fibrosis, steatosis, hypertension, systemic immune activation and metabolic dysregulation.[44] PIs may contribute specifically to the higher risk by increasing dyslipidemia and prelamin A accumulation.[41] Some studies have indicated PIs such as ritonavir may stimulate platelet production for proinflammatory markers, with mixed results for other ART (e.g., INSTIs and abacavir).[45] We also observed inconsistent results for a higher risk of HF for INSTIs compared with NNRTIs, which requires further investigation in other settings with longer follow-up, especially in light of research indicating that INSTIs may lead to weight gain and other cardiometabolic effects.[45, 46]
Our study had notable strengths, including the large sample size of PWH initiating ART, accurate ascertainment of a large number of clinical risk factors for HF and implementation of a causal framework. There are several study limitations. First, causal interpretation relies on several strong assumptions, including no unmeasured confounding and sources of selection bias, although we comprehensively considered known CVD risk factors. In addition, our inferences rely on the assumption of consistent estimation of treatment and right-censoring mechanisms. To address this, we used a machine learning algorithm (i.e., super learning[27]) that avoids reliance on a priori specified models for addressing confounding and selection bias.[28] Additionally, our HF case definition was based on an EHR-based algorithm relying on diagnostic codes, and it remains possible that our definition may lead to under-detection. However we anticipate the misclassification by ART is non-differential. There was also a high rate of non-adherence and switching of regimens. However, the consistency of results comparing ITT and PP analyses (which accounted for non-adherence) indicate this had minimal impact. An additional limitation is the low prevalence of INSTI use, which was primarily raltegravir during follow-up, which likely contributed to the inconsistent results. Thus results should be interpreted with caution, and requires investigation in other studies. Results for other ART regimens evaluated remain clinically relevant given many current PWH have remained on PIs and other regimens long-term. Finally, there may have been limited generalizability to women given 88% of the population were men.
In summary, we noted a higher risk of HF for PWH initiating PIs compared with NNRTIs, no difference between tenofovir-based and abacavir-based ART and inconsistent results for comparisons with INSTIs. These results may have important clinical implications about initial regimen choice for PWH regarding their future risk of HF but require confirmation with studies with extended follow-up for PWH initiating INSTIs. Nevertheless, the associations observed were marginal and only resulted in small absolute differences in HF risk between ART categories. Thus modification of other traditional cardiovascular risk factors, such as smoking, obesity, diabetes or hypertension will likely have greater impact on HF risk in PWH.
ACKNOWLEDGEMENTS
This work was s upported by the National Heart, Lung, and Blood Institute (R01HL132640, PI: MJS, ASG). The authors confirm contribution to the paper as follows: study conception and design: MJS, ASG; secured funding: MJS, ASG; biostatistical guidance: RN; data collection: WAL, TKL; data analysis: RN, WAL, NP; draft manuscript: MJS. All authors contributed to review and interpretation of results, critical revisions of draft manuscript, and approval of the final version of the manuscript.
Conflicts of Interest and Source of Funding:
This work was supported by the National Heart, Lung, and Blood Institute (NHLBI, R01HL132640, PI: MJS, ASG). APA is supported by a Mentored Patient-Oriented Research Career Development Award (K23HL150159) through the NHLBI. APA reports research grants not directly related to this study paid to his institution from Abbott, Amarin Pharma, Edwards Lifesciences, Esperion, Lexicon, and Novartis. WJT reports research grants not directly related to this study paid to his institution from Pfizer, Merck, Janssen, ViiV, GSK, and Moderna. No conflicts of interest are reported by other authors.
REFERENCES
- 1.Virani SS, Alonso A, Aparicio HJ, Benjamin EJ, Bittencourt MS, Callaway CW, et al. , American Heart Association Council on Epidemiology and Prevention Statistics Committee and Stroke Statistics Subcommittee. Heart Disease and Stroke Statistics-2021 Update: A Report From the American Heart Association. Circulation 2021; 143(8):e254–e743. [DOI] [PubMed] [Google Scholar]
- 2.Freiberg MS, Chang CC, Kuller LH, Skanderson M, Lowy E, Kraemer KL, et al. HIV Infection and the Risk of Acute Myocardial Infarction. JAMA Intern Med 2013; 173(8):614–622. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Silverberg MJ, Leyden WA, Xu L, Horberg MA, Chao CR, Towner WJ, et al. Immunodeficiency and risk of myocardial infarction among HIV-positive individuals with access to care. J Acquir Immune Defic Syndr 2014; 65(2):160–166. [DOI] [PubMed] [Google Scholar]
- 4.Butt AA, Chang CC, Kuller L, Goetz MB, Leaf D, Rimland D, et al. Risk of heart failure with human immunodeficiency virus in the absence of prior diagnosis of coronary heart disease. Arch Intern Med 2011; 171(8):737–743. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Go AS, Reynolds K, Avula HR, Towner WJ, Hechter RC, Horberg MA, et al. Human Immunodeficiency Virus Infection and Variation in Heart Failure Risk by Age, Sex, and Ethnicity: The HIV HEART Study. Mayo Clin Proc 2022; 97(3):465–479. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Lam JO, Leyden WA, Leong TK, Horberg MA, Reynolds K, Ambrosy AP, et al. Variation in Heart Failure Risk by HIV Severity and Sex in People With HIV Infection. J Acquir Immune Defic Syndr 2022; 91(2):175–181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Friis-Moller N, Reiss P, Sabin CA, Weber R, Monforte A, El-Sadr W, et al. Class of antiretroviral drugs and the risk of myocardial infarction. N Engl J Med 2007; 356(17):1723–1735. [DOI] [PubMed] [Google Scholar]
- 8.Durand M, Sheehy O, Baril JG, Lelorier J, Tremblay CL. Association between HIV infection, antiretroviral therapy, and risk of acute myocardial infarction: a cohort and nested case-control study using Quebec’s public health insurance database. J Acquir Immune Defic Syndr 2011; 57(3):245–253. [DOI] [PubMed] [Google Scholar]
- 9.Neesgaard B, Greenberg L, Miro JM, Grabmeier-Pfistershammer K, Wandeler G, Smith C, et al. Associations between integrase strand-transfer inhibitors and cardiovascular disease in people living with HIV: a multicentre prospective study from the RESPOND cohort consortium. Lancet HIV 2022; 9(7):e474–e485. [DOI] [PubMed] [Google Scholar]
- 10.Surial B, Chammartin F, Damas J, Calmy A, Haerry D, Stockle M, et al. Impact of Integrase Inhibitors on Cardiovascular Disease Events in People With Human Immunodeficiency Virus Starting Antiretroviral Therapy. Clin Infect Dis 2023; 77(5):729–737. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Sabin CA, Worm SW, Weber R, Reiss P, El-Sadr W, Dabis F, 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(9622):1417–1426. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Marcus JL, Neugebauer RS, Leyden WA, Chao CR, Xu L, Quesenberry CP Jr., et al. Use of abacavir and risk of cardiovascular disease among HIV-infected individuals. J Acquir Immune Defic Syndr 2015. [Epub ahead of print]. [DOI] [PubMed] [Google Scholar]
- 13.Ding X, Andraca-Carrera E, Cooper C, Miele P, Kornegay C, Soukup M, et al. No association of abacavir use with myocardial infarction: findings of an FDA meta-analysis. J Acquir Immune Defic Syndr 2012; 61(4):441–447. [DOI] [PubMed] [Google Scholar]
- 14.Cruciani M, Zanichelli V, Serpelloni G, Bosco O, Malena M, Mazzi R, et al. Abacavir use and cardiovascular disease events: a meta-analysis of published and unpublished data. Aids 2011; 25(16):1993–2004. [DOI] [PubMed] [Google Scholar]
- 15.Lundgren JD, Neuhaus J, Babiker A, Cooper D, El-Sadr W, Emery S, et al. Use of nucleoside reverse transcriptase inhibitors and risk of myocardial infarction in HIV-infected patients. Aids 2008; 22(14):F17–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Choi AI, Vittinghoff E, Deeks SG, Weekley CC, Li Y, Shlipak MG. Cardiovascular risks associated with abacavir and tenofovir exposure in HIV-infected persons. Aids 2011; 25(10):1289–1298. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hernan MA, Robins JM. Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available. Am J Epidemiol 2016; 183(8):758–764. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Go AS, Magid DJ, Wells B, Sung SH, Cassidy-Bushrow AE, Greenlee RT, et al. The Cardiovascular Research Network: a new paradigm for cardiovascular quality and outcomes research. Circ Cardiovasc Qual Outcomes 2008; 1(2):138–147. [DOI] [PubMed] [Google Scholar]
- 19.Go AS, Lee WY, Yang J, Lo JC, Gurwitz JH. Statin therapy and risks for death and hospitalization in chronic heart failure. JAMA 2006; 296(17):2105–2111. [DOI] [PubMed] [Google Scholar]
- 20.McManus DD, Hsu G, Sung SH, Saczynski JS, Smith DH, Magid DJ, et al. Atrial fibrillation and outcomes in heart failure with preserved versus reduced left ventricular ejection fraction. J Am Heart Assoc 2013; 2(1):e005694. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Go AS, Hsu CY, Yang J, Tan TC, Zheng S, Ordonez JD, et al. Acute Kidney Injury and Risk of Heart Failure and Atherosclerotic Events. Clin J Am Soc Nephrol 2018; 13(6):833–841. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Gurwitz JH, Magid DJ, Smith DH, Goldberg RJ, McManus DD, Allen LA, et al. Contemporary prevalence and correlates of incident heart failure with preserved ejection fraction. Am J Med 2013; 126(5):393–400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Gurwitz JH, Magid DJ, Smith DH, Tabada GH, Sung SH, Allen LA, et al. Treatment Effectiveness in Heart Failure with Comorbidity: Lung Disease and Kidney Disease. Journal of the American Geriatrics Society 2017; 65(12):2610–2618. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Hernan MA, Hernandez-Diaz S. Beyond the intention-to-treat in comparative effectiveness research. Clin Trials 2012; 9(1):48–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Bodnar LM, Davidian M, Siega-Riz AM, Tsiatis AA. Marginal structural models for analyzing causal effects of time-dependent treatments: an application in perinatal epidemiology. Am J Epidemiol 2004; 159(10):926–934. [DOI] [PubMed] [Google Scholar]
- 26.Hernan MA, Hernandez-Diaz S, Robins JM. A structural approach to selection bias. Epidemiology 2004; 15(5):615–625. [DOI] [PubMed] [Google Scholar]
- 27.van der Laan MJ, Polley EC, Hubbard AE. Super learner. Stat Appl Genet Mol Biol 2007; 6:Article25. [DOI] [PubMed] [Google Scholar]
- 28.Neugebauer R, Fireman B, Roy JA, Raebel MA, Nichols GA, O’Connor PJ. Super learning to hedge against incorrect inference from arbitrary parametric assumptions in marginal structural modeling. J Clin Epidemiol 2013; 66(8 Suppl):S99–109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Neugebauer R, Van der Laan M. Nonparametric Causal Effects based on Marginal Structural Models. Journal of Statistical Planning and Inference 2006; 137(2):419–434. [Google Scholar]
- 30.Petersen ML, Porter KE, Gruber S, Wang Y, van der Laan MJ. Diagnosing and responding to violations in the positivity assumption. Stat Methods Med Res 2012; 21(1):31–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Blake HA, Leyrat C, Mansfield KE, Tomlinson LA, Carpenter J, Williamson EJ. Estimating treatment effects with partially observed covariates using outcome regression with missing indicators. Biom J 2020; 62(2):428–443. [DOI] [PubMed] [Google Scholar]
- 32.Hernan MA, McAdams M, McGrath N, Lanoy E, Costagliola D. Observation plans in longitudinal studies with time-varying treatments. Stat Methods Med Res 2009; 18(1):27–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Stuart EA. Matching methods for causal inference: A review and a look forward. Stat Sci 2010; 25(1):1–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Hernan MA. The hazards of hazard ratios. Epidemiology 2010; 21(1):13–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Neugebauer R, Schroeder EB, Reynolds K, Schmittdiel JA, Loes L, Dyer W, et al. Comparison of Mortality and Major Cardiovascular Events Among Adults With Type 2 Diabetes Using Human vs Analogue Insulins. JAMA Netw Open 2020; 3(1):e1918554. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Alonso A, Barnes AE, Guest JL, Shah A, Shao IY, Marconi V. HIV Infection and Incidence of Cardiovascular Diseases: An Analysis of a Large Healthcare Database. J Am Heart Assoc 2019; 8(14):e012241. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Freiberg MS, Chang CH, Skanderson M, Patterson OV, DuVall SL, Brandt CA, et al. Association Between HIV Infection and the Risk of Heart Failure With Reduced Ejection Fraction and Preserved Ejection Fraction in the Antiretroviral Therapy Era: Results From the Veterans Aging Cohort Study. JAMA Cardiol 2017; 2(5):536–546. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Yen YF, Ko MC, Yen MY, Hu BS, Wang TH, Chuang PH, et al. Human Immunodeficiency Virus Increases the Risk of Incident Heart Failure. J Acquir Immune Defic Syndr 2019; 80(3):255–263. [DOI] [PubMed] [Google Scholar]
- 39.Lai YJ, Chen YY, Huang HH, Ko MC, Chen CC, Yen YF. Incidence of cardiovascular diseases in a nationwide HIV/AIDS patient cohort in Taiwan from 2000 to 2014. Epidemiol Infect 2018; 146(16):2066–2071. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Feinstein MJ, Steverson AB, Ning H, Pawlowski AE, Schneider D, Ahmad FS, et al. Adjudicated Heart Failure in HIV-Infected and Uninfected Men and Women. J Am Heart Assoc 2018; 7(21):e009985. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Choi H, Dey AK, Sharma G, Bhoite R, Burkholder G, Fedson S, et al. Etiology and pathophysiology of heart failure in people with HIV. Heart Fail Rev 2021; 26(3):497–505. [DOI] [PubMed] [Google Scholar]
- 42.Alvi RM, Neilan AM, Tariq N, Awadalla M, Afshar M, Banerji D, et al. Protease Inhibitors and Cardiovascular Outcomes in Patients With HIV and Heart Failure. J Am Coll Cardiol 2018; 72(5):518–530. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Chen R, Scherzer R, Hsue PY, Jotwani V, Estrella MM, Horberg MA, et al. Association of Tenofovir Use With Risk of Incident Heart Failure in HIV-Infected Patients. J Am Heart Assoc 2017; 6(4):e005387. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Toribio M, Neilan TG, Zanni MV. Heart Failure among People with HIV: Evolving Risks, Mechanisms, and Preventive Considerations. Curr HIV/AIDS Rep 2019; 16(5):371–380. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Hsue PY. Mechanisms of Cardiovascular Disease in the Setting of HIV Infection. Can J Cardiol 2019; 35(3):238–248. [DOI] [PubMed] [Google Scholar]
- 46.Rebeiro PF, Jenkins CA, Bian A, Lake JE, Bourgi K, Moore RD, et al. Risk of Incident Diabetes Mellitus, Weight Gain, and Their Relationships With Integrase Inhibitor-Based Initial Antiretroviral Therapy Among Persons With Human Immunodeficiency Virus in the United States and Canada. Clin Infect Dis 2021; 73(7):e2234–e2242. [DOI] [PMC free article] [PubMed] [Google Scholar]