Abstract
Background:
Prolonged exposure to hepatotoxic antiretroviral therapy (ART) could lead to chronic liver injury, which might increase risk of hepatocellular carcinoma (HCC). We used causal inference methods to evaluate if cumulative exposure to ART with known mechanisms of chronic liver injury is associated with HCC among people with HIV (PWH), by chronic liver disease (CLD) status.
Methods:
We conducted a cohort study of antiretroviral-naïve PWH with and without diagnosed CLD who initiated ART in the Veterans Affairs (VA) between 1999–2023. Exposure to ART with known mechanism of chronic liver injury (mitochondrial toxicity or accumulation of toxic intermediates) was determined monthly via prescription fills. Incident HCC was identified by VA National Cancer Registry and/or recorded diagnoses. Marginal structural models estimated odds ratios (ORs) with 95% confidence intervals (CIs) of HCC associated with cumulative exposure to chronic liver injury-inducing ART compared to ART without chronic liver injury mechanisms, by CLD status.
Results:
Among 11,854 PWH without CLD and 6,771 PWH with CLD, we observed 112 and 378 HCC events, respectively. Among PWH without CLD, ≥4 years of exposure to chronic liver injury-inducing ART was associated with HCC when compared to ≥4 years of use of ART without chronic liver injury mechanisms (OR=4.02 [95% CI, 1.00-16.24]). Among PWH with CLD, ≥4 years of exposure to chronic liver injury-inducing ART was not associated with HCC (OR=1.36; 95% CI, 0.68-2.72).
Conclusion:
Four or more years of ART with a chronic liver injury mechanism was associated with HCC among PWH without CLD, but not those with CLD.
INTRODUCTION
Despite the benefits of antiretroviral therapy (ART), antiretrovirals may be associated with drug toxicity and adverse effects. People with HIV (PWH) can develop liver enzyme elevations following ART initiation,1 and prolonged exposure to hepatotoxic ART could lead to chronic liver injury and promote hepatic fibrosis.2,3 Dideoxynucleoside analogues have been associated with mitochondrial dysfunction and chronic hepatic changes such as noncirrhotic portal hypertension.4,5 Protease inhibitors might contribute to chronic liver injury though production and accumulation of toxic intermediates via cytochrome P450 pathways.6 The chronic hepatic changes that develop due to exposure to these drug classes may increase risk of hepatocellular carcinoma (HCC).7
HCC is a growing cause of cancer mortality among PWH;8 however, studies have not examined whether cumulative use of ART capable of inducing chronic liver injury is associated with HCC. ART-induced chronic liver injury can occur by two main mechanisms: 1) mitochondrial toxicity, or 2) accumulation of toxic intermediates.9 As PWH age, greater cumulative exposure to chronic liver injury-inducing ART could lead to hepatic inflammation and oxidative stress, which might increase risk of HCC. Indeed, oxidative stress activates hepatocarcinogenic pathways that may lead to the development and progression of HCC.10,11 Further, the association between cumulative use of chronic liver injury-inducing ART and HCC may differ for people living with chronic liver disease (CLD). Patients with CLD often have underlying liver inflammation and fibrosis that can impair hepatic enzymatic pathways and hepatic drug metabolism, which might make them more susceptible to the adverse effects of ART-induced chronic liver injury and increased risk of HCC.12–14
To address these knowledge gaps, this study evaluated if cumulative exposure to ART with a known mechanism of chronic liver injury (i.e., mitochondrial toxicity or accumulation of toxic intermediates) is associated with HCC. Since CLD is the leading cause of HCC,15 diagnosed CLD may modify associations between chronic liver injury-inducing ART and HCC. Consequently, we stratified analyses by CLD status. We hypothesized that longer duration of use of chronic liver injury-inducing ART would be associated with HCC and that associations would be stronger among PWH with CLD.
METHODS
Study Design and Data Source
We conducted a cohort study of antiretroviral-naïve PWH who initiated ART in the VA system between October 1, 1999 and March 31, 2023. The VA comprises >1,300 hospitals and community outpatient clinics nationwide and is the largest HIV care provider in the US.16 VA electronic health record data include demographic data, outpatient and hospital diagnoses (recorded by International Classification of Diseases, Ninth Revision [ICD-9] or Tenth Revision [ICD-10], codes), laboratory results, and dispensed medications.16 Date of death is available from the VA Vital Status file. This study was approved by the Institutional Review Boards of the University of Pennsylvania, VA Connecticut Healthcare System, and Yale University.
Study Patients
PWH were included if they had: 1) newly initiated ART, defined as at least two antiretrovirals dispensed from two different classes,17 within the VA between October 1, 1999 and March 31, 2023, 2) HIV RNA measured within 180 days prior to or on the date of initiating ART with no measurement ≤500 copies/mL (to identify new ART initiators),18 and 3) at least 180 days of observation after initiating ART. Follow-up for patients began 180 days after ART initiation (Supplementary Figure 1). The 180 days prior to the start of follow-up represented the baseline period. We defined the baseline period as the first 180 days after initiation of ART because many PWH in the VA enter care at the time of ART initiation,18,19 and we wished to ensure sufficient time to identify prevalent HCC diagnoses. Further, we believed it was biologically unlikely that ART associated with chronic liver injury would result in HCC prior to 180 days of use. We excluded patients with HCC prior to start of follow-up.
We classified PWH with CLD if they had: 1) hepatitis B infection (detectable hepatitis B surface antigen, e antigen, or DNA) or hepatitis C infection (detectable hepatitis C RNA or genotype) infection, or 2) ≥1 hospital or ≥2 outpatient ICD-9/-10 diagnoses of a viral or non-viral CLD (Supplementary Table 1). Patients identified with a single outpatient CLD diagnosis code were excluded to reduce the likelihood of including patients with suspected or rule-out diagnoses.
Follow-up continued until incident HCC, death, loss to HIV care (defined as the last HIV RNA or CD4+ percentage measurement plus 540 days), or September 30, 2023. For PWH without baseline CLD, follow-up was additionally censored at incident CLD diagnosis. Patients were not censored at ART regimen switches, since our time-updated analysis accounted for time on and off ART. We also did not treat ART discontinuation as a censoring event because this might underestimate associations of HCC with cumulative antiretroviral use.
Exposure to Antiretrovirals Associated With Chronic Liver Injury
The primary exposure was use of ART with a known mechanism of chronic liver injury based on the National Institutes of Health LiverTox website (http://livertox.nih.gov).20 Using the mechanisms in Table 1, we classified antiretrovirals into two groups: 1) associated with chronic liver injury (i.e., mitochondrial toxicity or accumulation of toxic intermediates), or 2) no known mechanism of chronic liver injury. Some antiretrovirals lead to acute liver injury via hypersensitivity reaction, but this mechanism does not contribute to chronic liver injury. We therefore classified antiretrovirals associated with acute hypersensitivity reactions within the group of drugs with no known mechanism of chronic liver injury. Exposure to antiretrovirals with a known mechanism of chronic liver injury was determined monthly (1=exposed; 0=unexposed) via prescription fills and days’ supply. We allowed 30 days after the end of the days’ supply for an additional prescription to be filled to continue the current course (as a gap). Patients were classified as exposed for the entire 30-day period when at least one day in that period was covered by the prescription fill. Cumulative use was defined as the total number of months of past use. Antiretrovirals dispensed during the baseline period counted toward cumulative use. Months without ART use did not count towards the cumulative total. Patients receiving ≥2 antiretrovirals associated with chronic liver injury were not classified as having double the time on the drugs, since it is unclear if the effect of dual exposure is additive.
Table 1.
Antiretrovirals associated with a known mechanism of drug-induced liver injury. Mitochondrial toxicity and accumulation of toxic intermediates are mechanisms associated with chronic liver injury.
| Antiretroviral Drug* | Drug Class | Mechanism of Liver Injury | Likelihood Score† |
|---|---|---|---|
| Didanosine | NRTI | Mitochondrial toxicity | A (well known) |
| Stavudine | NRTI | Mitochondrial toxicity | A (well known) |
| Zalcitabine | NRTI | Mitochondrial toxicity | E* (unproven, but suspected) |
| Zidovudine | NRTI | Mitochondrial toxicity | B (highly likely) |
|
| |||
| Atazanavir | PI | Toxic intermediates | D (possible) |
| Darunavir | PI | Toxic intermediates | C (probably) |
| Indinavir | PI | Toxic intermediates | C (probably) |
| Lopinavir | PI | Toxic intermediates | D (possible) |
| Nelfinavir | PI | Toxic intermediates | D (possible) |
| Saquinavir | PI | Toxic intermediates | D (possible) |
| Tipranavir | PI | Toxic intermediates | E* (unproven, but suspected) |
|
| |||
| Abacavir | NRTI | Hypersensitivity | C (probably) |
| Delavirdine | NNRTI | Hypersensitivity | E* (unproven, but suspected) |
| Efavirenz | NNRTI | Hypersensitivity | A (well known) |
| Etravirine | NNRTI | Hypersensitivity | D (possible) |
| Fosamprenavir | PI | Hypersensitivity | D (possible) |
| Maraviroc | CCR5i | Hypersensitivity | D (possible) |
| Nevirapine | NNRTI | Hypersensitivity | A (well known) |
| Rilpivirine | NNRTI | Hypersensitivity | D (possible) |
|
| |||
| Bictegravir | INSTI | No known mechanism | X (unknown) |
| Cabotegravir | INSTI | No known mechanism | E* (unproven, but suspected) |
| Dolutegravir | INSTI | No known mechanism | D (possible) |
| Doravirine | NNRTI | No known mechanism | E* (unproven, but suspected) |
| Elvitegravir | INSTI | No known mechanism | E* (unproven, but suspected) |
| Emtricitabine | NRTI | No known mechanism | E (unlikely) |
| Enfuvirtide | FI | No known mechanism | E (unlikely) |
| Fostemsavir | AI | No known mechanism | E* (unproven, but suspected) |
| Ibalizumab | PAI | No known mechanism | E (unlikely) |
| Lamivudine | NRTI | No known mechanism | E (unlikely) |
| Lenacapavir | CI | No known mechanism | X (unknown) |
| Raltegravir | INSTI | No known mechanism | D (possible) |
| Tenofovir‡ | NRTI | No known mechanism | C (probably) |
Abbreviations: AI, attachment inhibitor; CCR5i, chemokine receptor 5 inhibitor; CI, capsid inhibitor; FI, fusion inhibitor; INSTI, integrase strand transfer inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; NNRTI, non-nucleoside reverse transcriptase inhibitor; PAI, post-attachment inhibitor; PI, protease inhibitor
Ritonavir and cobicistat were not included since they are used as a booster in antiretroviral therapy.
Likelihood score based on case report categories from: National Institute of Diabetes and Digestive and Kidney Diseases. LiverTox: clinical and research information on drug-induced liver injury. Updated June 20, 2024. Accessed September 29, 2025. https://www.ncbi.nlm.nih.gov/books/NBK547852/.
Includes tenofovir disoproxil fumarate or tenofovir alafenamide.
Main Study Outcome
HCC was defined by: 1) VA National Cancer Registry, which records cancers diagnosed or treated within the VA,21 or 2) ICD-9/10-CM diagnosis codes, requiring ≥1 inpatient or ≥2 outpatient diagnoses recorded within the VA (Supplementary Table 2). HCC diagnoses in the VA National Cancer Registry were determined by topography and histology codes from the International Classification of Diseases for Oncology, Third Edition (ICD-O-3).22 We further increased the specificity of HCC ICD-9/-10-CM diagnoses by requiring the absence of bile duct disease (intrahepatic bile duct carcinoma 155.1, malignant neoplasm of intrahepatic bile ducts C22.1). As secondary outcomes, we evaluated: 1) hepatic decompensation, using previously validated ICD-9/-10 diagnostic coding algorithms,23,24 and 2) all-cause mortality.
Covariates
We evaluated demographic, clinical, laboratory, and pharmacy variables at baseline and during each month of follow-up. Demographic variables included age, sex, race, ethnicity, body mass index (BMI), and year of ART initiation. We ascertained clinical variables that might influence selection or change of an ART regimen and affect risk of HCC. For example, a patient developing diarrhea as an adverse effect due to protease inhibitors may result in ART change,25 while drug-drug interactions resulting from statin use for cardiovascular diseases or anti-epileptic therapies may result in an ART change (Supplementary Figure 2).26,27 Clinical variables included tobacco use; alcohol use disorder (AUD); and diagnoses of diarrhea, epilepsy, hyperlipidemia, osteoporosis, diabetes mellitus, myocardial infarction, heart failure, peripheral vascular disease, cerebrovascular accident/transient ischemic attack, chronic obstructive pulmonary disease, and connective tissue disease (Supplementary Table 3).
Laboratory data (evaluated as continuous values) included HIV RNA, CD4+ percentage, estimated glomerular filtration rate (eGFR), white blood cells, hemoglobin, platelets, alanine aminotransferase, aspartate aminotransferase, total bilirubin, and albumin. We evaluated CD4+ percentage, rather than absolute CD4+ count, because absolute CD4+ count may decrease during cirrhosis due to portal hypertension-induced splenic sequestration.28 We selected baseline values from dates closest but prior to start of follow-up. During follow-up, when multiple results of the same test were measured within a month, we used the lowest value for albumin, CD4+ percentage, eGFR, hemoglobin, and platelets and highest value for BMI, HIV RNA, white blood cells, total bilirubin, and liver aminotransferases. When a laboratory test was not updated, we carried forward the value from the prior month. We did not lag laboratory values because we used them to model treatment decisions and not the outcome.
We evaluated exposure to the following drugs (based on coverage by dispensed fills) within 90 days of the start of follow-up or at each month of follow-up, since these may influence ART selection: proton pump inhibitors, H2 blockers, statins, rifamycins (rifabutin, rifapentine, rifampin), oral anticoagulants, and amiodarone. We also determined polypharmacy (use of ≥5 non-ART medications within the prior 90 days) and Hepatotoxicity Score (an index that accounts for co-administration of multiple hepatotoxic drugs by identifying medications with high rates of severe acute liver injury and incorporates weights based on these drugs’ rates).29
Statistical Analysis
We implemented marginal structural models, which account for time-varying treatments and confounders.30–32 Development of models involved two stages: 1) deriving patient-specific weights by modeling the probability of receiving chronic liver injury-inducing ART and the probability that a patient was censored at each month, and 2) evaluating the causal effect of chronic liver injury-inducing ART on HCC, estimated using a weighted regression.
Stage 1 - Derivation of patient-specific weights.
In modeling the probability of receiving chronic liver injury-inducing ART, we estimated the likelihood of each patient receiving this therapy given the patient’s covariate history at baseline and during follow-up. We modeled the probability of such treatment using logistic regression with chronic liver injury-inducing ART as the outcome and all covariates listed in the section above included as both baseline and time-varying independent variables. Next, the probability of censoring patients at each month, given the baseline and time-varying covariates, was estimated using logistic regression with splines on continuous variables. For each patient, the probability of receiving chronic liver injury-inducing ART at each month and the probability of remaining uncensored at each month were multiplied, and the inverse of these probabilities yielded the patient-specific weights.32 Weights were truncated at the 2.5th and 97.5th percentiles to prevent outlying data from exerting undue influence over the sample. In a sensitivity analysis, we explored truncation at the 5th and 95th percentiles.
Stage 2 – Association of HCC with use of chronic liver injury-inducing ART, by CLD status.
We used a marginal structural model to estimate associations between cumulative use of chronic liver injury-inducing ART and HCC, by CLD status. We used a discrete time failure model, which is based on a logistic regression, to estimate odds ratios (ORs) with 95% confidence intervals (CIs) of HCC for cumulative use of chronic liver injury-inducing ART. The model estimated ORs of HCC for specified durations of time exposed to chronic liver injury-inducing ART versus similar time on ART without such antiretrovirals. The models included all covariates above in addition to number of months of chronic liver injury-inducing ART (excluding current month), ART era (early ART [prior to 2014] versus modern ART [2014 and later]), and number of months of follow-up. Modern ART corresponded to HIV treatment guidelines recommending a once-daily integrase strand transfer inhibitor as part of first-line ART regimens.33 In one analysis, we dichotomized cumulative exposure (≥4 years versus <4 years) based on biological plausibility34 and our observation that rates of HCC were higher in exposed versus unexposed PWH at ≥4 years. In another analysis, we classified cumulative exposure as 4-8 years and ≥8 years to assess potential for a dose-response relationship. To assess the distribution of hepatotoxic ART use over time, we determined the proportion of patients receiving chronic liver injury-inducing ART by calendar year.35
Secondary analyses:
We identified the proportion of HCC events that developed during follow-up while patients were: 1) still receiving chronic liver injury-inducing ART, 2) off chronic liver injury-inducing ART for <5 years, or 3) off chronic liver injury-inducing ART for ≥5 years. Results were stratified according to cumulative exposure to chronic liver injury-inducing ART (no exposure; <4 years; ≥4 years). Given the shared pathway of chronic liver disease progression on HCC and hepatic decompensation, we additionally evaluated associations between chronic liver injury-inducing ART and incident hepatic decompensation, by CLD status. We evaluated death as a secondary outcome for cause-specific analysis for competing risk, using the same discrete time failure model as the primary analysis. To further explore the impact of temporal trends on the association between chronic liver injury-inducing ART and HCC, in a sensitivity analysis, we classified calendar year into five quantiles (≤2004, 2005-2009, 2010-2014, 2015-2019, ≥2020) and re-performed the analysis.
Analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC). We used multiple imputation with chained equations to assign values for missing baseline data, by means of 10 imputations using all baseline and time-updated variables and outcomes.36 Results across the 10 datasets were combined to arrive at CIs that accounted for within- and across-dataset variances.
RESULTS
Among 63,728 PWH dispensed ART during the observation period, we identified 19,472 new ART initiators (Supplementary Figure 3). Of these, 18,625 were eligible to be classified by CLD status (11,854 without CLD; 6,771 with CLD). Chronic liver injury-inducing ART was dispensed across all years of observation (see Supplementary Figure 4).
PWH without CLD.
Among 11,854 PWH without CLD, 5,582 initiated chronic liver injury-inducing ART, and 6,272 initiated ART without such antiretrovirals (Table 2). Exposed PWH were older; less frequently had BMI ≥25 kg/m2, AUD, hyperlipidemia, and diarrhea; and more frequently had higher Hepatotoxicity Scores than those unexposed.
Table 2.
Characteristics of persons with HIV at the time of antiretroviral therapy initiation in the Veterans Health Administration between October 1, 1999 and March 31, 2023, by chronic liver disease status.
| Patients without CLD (n=11,854) | Patients with CLD (n=6,771) | |||||
|---|---|---|---|---|---|---|
|
| ||||||
| Characteristic* | ART with Chronic Liver Injury Mechanism† (n=5,582) | ART without Chronic Liver Injury Mechanism† (n=6,272) | Std Diff | ART with Chronic Liver Injury Mechanism† (n=3,490) | ART without Chronic Liver Injury Mechanism† (n=3,281) | Std Diff |
| Median age (years, IQR) | 47 (40-54) | 45 (34-55) | 0.166 | 49 (44-54) | 53 (45-59) | 0.270 |
| Male sex | 5,409 (96.9%) | 5,979 (95.3%) | 0.081 | 3,404 (97.5%) | 3,167 (96.5%) | 0.060 |
| Race/ethnicity | 0.108 | 0.238 | ||||
| Black, non-Hispanic | 2,619 (46.9%) | 3,161 (50.4%) | 2,038 (58.4%) | 1,925 (58.7%) | ||
| Caucasian, non-Hispanic | 1,878 (33.6%) | 2,147 (34.2%) | 766 (21.9%) | 867 (26.4%) | ||
| Hispanic | 365 (6.5%) | 450 (7.2%) | 226 (6.5%) | 224 (6.8%) | ||
| Other/Unknown | 720 (12.9%) | 514 (8.2%) | 460 (13.2%) | 265 (8.1%) | ||
| Body mass index (prior year) | 0.260 | 0.166 | ||||
| Underweight (<18.50 kg/m2) | 285 (5.1%) | 184 (2.9%) | 209 (6.0%) | 133 (4.1%) | ||
| Normal (18.50-24.99 kg/m2) | 2,730 (48.9%) | 2,490 (39.7%) | 1,808 (51.8%) | 1,536 (46.8%) | ||
| Overweight (25.00-29.99 kg/m2) | 1,772 (31.7%) | 2,176 (34.7%) | 1,054 (30.2%) | 1,067 (32.5%) | ||
| Obesity (30.00-34.99 kg/m2) | 569 (10.2%) | 996 (15.9%) | 303 (8.7%) | 401 (12.2%) | ||
| Morbid obesity (≥35.00 kg/m2) | 191 (3.4%) | 413 (6.6%) | 101 (2.9%) | 140 (4.3%) | ||
| Missing weight and/or height in prior year | 35 (0.6%) | 13 (0.2%) | 15 (0.4%) | 4 (0.1%) | ||
| Alcohol use disorder diagnosis (ever prior) | 205 (3.7%) | 399 (6.4%) | 0.123 | 1,696 (48.6%) | 1,935 (59.0%) | 0.209 |
| Tobacco use (ever) | 3,269 (58.6%) | 3,418 (54.5%) | 0.180 | 2,547 (73.0%) | 2,322 (70.8%) | 0.155 |
| Comorbidities (ever prior) | ||||||
| Hyperlipidemia | 764 (13.7%) | 1,663 (26.5%) | 0.324 | 481 (13.8%) | 988 (30.1%) | 0.402 |
| Osteoporosis | 12 (0.2%) | 28 (0.4%) | 0.040 | 14 (0.4%) | 24 (0.7%) | 0.044 |
| Epilepsy | 55 (1.0%) | 75 (1.2%) | 0.020 | 60 (1.7%) | 87 (2.7%) | 0.064 |
| Diarrhea | 473 (8.5%) | 755 (12.0%) | 0.118 | 399 (11.4%) | 532 (16.2%) | 0.139 |
| Diabetes mellitus | 458 (8.2%) | 635 (10.1%) | 0.067 | 390 (11.2%) | 485 (14.8%) | 0.107 |
| Myocardial infarction | 64 (1.1%) | 98 (1.6%) | 0.036 | 75 (2.1%) | 79 (2.4%) | 0.017 |
| Congestive heart failure | 144 (2.6%) | 196 (3.1%) | 0.033 | 155 (4.4%) | 195 (5.9%) | 0.068 |
| Peripheral vascular disease | 113 (2.0%) | 171 (2.7%) | 0.046 | 112 (3.2%) | 135 (4.1%) | 0.048 |
| Cerebrovascular ischemic attack | 144 (2.6%) | 225 (3.6%) | 0.058 | 134 (3.8%) | 173 (5.3%) | 0.069 |
| Chronic obstructive pulmonary disease | 741 (13.3%) | 1,020 (16.3%) | 0.084 | 712 (20.4%) | 852 (26.0%) | 0.132 |
| Connective tissue disease | 39 (0.7%) | 48 (0.8%) | 0.008 | 40 (1.1%) | 47 (1.4%) | 0.025 |
| Median laboratory values (IQR; prior year) | ||||||
| HIV RNA (log copies/mL) | 11 (10-12) | 11 (10-12) | 0.152 | 11 (10-12) | 11 (10-12) | 0.113 |
| Not assessed in the prior year | 0 (0%) | 0 (0%) | -- | 0 (0%) | 0 (0%) | -- |
| CD4+ cell percentage (%)‡ | 14.0 (7.0-21.0) | 19.0 (12.0-27.0) | 0.489 | 14.0 (8.0-22.0) | 19.0 (11.0-27.0) | 0.385 |
| Not assessed in the prior year | 211 (3.8%) | 228 (3.6%) | 0.008 | 104 (3.0%) | 88 (2.7%) | 0.018 |
| Hemoglobin (g/dL) | 13.5 (12.0-14.7) | 13.9 (12.5-14.9) | 0.182 | 13.3 (11.8-14.5) | 13.5 (12.1-14.6) | 0.098 |
| Not assessed in the prior year | 66 (1.2%) | 34 (0.5%) | 0.069 | 23 (0.7%) | 8 (0.2%) | 0.062 |
| Alanine aminotransferase (U/L) | 30.0 (20.0-46.0) | 26.0 (19.0-39.0) | 0.181 | 37.6 (24.0-61.0) | 32.0 (21.0-52.0) | 0.224 |
| Not assessed in the prior year | 414 (7.4%) | 87 (1.4%) | 0.297 | 102 (2.9%) | 20 (0.6%) | 0.176 |
| Aspartate aminotransferase (U/L) | 30.0 (23.0-41.0) | 27.0 (21.0-36.0) | 0.227 | 40.0 (28.0-65.0) | 34.0 (25.0-55.0) | 0.268 |
| Not assessed in the prior year | 330 (5.9%) | 138 (2.2%) | 0.189 | 82 (2.3%) | 48 (1.5%) | 0.065 |
| Platelet count (x 106/L) | 207.0 (164.0-256.0) | 210.0 (170.0-255.0) | 0.040 | 193.0 (146.0-244.0) | 204.0 (161.0-252.0) | 0.154 |
| Not assessed in the prior year | 141 (2.5%) | 55 (0.9%) | 0.128 | 44 (1.3%) | 15 (0.5%) | 0.087 |
| White blood cell count (cells/μL) | 4.6 (3.5-5.8) | 4.9 (3.9-6.1) | 0.188 | 4.4 (3.4-5.6) | 4.8 (3.7-6.0) | 0.240 |
| Not assessed in the prior year | 46 (0.8%) | 30 (0.5%) | 0.043 | 22 (0.6%) | 6 (0.2%) | 0.070 |
| Albumin (g/dL) | 3.8 (3.4-4.2) | 4.0 (3.6-4.3) | 0.277 | 3.6 (3.2-4.0) | 3.8 (3.3-4.1) | 0.219 |
| Not assessed in the prior year | 509 (9.1%) | 289 (4.6%) | 0.179 | 187 (5.4%) | 99 (3.0%) | 0.117 |
| Total bilirubin (mg/dL) | 0.5 (0.4-0.7) | 0.5 (0.4-0.7) | 0.020 | 0.6 (0.4-0.8) | 0.5 (0.4-0.7) | 0.058 |
| Not assessed in the prior year | 276 (4.9%) | 153 (2.4%) | 0.133 | 62 (1.8%) | 27 (0.8%) | 0.084 |
| eGFR (mL/min/1.73m2) | 96.0 (80.0-111.0) | 99.0 (83.0-113.0) | 0.111 | 100.0 (84.0-113.0) | 98.0 (83.0-111.0) | 0.082 |
| Not assessed in the prior year | 205 (3.7%) | 39 (0.6%) | 0.212 | 47 (1.3%) | 8 (0.2%) | 0.124 |
| Medications dispensed (prior 90 days) | ||||||
| Polypharmacy (≥5 non-ART medications) | 798 (14.3%) | 840 (13.4%) | 0.026 | 733 (21.0%) | 624 (19.0%) | 0.050 |
| Proton pump inhibitors | 596 (10.7%) | 817 (13.0%) | 0.073 | 559 (16.0%) | 647 (19.7%) | 0.097 |
| H2 blockers | 393 (7.0%) | 226 (3.6%) | 0.154 | 278 (8.0%) | 168 (5.1%) | 0.115 |
| Statins | 369 (6.6%) | 614 (9.8%) | 0.116 | 174 (5.0%) | 361 (11.0%) | 0.223 |
| Rifamycins | 54 (1.0%) | 33 (0.5%) | 0.051 | 54 (1.5%) | 28 (0.9%) | 0.064 |
| Oral anticoagulants | 60 (1.1%) | 119 (1.9%) | 0.068 | 32 (0.9%) | 51 (1.6%) | 0.058 |
| Amiodarone | 8 (0.1%) | 10 (0.2%) | 0.004 | 2 (0.1%) | 6 (0.2%) | 0.036 |
| Hepatotoxicity Score at ART initiation (IQR) § | 0.416 | 0.265 | ||||
| 0 | 2,270 (40.7%) | 3,762 (60.0%) | 1,201 (34.4%) | 1,479 (45.1%) | ||
| Low | 225 (4.0%) | 316 (5.0%) | 1,056 (30.3%) | 960 (29.3%) | ||
| Medium | 2,359 (42.3%) | 1,782 (28.4%) | 840 (24.1%) | 621 (18.9%) | ||
| High | 728 (13.0%) | 412 (6.6%) | 393 (11.3%) | 221 (6.7%) | ||
Abbreviations: ART, antiretroviral therapy; CLD, chronic liver disease; eGFR, estimated glomerular filtration rate; HIV, human immunodeficiency virus; IQR, interquartile range; RNA, ribonucleic acid; std diff, standardized difference
Results reported as n (%) unless otherwise specified.
Mechanism of chronic liver injury includes mitochondrial toxicity and accumulation of toxic intermediates.
CD4+ percentage, rather than absolute CD4+ count, was evaluated because absolute CD4+ count may decrease during cirrhosis due to portal hypertension-induced splenic sequestration.28
The Hepatotoxicity Score uses weights based on the age-/sex-adjusted incidence rates of severe acute liver injury of the most hepatotoxic medications obtained from: Lo Re V, 3rd, Newcomb CW, Carbonari DM, et al. Hepatotoxicity Score: A new method to adjust for use of potentially hepatotoxic medications by chronic liver disease status. Pharmacoepidemiol Drug Saf. Dec 2024;33(12):e70069. doi:10.1002/pds.70069.Hepatotoxicity Score was categorized with categories of 0, low (3.0–4.9), medium (5.0–14.9), and high (≥ 15.0) for patients with no chronic liver disease, and 0, low (50.0–124.9), medium (125.0–249.9), and high (≥ 250.0) for patients with chronic liver disease.
Unadjusted rates of HCC for those exposed to chronic liver injury-inducing ART for ≥4 years and ≥8 years were higher than for those unexposed for each year of use (Supplementary Table 4); thus, we evaluated ART exposure at these thresholds (Figures 1 and 2).
Figure 1.
Incidence rates (per 10,000 person-years) with 95% confidence intervals of hepatocellular carcinoma among people with HIV (PWH) who newly initiated antiretroviral therapy (ART) within the within the United States Department of Veterans Affairs between October 1, 1999 and March 31, 2023, by antiretroviral mechanism, length of time on therapy, and chronic liver disease status.
Abbreviations: ART, antiretroviral therapy; PWH, people with HIV
Figure 2.
Cumulative incidence of hepatocellular carcinoma among people with HIV (PWH) without (Panel A) and with (Panel B) chronic liver disease who newly initiated antiretroviral therapy (ART) within the within the United States Department of Veterans Affairs between October 1, 1999 and March 31, 2023, by length of time on therapy.
Abbreviations: HCC, hepatocellular carcinoma; PWH, people with HIV
Within the marginal structural model employing stabilized weights truncated at the 2.5th and 97.5th percentiles, ≥4 years of exposure to chronic liver injury-inducing ART was associated with HCC compared to ≥4 years of use of ART without chronic liver injury mechanisms (OR=4.02 [95% CI, 1.00-16.24]; Table 3). Additionally, a dose-response relationship was observed. Greater duration of exposure to chronic liver injury-inducing ART was associated with higher ORs of HCC (4-8 years of cumulative use: OR=2.21 [95% CI, 0.25-19.48]; ≥8 years of cumulative use: OR=8.03 [95% CI, 1.64-39.22]). Results were similar with stabilized weights truncated at the 5th and 95th percentiles (Supplementary Table 5).
Table 3.
Odds ratios with 95% confidence intervals of hepatocellular carcinoma associated with specified durations of exposure to antiretroviral therapy with a chronic liver injury mechanism among people with HIV in the Veterans Health Administration, by chronic liver disease status.
| No. of Patients | No. of Events | Incidence Rate (95% CI), Events/10,000 person-years | Odds Ratio (95% Confidence Interval) |
|||
|---|---|---|---|---|---|---|
| Length of Exposure | Unadjusted* | Baseline Weighted | MSM Stabilized Weighted with 2.5% Truncation† | |||
|
PWH without chronic liver disease
| ||||||
| Model 1 | ||||||
| 4+ years of use of ART without chronic liver injury mechanism‡ | 6,627 | 11 | 2.6 (1.5-4.8) | Reference | Reference | Reference |
| 4+ years of use of ART with chronic liver injury mechanism‡ | 3,784 | 30 | 7.1 (4.9-10.1) | 3.10 (1.44-6.65) | 1.91 (0.85-4.30) | 4.02 (1.00-16.24) |
| Model 2 | ||||||
| 4-8 years of use of ART without chronic liver injury mechanism‡ | 6,627 | 4 | 1.6 (0.6-4.3) | Reference | Reference | Reference |
| 4-8 years of use of ART with chronic liver injury mechanism‡ | 3,784 | 12 | 5.0 (2.9-8.8) | 4.16 (1.29-13.39) | 2.09 (0.58-7.54) | 2.21 (0.25-19.48) |
| 8+ years of use of ART without chronic liver injury mechanism‡ | 3,556 | 7 | 4.2 (2.0-8.7) | Reference | Reference | Reference |
| 8+ years of use of ART with chronic liver injury mechanism‡ | 2,081 | 18 | 9.7 (6.1-15.4) | 2.59 (1.04-6.43) | 1.84 (0.69-4.91) | 8.03 (1.64-39.22) |
|
| ||||||
|
PWH with chronic liver disease
| ||||||
| Model 1 | ||||||
| 4+ years of use of ART without chronic liver injury mechanism‡ | 3,213 | 38 | 19.7 (14.4-27.1) | Reference | Reference | Reference |
| 4+ years of use of ART with chronic liver injury mechanism‡ | 1,994 | 75 | 38.7 (30.8-48.5) | 2.26 (1.48-3.45) | 2.09 (1.33-3.31) | 1.36 (0.68-2.72) |
| Model 2 | ||||||
| 4-8 years of use of ART without chronic liver injury mechanism‡ | 3,213 | 23 | 19.5 (12.9-29.3) | Reference | Reference | Reference |
| 4-8 years of use of ART with chronic liver injury mechanism‡ | 1,994 | 45 | 38.7 (28.9-51.8) | 2.35 (1.38-3.98) | 2.06 (1.14-3.69) | 1.22 (0.53-2.85) |
| 8+ years of use of ART without chronic liver injury mechanism‡ | 1,588 | 15 | 20.2 (12.2-33.4) | Reference | Reference | Reference |
| 8+ years of use of ART with chronic liver injury mechanism‡ | 993 | 30 | 38.7 (27.1-55.4) | 2.14 (1.13-4.05) | 2.13 (1.10-4.12) | 1.53 (0.46-5.09) |
Abbreviations: MSM, marginal structural model; PWH, persons with HIV
Unadjusted estimates of odds ratios with 95% confidence intervals did not require imputation. All other estimates were determined by multiple imputation using chained equations with 10 imputed datasets.
MSM stabilized weights were truncated at the 2.5th and 97.5th percentiles to prevent outlying data from exerting undue influence over the sample.
Mechanism of chronic liver injury includes mitochondrial toxicity and accumulation of toxic intermediates.
Of the HCC events occurring with <4 years of exposure to chronic liver injury-inducing ART, 39% occurred on treatment, while 50% and 11% occurred <5 years or ≥5 years after discontinuing chronic liver injury-inducing ART, respectively (Supplementary Figure 5). Of the HCC events occurring with ≥4 years of exposure to chronic liver injury-inducing ART, 50% occurred on treatment, while 40% and 10% occurred <5 years or ≥5 years after discontinuing chronic liver injury-inducing ART, respectively.
In secondary analyses, we observed a similar dose-response relationship between chronic liver injury-inducing ART and incident hepatic decompensation, such that rates of hepatic decompensation were greater among PWH without CLD and ≥4 years of chronic liver injury-inducing ART (Supplementary Table 6). Rates of death for those exposed for ≥1 year to chronic liver injury-inducing ART were higher than for those unexposed (Supplementary Table 7). Within the marginal structural model employing stabilized weights truncated at the 2.5th and 97.5th percentiles, ≥4 years of exposure to chronic liver injury-inducing ART was associated with all-cause mortality compared to ≥4 years of use of ART without chronic liver injury mechanisms (OR=1.20 [95% CI, 1.01-1.42]; Supplementary Table 8); however, no dose-response was observed. In the sensitivity analysis classifying calendar time into five quantiles, findings were similar (Supplementary Table 9).
PWH with CLD.
Among 6,771 PWH with CLD, 3,490 initiated chronic liver injury-inducing ART, and 3,281 initiated ART without such antiretrovirals (Table 2). Exposed PWH were younger; less frequently had BMI ≥25 kg/m2, AUD, hyperlipidemia, diarrhea, chronic obstructive pulmonary disease, and diabetes; and more frequently had higher Hepatotoxicity Scores than those unexposed.
Unadjusted incidence rates of HCC for those receiving chronic liver injury-inducing ART were higher with ≥4 years and ≥8 years of exposure than for those unexposed (Supplementary Table 4); therefore, we chose these thresholds for analysis (Figures 1 and 2).
Within the marginal structural model employing stabilized weights truncated at the 2.5th and 97.5th percentiles, ≥4 years of exposure to chronic liver injury-inducing ART was not associated with HCC (OR=1.36; 95% CI, 0.68-2.72; Table 3). Evaluating a dose-response relationship showed no association (4-8 years of cumulative use: OR=1.22 [95% CI, 0.53-2.85]; ≥8 years of cumulative use: OR=1.53 [95% CI, 0.46-5.09]). Results were similar with stabilized weights truncated at the 5th and 95th percentiles (Supplementary Table 5).
Of the HCC events occurring with <4 years of exposure to chronic liver injury-inducing ART, 45% occurred on treatment, while 39% and 16% occurred <5 years or ≥5 years after discontinuing chronic liver injury-inducing ART, respectively (Supplementary Figure 5). Of the HCC events occurring with ≥4 years of cumulative exposure to chronic liver injury-inducing ART, 57% occurred on treatment, while 33% and 9% occurred <5 years or ≥5 years after discontinuing chronic liver injury-inducing ART, respectively.
In secondary analyses, we observed that rates of hepatic decompensation were greater among PWH with CLD who had ≥4 years of chronic liver injury-inducing ART (Supplementary Table 6). Rates of death for those exposed for ≥1 year to chronic liver injury-inducing ART were higher than for those unexposed (Supplementary Table 7). Within the marginal structural model using stabilized weights truncated at the 2.5th and 97.5th percentiles, ≥4 years of exposure to chronic liver injury-inducing ART was not associated with all-cause mortality compared to ≥4 years of use of ART without chronic liver injury mechanisms (OR=1.10 [95% CI, 0.93-1.31]); Supplementary Table 8). In the sensitivity analysis classifying calendar time into five quantiles, findings were similar (Supplementary Table 9).
DISCUSSION
In this study of PWH initiating ART in the VA system, we found that among 11,854 PWH without CLD, ≥4 years of exposure to chronic liver injury-inducing ART was significantly associated with HCC compared to ≥4 years of use of ART without chronic liver injury mechanisms. Importantly, a dose-response relationship was observed, and greater duration of exposure to chronic liver injury-inducing ART was associated with higher ORs of HCC. In contrast, among 6,771 PWH with CLD, ≥4 years of exposure to chronic liver injury-inducing ART was not associated with HCC. Regardless of CLD status, approximately half of all HCC events occurring after use of chronic liver injury-inducing ART were diagnosed after discontinuing this therapy, suggesting that the adverse hepatic effects of these antiretrovirals may persist despite their discontinuation. While the majority of chronic liver injury-inducing antiretrovirals are no longer included in first-line ART regimens, these findings may assist clinicians in further tailoring HCC risk assessment among PWH previously exposed to ART that included such medications.
Few studies have evaluated associations between cumulative use of antiretrovirals and HCC, and none specifically examined the impact of chronic liver injury-inducing antiretrovirals. In one cohort study of 45,544 PWH in Australia, Europe, and the United States followed in the Data Collection on Adverse Events of Anti-HIV Drugs (D:A:D) study (1999-2014), cumulative exposure (per 5 years) to stavudine (relative rate [RR]=1.46 [95% CI, 1.20-1.77]), didanosine (RR=1.32 [95% CI, 1.07-1.63]), tenofovir (RR=1.46 [95% CI, 1.11-1.93]) and (fos)amprenavir (RR=1.47 [95% CI, 1.01-2.15]) was associated with increased rate of a composite of HCC/hepatic decompensation.37 Our study extends these findings by evaluating the effect of antiretrovirals capable of inducing chronic liver injury, which represents a mechanistically plausible contributor to HCC, and presents results stratified by CLD, which could modify the long-term effects of such antiretrovirals.
Among PWH with CLD, we did not observe an association between cumulative use of chronic liver injury-inducing ART and HCC. While CLD includes a heterogenous group of conditions, the majority of patients had chronic viral hepatitis and/or hazardous alcohol use.38 Viral hepatitis and alcohol-related liver disease have known mechanisms increasing the risk of HCC, either through development of cirrhosis, direct cellular toxicity, or oncogenic effects.39 The risk of HCC driven by these known mechanisms may dwarf any additional impact of chronic liver injury-inducing ART, resulting in no observed association.
Among people with ≥4 years of cumulative exposure to chronic liver injury-inducing ART, approximately 50% of HCC events occurred after discontinuation of these medications. The greatest proportion of HCC events occurring after discontinuing chronic liver injury-inducing ART were identified within 5 years of discontinuation, suggesting heightened vigilance for HCC may be warranted during this period. Prior research has demonstrated that a heightened risk of hepatic decompensation persists following discontinuation of dideoxynucleoside analogues.7,37 These findings highlight the legacy effect that these medications exert, suggesting that the hepatic injury inflicted might persist after discontinuation.
Our study has several limitations. First, there is the possibility that we might have misclassified CLD status, particularly among PWH with as-yet undiagnosed CLD. Second, residual or unmeasured confounding could have impacted our findings given the observational nature of the study. While we accounted for hepatotoxic non-ART medications using the Hepatotoxicity Score,29 we did not ascertain potentially hepatotoxic herbal or dietary supplements, since dispensed fills for these products are not routinely recorded in electronic health record databases. Third, AUD diagnoses were evaluated to capture significant alcohol exposure, since HCC risk is most strongly associated with history of AUD.40 Non-severe alcohol use was not captured in this study; thus, the impact of low-level use could not be evaluated. Fourth, while our analyses allow for causal inference, formal causality assessment to adjudicate whether chronic liver injury-inducing ART was implicated in HCC was beyond the scope of this study. Future work evaluating histopathologic features or epigenetic signatures associated with drug-induced chronic liver injury might improve understanding of the hepatocarcinogenic potential of these antiretrovirals. Fifth, many of the point estimates for the primary marginal structural models had wide confidence intervals, indicating a notable degree of uncertainty. Replication of our study in different settings is warranted to confirm our findings. Use of complementary approaches, such as target trial emulation, may further elucidate effects of chronic liver injury-inducing ART, including class-specific effects. Finally, since our study predominantly included male patients, the results may not be generalizable to females.
In conclusion, this study found that ≥4 years of chronic liver injury-inducing ART was associated with HCC among PWH without CLD, and a dose-response relationship was observed. Among PWH with CLD, cumulative exposure to chronic liver injury-inducing ART was not significantly associated with HCC. While ART has provided dramatic life-saving benefits to millions of PWH, adverse effects of these medications, both with current, as well as with past, use are important to understand to determine if clinical interventions are needed. This study’s findings, if replicated, may assist clinicians in further refining HCC risk assessment among PWH with cumulative exposure to chronic liver injury-inducing ART.
Supplementary Material
ACKNOWLEDGEMENTS
The authors would like to acknowledge the contributions of Sara Lodi, PhD, to this study.
The contents of this article do not represent the views of the Department of Veterans Affairs or the United States Government. This work uses data provided by patients and collected by the VA as part of their care and support. Due to US Department of Veterans Affairs (VA) regulations and our ethics agreements, the analytic datasets used for this study are not permitted to leave the VA without a Data Use Agreement. This limitation is consistent with other studies based on VA data. However, VA data are made freely available to researchers with an approved VA study protocol. For more information, please visit https://www.virec.research.va.gov or contact the VA Information Resource Center (VIReC) at VIReC@va.gov.
Funding:
This study was funded, in part, by the National Cancer Institute (R01CA206465, ACJ and VLR), National Institute on Alcohol Abuse and Alcoholism (P01AA029545, ACJ; U24AA020794, ACJ; U01AA020790, ACJ; U24AA022001, ACJ; U01AA013566, ACJ), National Institute of Diabetes and Digestive and Kidney Diseases (K08DK132977, JT), and National Institute of Allergy and Infectious Diseases (T32AI055435).
Footnotes
Conflict of Interest: VLR has received consulting fees from Entasis, GlaxoSmithKline, Takeda, and Urovant Sciences.
Ethics Statement: This study was approved by the Institutional Review Boards of the University of Pennsylvania, Yale University, and VA Connecticut Healthcare System.
REFERENCES
- 1.Gowda C, Newcomb CW, Liu Q, et al. Risk of acute liver injury with antiretroviral therapy by viral hepatitis status. Open Forum Infect Dis 2017; 4(2): ofx012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Stine JG, Chalasani N. Chronic liver injury induced by drugs: a systematic review. Liver Int 2015; 35(11): 2343–53. [DOI] [PubMed] [Google Scholar]
- 3.Bjornsson ES, Andrade RJ. Long-term sequelae of drug-induced liver injury. J Hepatol 2022; 76(2): 435–45. [DOI] [PubMed] [Google Scholar]
- 4.Walker UA, Setzer B, Venhoff N. Increased long-term mitochondrial toxicity in combinations of nucleoside analogue reverse-transcriptase inhibitors. AIDS 2002; 16(16): 2165–73. [DOI] [PubMed] [Google Scholar]
- 5.Cotte L, Benet T, Billioud C, et al. The role of nucleoside and nucleotide analogues in nodular regenerative hyperplasia in HIV-infected patients: a case control study. J Hepatol 2011; 54(3): 489–96. [DOI] [PubMed] [Google Scholar]
- 6.Ji C. Molecular Factors and Pathways of Hepatotoxicity Associated with HIV/SARS-CoV-2 Protease Inhibitors. Int J Mol Sci 2023; 24(9). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Lo Re V 3rd, Zeldow B, Kallan MJ, et al. Risk of liver decompensation with cumulative use of mitochondrial toxic nucleoside analogues in HIV/hepatitis C virus coinfection. Pharmacoepidemiol Drug Saf 2017; 26(10): 1172–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Engels EA, Yanik EL, Wheeler W, et al. Cancer-attributable mortality among people with treated human immunodeficiency virus infection in North America. Clin Infect Dis 2017; 65(4): 636–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Navarro VJ, Senior JR. Drug-related hepatotoxicity. N Engl J Med 2006; 354(7): 731–9. [DOI] [PubMed] [Google Scholar]
- 10.Jungst C, Cheng B, Gehrke R, et al. Oxidative damage is increased in human liver tissue adjacent to hepatocellular carcinoma. Hepatology 2004; 39(6): 1663–72. [DOI] [PubMed] [Google Scholar]
- 11.Li Y, Yu Y, Yang L, Wang R. Insights into the role of oxidative stress in hepatocellular carcinoma development. Front Biosci (Landmark Ed) 2023; 28(11): 286. [DOI] [PubMed] [Google Scholar]
- 12.Keiding S, Andreasen PB. Hepatic clearance measurements and pharmacokinetics. Pharmacology 1979; 19(3): 105–10. [DOI] [PubMed] [Google Scholar]
- 13.Morgan DJ, McLean AJ. Clinical pharmacokinetic and pharmacodynamic considerations in patients with liver disease. An update. Clin Pharmacokinet 1995; 29(5): 370–91. [DOI] [PubMed] [Google Scholar]
- 14.Palatini P, De Martin S. Pharmacokinetic drug interactions in liver disease: An update. World J Gastroenterol 2016; 22(3): 1260–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Russo FP, Zanetto A, Pinto E, et al. Hepatocellular carcinoma in chronic viral hepatitis: Where do we stand? Int J Mol Sci 2022; 23(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.US Department of Veterans Affairs. Corporate Data Warehouse. Accessed at: https://www.hsrd.research.va.gov/for_researchers/vinci/cdw.cfm. Accessed on February 28, 2025.
- 17.Gandhi RT, Landovitz RJ, Sax PE, et al. Antiretroviral Drugs for Treatment and Prevention of HIV in Adults: 2024 Recommendations of the International Antiviral Society-USA Panel. JAMA 2025; 333(7): 609–28. [DOI] [PubMed] [Google Scholar]
- 18.Gandhi NR, Tate JP, Rodriguez-Barradas MC, et al. Validation of an algorithm to identify antiretroviral-naive status at time of entry into a large, observational cohort of HIV-infected patients. Pharmacoepidemiol Drug Saf 2013; 22(9): 1019–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lo Re V 3rd, Kallan MJ, Tate JP, et al. Hepatic decompensation in antiretroviral-treated patients co-infected with HIV and hepatitis C virus compared with hepatitis C virus-monoinfected patients: a cohort study. Annals of internal medicine 2014; 160(6): 369–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.National Institute of Diabetes and Digestive and Kidney Diseases. LiverTox: clinical and research information on drug-induced liver injury. Accessed at: https://www.ncbi.nlm.nih.gov/books/NBK547852/. Accessed on July 2, 2024.
- 21.Park LS, Tate JP, Sigel K, et al. Association of viral suppression with lower AIDS-defining and non-AIDS-defining cancer incidence in HIV-infected veterans: A prospective cohort study. Ann Intern Med 2018; 169(2): 87–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Fritz A, Percy C, Jack A, Shanmugaratnam K, Sobin L, Parkin DM, et al. , eds. International Classification of Diseases for Oncology. 3rd ed. Geneva: World Health Organization; 2000. [Google Scholar]
- 23.Lo Re V 3rd, Lim JK, Goetz MB, et al. Validity of diagnostic codes and liver-related laboratory abnormalities to identify hepatic decompensation events in the Veterans Aging Cohort Study. Pharmacoepidemiol Drug Saf 2011; 20(7): 689–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Haque LY, Tate JP, Chew M, Caniglia EC, Taddei TH, Lo Re V 3rd. A validated algorithm to identify hepatic decompensation in the Veterans Health Administration electronic health record system. Pharmacoepidemiol Drug Saf 2024; 33(11): e70024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Bonfanti P, Valsecchi L, Parazzini F, et al. Incidence of adverse reactions in HIV patients treated with protease inhibitors: a cohort study. Coordinamento Italiano Studio Allergia e Infezione da HIV (CISAI) Group. J Acquir Immune Defic Syndr 2000; 23(3): 236–45. [DOI] [PubMed] [Google Scholar]
- 26.Zaporojan L, McNamara PH, Williams JA, Bergin C, Redmond J, Doherty CP. Seizures in HIV: The case for special consideration. Epilepsy Behav Case Rep 2018; 10: 38–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Spanakis M, Alexakis K, Ioannou P. Antiretroviral therapy and associated drug interactions with cardiovascular drugs: a critical review. Front Pharmacol 2025; 16: 1685710. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.McGovern BH, Golan Y, Lopez M, et al. The impact of cirrhosis on CD4+ T cell counts in HIV-seronegative patients. Clin Infect Dis 2007; 44(3): 431–7. [DOI] [PubMed] [Google Scholar]
- 29.Lo Re V 3rd, Newcomb CW, Carbonari DM, et al. Hepatotoxicity Score: A new method to adjust for use of potentially hepatotoxic medications by chronic liver disease status. Pharmacoepidemiol Drug Saf 2024; 33(12): e70069. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Joffe MM, Ten Have TR, Feldman HI, Kimmel SE. Model selection, confounding control, and marginal structural models: review and new applications. Am Statistician 2004; 58: 272–9. [Google Scholar]
- 31.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(5): 561–70. [DOI] [PubMed] [Google Scholar]
- 32.Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology 2000; 11(5): 550–60. [DOI] [PubMed] [Google Scholar]
- 33.Gunthard HF, Aberg JA, Eron JJ, et al. Antiretroviral treatment of adult HIV infection: 2014 recommendations of the International Antiviral Society-USA Panel. JAMA 2014; 312(4): 410–25. [DOI] [PubMed] [Google Scholar]
- 34.Woodward C, Smith J, Acreman D, Kumar N. Hepatocellular carcinoma in body builders; an emerging rare but serious complication of androgenic anabolic steroid use. Ann Hepatobiliary Pancreat Surg 2019; 23(2): 174–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Moodie EEM, Schulz J. A simple diagnostic for the positivity assumption for continuous exposures. Stat Med 2025; 44(15-17): e70194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Royston P. Multiple imputation of missing values: update. The STATA Journal 2005; 5(2): 188–201. [Google Scholar]
- 37.Ryom L, Lundgren JD, De Wit S, et al. Use of antiretroviral therapy and risk of end-stage liver disease and hepatocellular carcinoma in HIV-positive persons. AIDS 2016; 30(11): 1731–43. [DOI] [PubMed] [Google Scholar]
- 38.Soriano V, Barreiro P, Sherman KE. The changing epidemiology of liver disease in HIV patients. AIDS Rev 2013; 15(1): 25–31. [PubMed] [Google Scholar]
- 39.Farazi PA, DePinho RA. Hepatocellular carcinoma pathogenesis: from genes to environment. Nat Rev Cancer 2006; 6(9): 674–87. [DOI] [PubMed] [Google Scholar]
- 40.Morgan TR, Mandayam S, Jamal MM. Alcohol and hepatocellular carcinoma. Gastroenterology 2004; 127(5 Suppl 1): S87–96. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.