Ethnicity, Race, and Gender: Differences in Serious Adverse Events Among Participants in an Antiretroviral Initiation Trial: Results of CPCRA 058 (FIRST Study)

Background: Differences in adverse events by gender and race/ethnicity have not been described extensively in randomized clinical trials of HIV antiretroviral therapy (ART).

Methods: Antiretroviral-naive HIV-infected participants enrolled in a long-term randomized clinical trial of 3 different initial ART strategies-protease inhibitor (PI), nonnucleoside reverse transcriptase inhibitor (NNRTI), or PI plus NNRTI-based combinations-with a median follow-up of 5 years, were compared by gender and race for 14 categories of grade 4 adverse events, discontinuation of initial antiretroviral regimen, and all-cause mortality. Multivariate analysis was used to identify predictors of events and death.

Results: Among 1301 participants with complete data, there were 701 blacks, 225 Latinos, and 263 women. Several baseline characteristics differed by gender and race, including age, HIV transmission risk, hepatitis B or C coinfection, viral load, diagnosis of AIDS, body mass index, and baseline hypertension. Grade 4 events occurred in 409 participants (rate: 8.9/100 person-years). There were 176 deaths (rate: 3.0/100 person-years) and 523 discontinuations of regimen for any toxicity (rate: 13/100 person-years). In the fully adjusted regressions, blacks had greater risk for cardiovascular (hazard ratio [HR] = 2.64, 95% confidence interval [CI]: 1.04 to 6.67) and renal (HR = 3.83, 95% CI: 1.28 to 11.5) events. Black men had more psychiatric events (HR = 2.45, 95% CI: 1.13 to 5.30). Women had a higher risk for anemia (HR = 2.34, 95% CI: 1.09 to 4.99).

Conclusion: Among HIV-infected participants initiating ART, there were significant risk-adjusted differences for specific adverse events by gender and race but not in the overall adverse event rates, all-cause mortality, or rates of toxicity-related treatment discontinuations.

In the past decade of combination antiretroviral therapy (ART) for HIV infection, adverse events have been associated with a variety of factors, including coinfection with viral hepatitis,^1-4 immunologic status,^5-8 genetics and drug metabolism,^9-11 and body mass index (BMI)^12,13 among others. There are also complications of treatment that have been attributed to gender^14,15 or race.^16,17 Among women, for example, there have been reports of increased frequency of rash with nevirapine¹⁸ or lactic acidosis with combined use of didanosine and stavudine in pregnancy.¹⁹ Among persons of African descent, there are apparent differences in genetic polymorphisms in the cytochrome P450 system that may account for prolonged metabolic clearance of efavirenz and increased central nervous system (CNS) effects.⁹

Development of ART-associated toxicities may be further complicated and more frequent among HIV-infected subjects with underlying renal, hepatic, cardiovascular, or other comorbid conditions. The higher prevalence of specific preexisting comorbid conditions among racial/ethnic minorities makes evaluation of drug-associated toxicities in diverse HIV-infected populations of critical importance.^20-22

In many randomized clinical trials of initial highly active antiretroviral therapy (HAART) for HIV, the percentage of women and racial/ethnic minorities participating in these studies has been small. The published experience on the rate and types of adverse events by gender and race, therefore, is limited.^23-27 We compared the frequency and types of serious adverse events by gender and race in a large, multicenter, randomized clinical strategy trial of participants initiating ART for HIV infection. We also evaluated all-cause mortality rates and rates of toxicity-related therapy discontinuations. Use of these results could inform HIV treating clinicians about particular issues to consider in the selection of antiretroviral regimens for diverse populations.

METHODS

Study Population

The Terry Beirn Community Programs for Clinical Research on AIDS (CPCRA) is a clinical trials network funded by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health to promote participation in AIDS research activities by traditionally underrepresented groups such as women, people of color, and injecting drug users within their usual sites of medical care. The 18 participating units with 160 sites include community-, veterans-, and hospital-based clinics; private practices; and health maintenance organizations in 16 US locations. These units provide HIV primary care to more than 60,000 individuals.

Study Design

The Flexible Initial Retrovirus Suppressive Therapy (FIRST CPCRA 058) study was a randomized open-label clinical trial comparing the long-term effectiveness of 3 strategies of initial HAART in antiretroviral-naive HIV-1-infected persons.²⁸ The eligibility criteria included documented HIV infection and being at least 13 years of age. Patients were excluded if they were pregnant or breast-feeding, had any prior use of protease inhibitors (PIs) or nonnucleoside reverse transcriptase inhibitors (NNRTIs), or had a cumulative total of more than 4 weeks of nucleoside reverse transcriptase inhibitor (NRTI) use or more than 1 week of lamivudine (3TC) use. Patients were categorized based on self-reported race/ethnicity and gender at birth. The protocol compared 3 arms of treatment in a 1:1:1 allocation: a PI-containing regimen, an NNRTI-containing regimen, and a combination PI- and NNRTI-containing regimen. The study's objectives included the virologic and immunologic effectiveness and durability, development of drug resistance, disease progression, and mortality and safety of each strategy from January 1999 until September 2005. The combined primary outcome was the composite of an AIDS-defining event, death, or CD4 count <200 cells/mm³.

In CPCRA 058, participants randomized to a particular antiretroviral strategy could elect to be enrolled in substudies that specified the antiretroviral agents in each arm or they could select a drug combination based on clinician and patient choice.

Antiretroviral Management

Antiretroviral medications could be changed during the course of the trial for any reason, including drug toxicity, virologic or immunologic failure, or patient and/or clinician choice. Participants were encouraged, however, to stay within their randomized strategy if possible.

Procedures

After signing the informed consent, study participants had a baseline evaluation and were seen every 4 months. At each study visit, research staff did a standardized assessment that collected information on adverse events, new clinical events, treatments or diagnoses (including supplemental hepatic, circulatory, and metabolic conditions), and adherence. The Clinical Events Committee reviewed any new AIDS-defining clinical events or death for confirmation.²⁹ If a participant's vital status was unknown or if he or she could not be contacted, research personnel searched the Social Security Death Index and National Death Index.

Baseline laboratory data included selected measures of alanine aminotransferase (ALT) and aspartate aminotransferase (AST), lipids, glucose, hepatitis B virus surface antigen, total hepatitis B core and surface antibody, and hepatitis C virus antibody. Serum creatinine was not routinely collected. Chronic hepatitis B was defined as persistence of hepatitis B surface antigen for more than 6 months, and chronic hepatitis C infection was defined as a reactive hepatitis C virus antibody test. A central laboratory performed testing for viral load (HIV RNA polymerase chain reaction [PCR] assay), and local laboratories performed CD4 cell counts.

Data Collection

Grade 4 adverse events were divided into categories based on the Division of AIDS (DAIDS) toxicity tables, which include laboratory and clinical symptoms and signs.³⁰

The adverse events were divided into 14 categories as follows: any grade 4 event, cardiovascular (myocardial infarction, congestive heart failure, dysrhythmia, chest pain, hypertension, or hypotension), vascular (deep venous thrombosis, pulmonary embolism, or cerebrovascular accident [CVA]), renal (kidney stone, renal failure, or elevated creatinine), neurologic (peripheral neuropathy), psychiatric (mental status changes, depression, or suicidal ideation/attempt), hepatic (transaminase elevations, elevated bilirubin, or liver failure), gastrointestinal (diarrhea, pancreatitis, nausea/vomiting, abdominal pain, or dehydration), anemia, other hematologic (thrombocytopenia or neutropenia), metabolic (hypertriglyceridemia, hyperkalemia, or metabolic acidosis), pulmonary (dyspnea), infection (sepsis or pneumonia), fever, and dermatologic (rash or dermatitis).

In those instances in which a single participant had multiple grade 4 events, we reviewed the clinical record to assign the event to the most appropriate category or categories (eg, grade 4 hypotension with grade 4 diarrhea was categorized as a gastrointestinal adverse event rather than as a grade 4 vascular adverse event). As another example, grade 4 anemia and a grade 4 gastrointestinal event diagnosed 2 weeks apart were treated as 2 distinct grade 4 events. The authors reviewed reports of near-simultaneous adverse events in multiple categories for selected participants to determine if there was a clinically appropriate assignment to a single category. Seven cases of dyspnea, 4 cases of fever, 7 cases of dehydration, 6 cases of hypotension, and 4 cases of hyperkalemia were reassigned to prevent inaccurate overreporting of adverse events by category. In analyzing adverse events in a given category, only the first adverse event was considered and all data after that time point were censored.

Statistical Analysis

Baseline characteristics were tabulated by gender and race/ethnicity for all patients with complete data. Self-reported race/ethnicity was allocated to 3 groups: Latino or Hispanic, African-American or black, and white/other. F tests and χ² tests were used to detect differences between groups. We tested for any differences among the 6 categories (male/female, further split by 3 racial/ethnic groups) and also tested for gender-related differences adjusted only for race/ethnicity and race/ethnicity adjusted for gender.

Rates per 100 person-years of exposures were calculated for each of the 14 grade 4 adverse event categories cited previously and for each combination of gender and race/ethnicity. Event rates per 100 person-years were also calculated for discontinuations attributable to toxicity; all-cause death; and a combined outcome of death, discontinuations, and any grade 4 adverse events. Proportional hazards regressions, adjusted only for gender and race/ethnicity, were used to test for significant differences in crude event rates.

Fully adjusted proportional hazards regression models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for predictors of each of the 17 outcomes cited previously. Predictors included gender, HIV transmission category, race/ethnicity, age, AIDS diagnosis at baseline, CD4 cell count, log₁₀ HIV RNA level, history of antiretroviral use as outlined in the inclusion criteria, chronic hepatitis B infection, chronic hepatitis C infection, BMI, diabetes, hypertension, and randomization strategy. Two of the adjusting variables in particular, diabetes and hypertension, were included because they are known risk factors for some of the outcomes and because baseline analyses showed that they were not evenly distributed among the study's racial/ethnic groups. They are also known to occur more frequently among blacks and Latinos in the comparably aged non-HIV-infected US population.³¹ For anemia, we performed sensitivity testing by further adjusting for utilization of AZT (zidovudine) or 3TC in the initial regimen.

Separate subanalyses were performed using only events from the first 8 months of the trial. The 8-month time point was a prespecified secondary endpoint for viral load. The expectation was that analyses would be possible because (1) a high proportion of patients would still be on their initial regimen at 8 months and (2) a sufficient number of events would have occurred to provide statistical power.

RESULTS

Baseline Characteristics

Among 1397 randomized trial participants in the FIRST study, 1301 (93% of total participants) had complete baseline data and were included in this study. Of those, 263 (20%) were women, 225 (17%) were Latino/Latina, and 701 (54%) were black (Table 1). There were 375 persons who were neither Latino/Latina nor black and were categorized as white/other (338 self-described whites and 37 in all other groups). Latino men were 2 years younger than the overall population, and Latina women were 2 years older.

Table 1

There were differences among HIV risk categories by race/ethnicity. Men who have sex with men (MSM) were more commonly reported among the whites/other category as compared with black or Latino men (P < 0.0001). There were no significant differences between groups regarding history of injection drug use (IDU). Latinas and white men had the lowest percentage of hepatitis C virus coinfection (9% and 15%, respectively; P = 0.05). Among the 6% of participants who were chronically infected with hepatitis B virus, rates were higher among men (7%) than among women (2%) (P = 0.003, data not shown).

Overall, 38% of participants entered the study with a previous AIDS-defining illness. Overall, blacks and Latinos had higher rates of baseline AIDS and lower CD4 cell counts compared with whites/others (P = 0.001 and P = 0.003, respectively; data not shown). Latinas were the least likely, however, to have had an AIDS-defining illness at baseline. After adjusting for race/ethnicity, women tended to have higher CD4 cell counts (P = 0.07, data not shown) and lower HIV viral loads (P = 0.03, data not shown) compared with men. There were no differences in randomization to antiretroviral strategy by gender or racial/ethnic groups. AZT and 3TC were components in the initial antiretroviral strategy for most patients in all groups, with 3TC being used slightly less frequently among blacks compared with other racial/ethnic groups (78% vs. 85% to 86%; P = 0.01).

Nelfinavir was the most commonly prescribed PI (61%), followed by a ritonavir-boosted PI (26%), whereas efavirenz was prescribed for 63% and nevirapine for 37% of participants in the NNRTI category.

Baseline hypertension was more prevalent among blacks (P < 0.0001), whereas the percentage with baseline diabetes did not differ significantly by race/ethnicity or gender (P = 0.08).

Adverse Events, Discontinuation, and Death

Among 5929 person-years of follow-up, a total of 409 participants had 1 or more grade 4 adverse events, with an event rate of 8.9 per 100 person-years (Table 2). Of the total number of adverse events, most were among male participants (79%) commensurate with their share of the study population.

Table 2

The frequency of grade 4 adverse events was highest in the hepatic category, followed by gastrointestinal, other hematologic, psychiatric, anemia, cardiovascular, and renal event categories (see Table 2). There were a total of 176 deaths (13.5%) over the 5 years of the study for a rate of 3.0 per 100 person-years. There were no deaths attributed to antiretroviral toxicity, and there was no difference in mortality by randomization strategy. In the median follow-up time of 60 months for the study, approximately 18% of those participants who died had no follow-up clinical information that could be obtained.

In analyses that were adjusted for only gender and race/ethnicity, blacks had higher rates of grade 4 adverse events (P = 0.03) and deaths (P = 0.01) than Latinos or white/other participants (data not shown). They also had significantly more grade 4 cardiovascular events (P = 0.03) and renal events (P = 0.01) but had significantly fewer pulmonary events (P = 0.03). There were 523 (40%) participants who discontinued their initial ART regimen because of toxicity (rate: 13 per 100 person-years). Within each gender, blacks had the lowest discontinuation rates, although the difference was not statistically significant (P = 0.06). Black men had higher rates of psychiatric events (P = 0.01), whereas, overall, women had more anemia events (P = 0.01). No significant differences in the event rates were observed between Latino and white participants.

Predictors of Grade 4 Adverse Events, Discontinuations, and Death by Race/Ethnicity and Gender

In the fully adjusted multivariate analysis (Table 3), female gender was associated with an increased risk of grade 4 anemia (adjusted HR = 2.34, 95% CI: 1.09 to 4.99; see Table 3). Further adjustments for use of 3TC and AZT at baseline did not materially change the result, and there were no significant interactions between gender and use of these drugs. There was no difference by gender or race/ethnicity in other hematologic grade 4 events. Black race predicted increased cardiovascular events (adjusted HR = 2.64, 95% CI: 1.04 to 6.67) and renal events (adjusted HR = 3.83, 95% CI: 1.28 to 11.5). Black men had more grade 4 psychiatric events (HR = 2.45, 95% CI: 1.13 to 5.30; shown in Table 3 footnote). The negative association of black race and discontinuation of ART for any toxicity neared significance in the fully adjusted regression model (HR = 0.82, 95% CI: 0.66 to 1.01). Race/ethnicity or gender was not predictive of the rate of overall adverse events or death. Using white/other as the referent group, Latino ethnicity was not predictive of any of the outcomes modeled.

Table 3

In the separate 8-month analysis described previously, 73% of participants were still on their initial randomization strategy. There were 185 grade 4 adverse events, 280 discontinuations attributable to toxicity, and 27 deaths. Analyses of 3 overall categories (any grade 4 adverse event, discontinuation attributable to toxicity, or the combined endpoint of the first two, or all-cause death) failed to reveal significant differences by gender or race/ethnicity. Because of the small numbers, more detailed analyses by specific grade 4 events could not be performed.

DISCUSSION

The FIRST study is the largest randomized clinical strategy trial of antiretroviral-naive patients followed for a median of 5 years that has included sufficient numbers of women and racial/ethnic minority participants to support an analysis of differences in adverse events by gender or race/ethnicity.^23-26 After adjusting for baseline risk factors, this study demonstrated significant racial/ethnic- and gender-related differences for specific types of grade 4 adverse events but not for overall adverse event rates, all-cause mortality rates, or rates of toxicity-related discontinuation of therapy. Specifically, cardiovascular and renal events were reported more frequently among all blacks, and psychiatric events were more common among black men. Women had more grade 4 anemia.

The higher unadjusted rate of grade 4 cardiovascular and renal events among blacks is consistent with the greater rates of heart and kidney disease found in the non-HIV-infected African-American populations.^20,21,32 The prevalence of HIV-associated nephropathy (HAN) among blacks has been reported previously.^33,34 Indeed, HAN is seen almost exclusively in blacks and is associated with lower CD4 cell counts. Black participants were more immunosuppressed than whites/others at baseline, with a median (interquartile range [IQR]) CD4 count of 144 (28 to 311) cells/mm³ compared with 214 (62 to 373) cells/mm³ (data not shown). Although HAN could have accounted for the predominance of renal complications, baseline creatinine levels or other markers of kidney function were not collected to examine this hypothesis. The higher rate of baseline hypertension among blacks in the FIRST study is another likely contributor to the frequency of renal and cardiac morbidity.

Cardiovascular events in HIV-infected persons have been associated with traditional risk factors such as smoking, age, male gender, hypertension, diabetes and dyslipidemia, and long-term exposure to ART.³⁵ In this study of antiretroviral-naive patients, the disproportionately high cardiovascular event rate among blacks could be partially attributable to higher rates of hypertension and diabetes at baseline; the significant difference (P = 0.02) in the crude event rates became merely suggestive after adjusting for baseline risk factors. Other cardiovascular risk factors such as smoking and dyslipidemia may be critically important to these outcomes, but this information was not collected at baseline. The particular influence of initial ART strategy, per se, was not a significant factor in the likelihood of the development of renal or cardiac events.^36-39

Rates of serious psychiatric adverse events were higher among black men. Depression is noted to be prevalent among HIV-infected women; however, mood disorders or mental status changes among black men in clinical trials have not been described extensively.^40-42 The reports of mental illness or depression in HIV-infected men have focused on homosexual cohorts such as the Multicenter AIDS Cohort Study (MACS).⁴³ It is likely that grade 4 psychiatric adverse events represent a constellation of factors that include psychosocial and biologic associations (eg, undiagnosed mental illness and/or CNS effects of HIV infection itself). Although efavirenz is known to have CNS effects and delayed metabolic clearance among blacks, there was no significant difference in psychiatric adverse event rates by NNRTI strategy. Given the prevalence of depressive disorders among women with chronic disease, it is surprising that a greater number of psychiatric events were not reported. Potential factors influencing these results include the inherent differences between participants of clinical trials versus those seen through general medical practices. Many of the participating clinical trial units had additional social support services available to participants, perhaps favorably affecting the quality of life of these participants during the course of the clinical trial. Indeed, most sites enrolled in the FIRST study were also enrolled in a companion adherence study that included various strategies such as medication managers.

The higher risk for anemia as a grade 4 adverse event among women was not unexpected. The female participants included in this study were predominantly premenopausal and African American.⁴⁴ The most common nucleoside backbone in the trial was AZT/3TC, which was used by 55% of all study participants, and anemia would be an anticipated adverse event. Adding baseline use of 3TC and ZDV to the model failed to change the estimated female/male HR significantly, however, suggesting that other factors not assessed in this study may have contributed to the higher risk for anemia.

During this study's 5929 person-years of exposure, there were only 4 grade 4 adverse events in the dermatologic category. It was not possible to detect any difference in adverse events among the groups. The low event rate was surprising, because the medications abacavir, nevirapine, and efavirenz, which were commonly used (12% on abacavir, 27% on nevirapine, and 39% on efavirenz, with a total of 70% on 1 of 2 of these drugs at baseline), are associated with the development of rash. There were 43 episodes of lower grade rashes that led to discontinuation of ART (data not shown). Presumably, clinicians chose to suspend therapy with less serious skin reactions. With a median (IQR) CD4 count of 163 (36 to 331) cells/mm³ among study participants, the risks of rash are diminished for nonnucleoside agents.⁴⁵

Although racial/ethnic minorities and women were randomized proportionately across the 3 strategies in the FIRST study, there were several differences in baseline characteristics that parallel the demographics of the US HIV-infected population and could account for some of the adverse event findings.⁴⁶ Racial/ethnic minorities were more likely to enter this trial of naive participants with a history of AIDS-defining illnesses and, on average, had lower CD4 cell counts than whites and Latinos at entry. Diagnosis and entry into care of HIV infection are frequently delayed among people of color in the United States.^47-50 Persons who initiate ART with more advanced immunosuppression may also be at greater risk for complications from immune reconstitution or ART toxicity.^51,52 We were not able to analyze the grade 4 events for the possibility of immune reconstitution, which could be part of the toxicity profile seen in participants with CD4 counts <200 cells/mm³, because there were not clearly defined criteria for this syndrome. The study does remind us that health care disparities by race/ethnicity have clinical implications that have an impact on the experience of HIV treatment.

The overall coinfection rate with hepatitis C (19%) corresponded to the background rate of IDU (15%) and is comparable to that of other US clinical trials networks.⁵³ Coinfection with hepatitis C is a frequently cited factor in the antiretroviral toxicity experience among HIV-infected persons,^1-4,54 and that was true in this study as well (HR = 2.31, 95% CI: 1.23 to 4.32). Although hepatitis C coinfection was reported frequently among black and Latino men, there was no association identified between gender and race/ethnicity among the reported hepatic events. Coinfection with hepatitis B nearly obtained significance as a predictor of grade 4 hepatitic events (HR = 1.95, 95% CI: 0.96 to 3.96). Hepatitis B was found much less often in women than in men; women had correspondingly lower rates of grade 4 hepatic events, but the difference in hepatic event rates was not statistically significant.

As noted in other observational cohorts of US HIV-infected persons, women in this study had higher BMIs, higher CD4 cell counts, and lower viral loads at baseline compared with men.^55-58 In a recent analysis of the HIV Outpatient Study (HOPS) cohort for those participants who initiated HAART, a higher BMI was not associated with increased toxicity or discontinuation rates.⁵⁹ Overweight and obesity may increase the risk of metabolic and cardiovascular complications over time; however, in this analysis, weight was not a significant predictor of adverse events, discontinuation of therapy, or death.

In analyses adjusted only for gender and race/ethnicity, blacks had higher overall adverse event rates, higher mortality rates, and marginally lower rates of discontinuation attributable to toxicity. These apparent disparities disappeared after adjustment for recognized predictors (eg, lower CD4 cell counts, prior AIDS). The overall rate of discontinuation of the initial regimen (40%) in all groups is comparable to that of the other cohorts in randomized trials of naive patients. Rather than the grade 4 events driving this cessation of therapy, there seem to be lower levels of toxicities or other unrecognized factors that were not measured in this study.

We included all-cause mortality in this analysis to be certain that there were no deaths attributable to antiretroviral toxicity by strategy group. There were no deaths that were related to ART, per se.

There are some limitations to our study. The FIRST study was a clinical strategy trial of initial combination ART, and the gender-specific or racial/ethnic-specific toxicity of single antiretroviral drugs was not a defined analysis. In addition, more than half of the participants in the FIRST study had AZT or stavudine as part of their nucleoside backbone, and nelfinavir was the predominant PI used. The serious adverse event profile by gender or race could be different in the era of boosted PI combinations or other newer generation nucleoside agents. Second, we were not able to distinguish between early adverse events related to immune reconstitution compared with drug toxicity itself. Third, we did not collect information on specific confounders for all participants that could have influenced the development of specific adverse events (eg, smoking, current substance and alcohol abuse).

CONCLUSIONS

In summary, gender and race/ethnicity did influence certain grade 4 adverse events in this large randomized trial of antiretroviral-naive persons followed for a median of 5 years. The HIV clinician who treats women and racial/ethnic minorities needs to consider the underlying comorbid diseases and risks for cardiovascular, renal, and psychiatric complications when selecting an antiretroviral regimen. The long-term success of HIV therapy, however, ultimately depends on earlier identification of HIV-infected individuals and entry into care to minimize antiretroviral toxicity and maximize survival.