Methods

Participants and study design. The design and data collection methods of the SMART trial have previously been reported [6]. In brief, the SMART study was a randomized clinical trial that compared 2 distinct strategies of using antiretroviral therapy in a cohort of participants aged >13 years who had confirmed HIV-1 infection and CD4⁺ cell counts of ⩾350 cells/µL at the time of screening.

One strategy was the drug conservation strategy, where participants deferred antiretroviral therapy until the CD4⁺ cell count decreased to <250 cells/µL. Antiretroviral therapy was initiated or resumed until the CD4⁺ cell count reached ⩾350 cells/µL and then suspended again. Cycles of antiretroviral therapy were based on those CD4⁺ cell counts or the presence of HIV-related symptoms or if the CD4⁺ cell percentage decreased to <15%. The protocol allowed for participants with chronic HBV infection to use single-drug anti–hepatitis virus medication (e.g., adefovir) while not receiving antiretroviral therapy. The other strategy was viral suppression, which stipulated that participants should initiate or continue antiretroviral therapy with the goal of maximum viral suppression in accordance with HIV treatment guidelines [15]. The choice of antiretroviral agents and combinations was based on clinician or participant preference, and therapy was continued without interruption.

The primary end point of the study was the development of a new or recurrent opportunistic disease or death from any cause. Opportunistic diseases included those defined by the 1993 Centers for Disease Control and Prevention criteria [16], as well as additional conditions related to immunodeficiency. An end point review committee that was unaware of treatment assignment reviewed each opportunistic disease event and death.

Hepatitis status. During screening, participants medical records were reviewed for documentation of hepatitis B and C status. If there was no laboratory evidence of 1 positive hepatitis B surface antibody or 2 positive hepatitis B surface antigen results obtained at least 6 months apart, these tests were performed, as well as either IgG core antibody or total core antibody testing. If there was no evidence of a positive hepatitis C antibody result from anytime in the past or a negative result from within the previous year, an antibody test was performed. Chronic hepatitis C was defined as the presence of hepatitis C antibody and denoted as “HCV positive.” Chronic hepatitis B was defined as the persistence of hepatitis B surface antigen throughout 6 months and denoted “HBV positive.”

In a separate analysis, baseline plasma samples obtained from HBV-positive and/or HCV-positive participants were analyzed for levels of HCV RNA and HBV DNA using branched DNA assays (Versant HCV RNA 3.0 and Versant HBV DNA 3.0, respectively; Bayer Diagnostics), which have lower levels of detection of 615 and 357 IU/mL, respectively.

Data collection and follow-up. Before randomization, the following participant information was collected: antiretroviral therapy history, nadir CD4⁺ cell count and peak viral load, prior 3 laboratory results for CD4⁺ cell count and percentage, and HIV load. Participants were seen at 1 month and every 2 months during year 1 and every 4 months from year 2 onward.

Statistical analysis. Baseline characteristics (table 1) were compared between drug conservation and viral suppression coinfected participants and between all coinfected and all HIV-monoinfected participants using Pearson's Χ² test for binomial proportions for categorical variables and Wilcoxon rank tests for continuous variables.

Cox proportional hazards models were used to compare the coinfected and the HIV-monoinfected groups in the drug conservation and viral suppression arm with respect to event rates for the primary end point (opportunistic disease or death from all causes) and its components (fatal or nonfatal opportunistic disease and non—opportunistic disease death). The interaction between treatment group and coinfection status was examined by including an interaction term in expanded Cox models. Adjusted models included the following baseline covariates: age, sex, race, CD4⁺ cell count, nadir CD4⁺ cell count, HIV RNA level (⩽400 vs. >400 copies/mL), antiretroviral therapy status, prior opportunistic disease event, and history of alcohol abuse. The number of participants needed to be enrolled for one to experience a non—opportunistic disease death if randomized to drug conservation versus viral suppression (number needed to harm) was calculated by dividing 1 by the difference in incidence rate (number of events per person-year) between the drug conservation and viral suppression groups.

Time-to-event analyses were censored at the earliest date of opportunistic disease or death, date lost to follow-up, or 11 January 2006. Only 1 event of a given type was counted for each participant. All P values are 2-sided. Analyses were performed using SAS statistical software, version 9.1 (SAS Institute).

Because of the relatively few HBV-positive participants enrolled in the SMART study, HBV-positive and HCV-positive participants were grouped together in the primary analyses. A sensitivity analysis, excluding HBV-positive patients, was also performed.

Results

Baseline characteristics. There were 5472 participants enrolled from 8 January 2002 through 11 January 2006, with 2752 in the viral suppression arm and 2720 in the drug conservation arm. Overall, 930 participants (17%) were coinfected (i.e., HBV and/or HCV positive). A total of 120 (2.2%) were HBV positive, 796 (14.5%) were HCV positive, and 14 (0.3%) were HBV and HCV positive.

Compared with the HIV-monoinfected participants, coinfected participants were more likely to be older and black, to have a history of alcohol abuse, to have lower baseline CD4⁺ cell counts, to have unsuppressed baseline viral load, have a prior AIDS event, to not be receiving antiretroviral therapy at baseline, and to have a longer history of antiretroviral therapy use (table 1). The baseline median CD4⁺ cell count in coinfected participants was 582 cells/µL, and the nadir CD4⁺ cell count was 250 cells/µL.

The study's enrollment was halted on 11 January 2006, when the Data and Safety Monitoring Board determined that there were safety concerns associated with the drug conservation strategy [6]. At that time, the follow-up time was 1469 person-years (median duration of follow-up, 1.3 person-years per participant) for the coinfected and 5879 (0.9 person-years per participant) for the HIV-monoinfected participants. The rates of loss to follow-up were 2.5% and 1.5% in the coinfected and monoinfected participants, respectively.

Primary end point and its components in the drug conservation versus viral suppression groups. During follow-up in the SMART study, overall as of January 2006, 172 participants experienced at least 1 primary end point (i.e., an opportunistic disease or death due to any cause; 122 primary end point [rate, 3.4 events per 100 person-years; 95% CI, 2.8–4.0 events per 100 person-years] in the drug conservation group and 50 [rate, 1.4 per 100 person-years; 95% CI, 1.0–1.8] in the viral suppression group; hazard ratio [HR], 2.5; 95% CI, 1.8–3.5; P<.001). The overall number of participants who died among those in the drug conservation group was 55 (rate, 1.5 deaths per 100 person-years; 95% CI, 1.1–1.9 deaths per 100 person-years), compared with 30 (rate, 0.8 deaths per 100 person-years; 95% CI, 0.5–1.1 deaths per 100 person-years) among those in the viral suppression group (HR, 1.8; 95% CI, 1.2–2.9; P=.007).

In the HBV- and/or HCV-coinfected cohort, 42 participants (rate, 5.7 events per 100 person-years; 95% CI, 4.0–7.4 events per 100 person-years) in the drug conservation group experienced a primary event, compared with 15 (rate, 2.1 events per 100 person-years; 95% CI, 1.0–3.2 events per 100 person-years) in the viral suppression group (drug conservation/viral suppression HR, 2.7; 95% CI, 1.5–4.8) (table 2). For HIV-monoinfected patients, 80 events (2.8 events per 100 person-years; 95% CI, 2.2–3.4 events per 100 person-years) occurred in the drug conservation group, compared with 35 (rate, 1.2 events per 100 person-years; 95% CI, 0.8–1.6 events per 100 person-years) in the viral suppression group (drug conservation/viral suppression HR, 2.4; 95% CI, 1.6–3.5). The drug conservation/viral suppression HRs did not vary by coinfection status (interaction P=.72).

Both the group of patients with concurrent hepatitis and HIV-monoinfected participants randomized to the drug conservation arm versus the viral suppression arm were at increased risk of an opportunistic disease event (drug conservation/viral suppression HR, 6.2 [95% CI, 1.8–20.9] and 2.8 [95% CI, 1.7–4.6], respectively; interaction P=.24) (table 2).

The risk of non—opportunistic disease death was increased in the drug conservation group, compared with the viral suppression group, irrespective of viral hepatitis coinfection status (drug conservation/viral suppression HR, 1.9 [95% CI, 1.0–3.9] and 1.8 [95% CI, 0.9–3.4], respectively; interaction P=.84) (table 2).

Primary end point and its components in HBV- or HCV-coinfected versus HIV-monoinfected participants. In the cohort of HBV-positive and/or HCV-positive coinfected patients, 57 participants (rate, 3.9 events per 100 person-years; 95% CI, 2.9–4.9 events per 100 person-years) had an opportunistic disease or died of any cause compared with 115 (rate, 2.0 events per 100 person-years; 95% CI, 1.6–2.4 events per 100 person-years) opportunistic disease or deaths in the HIV-monoinfected group (unadjusted coinfected/monoinfected HR, 1.9; 95% CI, 1.4–2.7). Adjustment for baseline variables did not change these findings (coinfected/monoinfected HR, 1.6; 95% CI, 1.2–2.3) (figure 1).

Among the coinfected subgroup, 22 patients (rate, 1.5 events per 100 person-years; 95% CI, 0.9–2.1 events per 100 person-years) experienced at least 1 opportunistic disease event (of which 2 were fatal), compared with 78 patients (rate, 1.3 events per 100 person-years; 95% CI, 1.0–1.6 events per 100 person-years; 5 events were fatal) in the HIV-monoinfected group (unadjusted and adjusted coinfected/monoinfected HR, 1.1 [95% CI, 0.7–1.8] and 0.9 [95% CI, 0.6–1.5], respectively) (figure 1).

When focusing on non—opportunistic disease deaths, a different picture emerged. of the 78 participants who died of those causes, 37 were HBV and/or HCV coinfected, and 41 were not, providing rates of 2.5 events per 100 person-years (95% CI, 1.7–3.3 events per 100 person-years) and 0.7 events per 100 person-years (95% CI, 0.5–0.9 events per 100 person-years) in the 2 groups, respectively (unadjusted and adjusted coinfected/monoinfected HR, 3.6 [95% CI, 2.3–5.6] and 2.9 [95% CI, 1.8–4.6], respectively) (figure 1).

Although the relative risk of being randomized to the drug conservation group was comparable for coinfected and HIV-monoinfected participants, a higher underlying risk of non—opportunistic disease death among coinfected patients resulted in a higher absolute risk of non—opportunistic disease death in this subgroup. We found 1.6 more non—opportunistic disease deaths per 100 person-years in drug conservation, compared with viral suppression, in the coinfected group and 0.4 more non—opportunistic disease deaths per 100 person-years in drug conservation, compared with viral suppression, in the monoinfected group. In other words, the numbers of participants enrolled for 1 patient to experience a non—opportunistic disease death if randomized to the drug conservation group were 61 and 255 for coinfected and monoinfected participants, respectively.

The causes of the 37 non—opportunistic disease deaths in the coinfected group included renal (n=3), hepatic (n=2), non—opportunistic disease malignancy (n=7), substance abuse (n=7), other (n=8), and unknown (n=10) (figure 2). Both hepatic deaths occurred in patients who were HCV positive. The differences between the drug conservation and viral suppression arms in the coinfected group were seen primarily in renal and unknown deaths. The small number of events precluded more formal statistical analysis.

In the drug conservation arm, 1 non—opportunistic disease death patient was HBV positive, 17 were HCV RNA positive, 5 were HCV aviremic, and 2 were HCV positive but did not have HCV RNA data (did not consent). In the viral suppression arm, 1 patient was HBV positive, 9 were HCV RNA positive, and 2 were HCV aviremic.

Sensitivity analyses. In the HBV-positive and/or HCV-positive coinfected cohort, for 877 of the 930 participants, a baseline plasma sample was available: 752 for HCV-positive participants (752 [93%] of 810) and 121 for HBV-positive participants (121 [90%] of 134). Analysis of these samples showed that, among the HCV-positive participants, 554 (73%) had detectable HCV RNA, and among the HBV-positive participants, 65 (53.7%) had detectable HBV DNA.

When analyzed by the presence or absence of detectable HCV RNA and/or HBV DNA viremia at study entry, comparable findings to what is described in the text and in table 2 and figure 1 were discovered (tables 3 and 4). of note, the aviremic HBV-positive and HCV-positive participants had an excess risk of non—opportunistic disease death of a comparable magnitude compared with the viremic participants. Another sensitivity analysis, which excluded HBV-positive and analyzed HCV-positive patients only, resulted in a comparable overall conclusion (results not shown).

Discussion

The interruption of antiretroviral therapy can be hazardous and lead to HIV disease progression and non–AIDS-related cardiac, renal, and hepatic complications [6]. The SMART study was the first randomized study to compare the effect of antiretroviral therapy interruption in persons with HBV and/or HCV coinfection. Thus, the present set of analyses adds nuances to this general message.

When the SMART study was designed, it was thought that non—opportunistic disease deaths would be seen more frequently in the viral suppression group than in the drug conservation group because of a higher risk of death associated with adverse effects of antiretroviral therapy [6]. Randomization of both the HBV- and/or HCV-coinfected and HIV-monoinfected participants to the antiretroviral therapy interruption strategy (the drug conservation arm) resulted in increased risk of developing both opportunistic disease or dying due to any cause. The relative risk of non—opportunistic disease death in participants randomized to the drug conservation arm versus the viral suppression arm was comparable irrespective of hepatitis status. However, because the coinfected population had a high a priori underlying risk of non—opportunistic disease death, the absolute risk of non—opportunistic disease death by being randomized to the drug conservation arm was much higher for coinfected participants than for HIV-monoinfected participants. Thus, although constituting only 17% of participants in the SMART study, almost one-half of all non—opportunistic disease deaths occurred in this population.

The number of participants who had to be randomized to the drug conservation arm for one of them to experience a non—opportunistic disease death [17] was 4-fold lower in the coinfected group than in HIV-monoinfected group. Contrary to the situation for non—opportunistic disease deaths, coinfected participants were not at excess risk of opportunistic disease. This finding is consistent with other cohort studies that have also not found an excess risk of these diseases in either HBV- or HCV-coinfected persons [18, 19].

The coinfected participants were slightly older and more often had a history of alcohol abuse. A higher prevalence of Africans or African Americans among the coinfected patients may indicate lower socioeconomic status. The excess risk in the hepatitis virus coinfected population was also seen if participants with ongoing hepatitis virus replication were excluded. This information is important and supports the assumption that differences in lifestyle and socioeconomic factors with expected survival disadvantages explain the higher risk of non—opportunistic disease death in coinfected participants, compared with HIV-monoinfected participants [7, 19]. of note, the coinfected group was only slightly less likely to be undergoing antiretroviral therapy at baseline than was the HIV-monoinfected group, and for those receiving treatment, the chance of antiretroviral therapy being able to fully suppress HIV replication was comparable irrespective of hepatitis virus status.

In the pre–antiretroviral therapy era, most persons coinfected with HBV and/or HCV infection died of the underlying immunodeficiency caused by uncontrolled HIV infection. Although this situation remains important also after effective antiretroviral therapy has been introduced [20, 21], an increasing proportion of deaths in recent years among coinfected persons are due to liver-related causes, cardiovascular diseases, non-AIDS malignancies, violent deaths, overdose of narcotic drugs, and other causes [7, 11]. In this study, among participants in the viral suppression group, substance abuse and non-AIDS cancers were the leading causes of death. Only 1 participant died of liver-related causes. Because this cohort was selected on the basis of the ability to participate in a trial and had to have a good immune function at study entry, it is not surprising that we saw few opportunistic disease–related deaths in our coinfected cohort. Although hepatitis virus likely has contributed to the liver-related death [11], it is less likely that other causes of death are due to the viral hepatitis infection [22, 23]. Thus, neither HIV nor hepatitis viruses appear to have contributed to any major extent to the profile of causes of non—opportunistic disease deaths among coinfected participants in the viral suppression arm. However, because the study was interrupted prematurely, and because participants were observed only for short periods [6], it is likely that the number of liver-related deaths would have increased had the study continued.

A wide variety of non—opportunistic disease causes of death were also observed in the drug conservation arm, with no clear dominance of any particular cause. Also in this arm, only 1 participant died of liver-related causes. The high median CD4⁺ cell count at baseline and the fact that the study was halted prematurely and follow-up was subsequently short might explain why we could not detect any adverse influence on liver-related outcomes. Analyses are ongoing to assess whether surrogates of liver injury were adversely affected by the drug conservation strategy.

Three coinfected participants (all in the drug conservation arm) died of renal failure. It is unknown whether any of these deaths were due to hepatorenal syndrome, HIV-related renal failure, or other causes. Previous studies have shown that HCV coinfection, low CD4⁺ cell count, and increasing HIV-RNA level increase the risk of developing renal disease in HIV-1–infected persons [24].

Ten coinfected participants in the drug conservation arm died of unknown causes. Each event in the SMART study was thoroughly reviewed by an end point review committee, and no cause could be determined. Specifically, the committee found no evidence suggesting that any of these deaths were due to liver-related causes.

The finding of a diverse profile of causes of non—opportunistic disease deaths suggests that interruption of antiretroviral therapy affects multiple and not a single pathological process. Recently, data from analyses of various biomarkers in stored samples from the SMART study suggested that ongoing inflammatory and coagulation activity was strongly associated with all-cause death and, furthermore, that interruption of antiretroviral therapy exacerbated these processes [25]. Further understanding of whether these processes are part of the causal pathway to affect the diverse conditions leading to death or merely epiphenomena of other processes awaits further research.

Liver enzyme (alanine aminotransferase and aspartate aminotransferase) levels were not routinely determined in the SMART study, so we do not know whether any of the HBV-positive patients had hepatic flares after antiretroviral therapy interruption, but importantly none of the 2 hepatic deaths occurred in the HBV-positive group.

In summary, although antiretroviral therapy interruption has proven to be deleterious in the general HIV-1–infected population, it is particularly unsafe for viral hepatitis–coinfected persons because of an elevated risk of non—opportunistic disease death. Antiretroviral therapy interruption in the coinfected participants resulted in increased mortality from a wide variety of causes with no clear predominance. However, we did not find that antiretroviral therapy interruption resulted in increased risk of liver-related death in coinfected participants. This finding might be attributable to the high CD4⁺ cell count at baseline and the short follow-up period. Studies that use surrogate markers of liver disease are currently ongoing to further examine these findings.