Non-adherence to highly active antiretroviral therapy predicts progression to AIDS

The introduction of highly active antiretroviral therapy (HAART) has produced a dramatic reduction in mortality among HIV-infected individuals 1–4. Whereas the level of adherence to HAART is closely associated with suppression of the HIV viral load in plasma 5–14, a relationship between adherence and disease progression has not been established. Here we examine the relationship between pill-count adherence 10 and progression to AIDS, in a representative cohort of homeless and marginally housed individuals. Three hundred and thirty HIV-positive individuals were recruited into the Research in Access to Care in the Homeless (REACH) cohort between July 1996 and April 2000. We performed adherence assessments in 76 of the 81 (94%) individuals without AIDS who were taking HAART on or after January 1998. Pill counts are conducted every 3–6 weeks at the subject's usual place of residence (single room occupancy hotel, shelter, or other) on an unannounced day as previously described 10. The HIV-1 viral load was determined monthly and the CD4 cell count was determined quarterly. Record match, through the San Francisco Department of Public Health AIDS Surveillance Registry, was used to identify opportunistic infections not otherwise detected by study protocol. No patients were lost to follow-up during the study. Progression was defined as a decline in the CD4 cell count to below 200 cells/μl or the development of an opportunistic infection during follow-up. Adherence to antiretroviral medication was based on unannounced pill counts, and was measured as the mean percentage of pills taken over the observation period (0–100%), and categorized as 50% or less, 51–90%, or over 90%. The relationship between adherence and progression-free survival was assessed using the Kaplan–Meier method with the three adherence groups compared by log-rank test. A Cox proportional hazards model was used to calculate relative risk (RR) with 95% confidence intervals (CI), and to adjust the relationship of adherence with progression-free survival for CD4 cell count, viral load, and months of HAART therapy at baseline. Baseline characteristics with a P value of less than 0.10 were considered candidates for adjustment. We tested each potential confounder in a two-predictor Cox proportional hazards model that included the potential confounder and adherence. The sample was composed mostly of non-white (59%) men (86%) who had a high prevalence of recent injection drug use (29%), lifetime psychiatric hospitalization (32%), and homelessness (93%). At entry, 47 individuals (62%) were on a protease inhibitor-containing regimen; 23 (30%) were on a non-nucleoside-based regimen; five (7%) were taking a regimen containing both a protease inhibitor and a non-nucleoside reverse transcriptase inhibitor; and one (1%) was taking three nucleoside analog reverse transcriptase inhibitor. A total of 47 (61%) individuals were antiretroviral naive before their current HAART regimen. At baseline, subjects had already received a median of 14 months of HAART. Median pill-count adherence in the 76 individuals was 65%. One-third (31%) had low adherence (≤ 50%); half (49%) had moderate adherence (51–90%); and 20% had high adherence (> 90%). None of the high-adherence group developed an AIDS event during observation (mean observation time 16 months), compared with 8% with moderate adherence (13 months) and 41% with low adherence (11 months) (P = 0.0012, log-rank test). In a bivariate analysis, each 10% difference in mean adherence was associated with a 28% reduction in risk of progression to AIDS (RR 0.72; 95% CI 0.59–0.87). In addition to adherence, baseline CD4 cell count, baseline viral load, and the proportion of months with viral suppression all predicted progression to AIDS (P < 0.05). Failure to complete high school (P = 0.07), total months of HAART therapy at enrollment (P = 0.10), and injection drug use within 30 days of baseline (P = 0.08) were marginally associated with progression to AIDS. A 10% difference in average adherence was associated with a 21% reduction in relative risk (RR 0.79; 95% CI 0.63–0.99) when adjusted for baseline CD4 cell count, a 22% reduction when adjusted for baseline viral load (RR 0.78; CI 0.62–1.00), a 27% reduction when adjusted for baseline number of months of HAART therapy (RR 0.73; CI 0.60–0.89), a 28% reduction when adjusted for failure to complete high school (RR 0.72; CI 0.60–0.88), and a 27% reduction when adjusted for baseline injection drug use (RR 0.73; CI 0.59–0.89) (see Fig. 1).Fig. 1.: Proportion of individuals who were AIDS free by adherence level. Dotted line, 90–100% adherence; solid line, 50–89% adherence; dashed line, 0–49% adherence. Dots represent censored observations.In summary, we found a strong relationship between the level of adherence to antiretroviral therapy and the risk of progression to AIDS in a population-based cohort of HIV-positive, urban poor adults with a high risk of non-adherence. These findings are consistent with findings presented by Hogg et al.15, indicating that failure to obtain scheduled pharmacy refills is associated with a more rapid progression to AIDS and death. There are several limitations to our analysis. Although we found a strong and statistically significant relationship between adherence and disease progression, the number of events is small. Because of the small number of events, we controlled for several important confounders in separate two-predictor models including adherence and the potential confounder, and found that the effect persisted. We were not able to control for all potential confounders simultaneously. The median follow-up of the group was 13 months; thus our observations are relevant to short-term disease progression, and factors predicting long-term disease progression may be different. Finally, it is possible that adherence assessment may have changed adherence behavior in some individuals. Whereas most subjects did not achieve the more than 90% level of adherence required for reliable viral suppression 10,13, few individuals progressed to AIDS in the moderate (50–89%) adherence group and no one progressed in the 90–100% adherence group. Although moderate adherence is suboptimal for virological control, few individuals in this group progressed to AIDS when compared with the group with less than 50% adherence. This observation raises the hypothesis that the level of adherence required to produce clinical benefit may be lower than that required for viral suppression. If true, this hypothesis would explain the apparent contradiction between the dramatic declines in AIDS-related mortality 1–4 throughout the developed world after the introduction of HAART, in spite of both suboptimal adherence and viral suppression in most clinical cohorts 13,16. Although the adequacy of a less rigorous adherence level for clinical benefit would be reassuring, our conclusion requires confirmation, and the goal of treatment should be 100% adherence in all patients on HAART. Acknowledgements The authors would like to acknowledge Dr Peter Bacchetti for comments regarding the statistical analysis; Ms Paula Zenti, Ms Johanna Crane, and Ms Maureen Morgan for conducting the patient adherence assessments; and Ms Nelia Dela Cruz, Mr John Day, and Mr Jason Bonafacio for assisting with the REACH Cohort. They would also like to thank Ms Priscilla Chu at the San Francisco Department of Public Health for assistance performing the AIDS surveillance registry match. David R. Bangsbergab Sharon Perryc Edwin D. Charleboisa Richard A. Clarkc Marjorie Roberstond Andrew R. Zolopae Andrew Mossc