Background:  Because the HIV viral reservoir decreases with the duration of viral suppression, we hypothesized that the level of adherence necessary to maintain viral suppression would also decrease with past duration of viral suppression.

Methods:  The REACH cohort of HIV⁺ homeless and marginally housed individuals on stable highly active antiretroviral therapy and achieving HIV RNA <50 copies/mL with monthly unannounced home pill counts were eligible. HIV RNA was measured monthly. Virologic failure was defined as HIV RNA >50 copies/mL. Adherence was categorized into 4 levels (0 to 49%, 50 to 74%, 75 to 89%, and 90 to 100% pills taken). We used a marginal structural model, and controlled for confounding by CD4 nadir, regimen characteristics, past adherence, age, sex, depression, ethnicity, and drug or alcohol use, among others. Targeted maximum likelihood estimation was used, and inference was based on the nonparametric bootstrap using 500 iterations.

Results:  A total of 221 subjects was studied (mean age 46; mean CD4 nadir 233). Most were either on an NNRTI (38%), a boosted PI (32%), or both (8%). The mean time on current regimen at date of adherence monitoring was 23 months; 87, however, had adherence data from the start of a period of viral suppression. Comparing the probability of failure just after achieving suppression versus after 12 consecutive months of suppression, there was a statistically significant decrease in the probability of virologic failure for those with at least 50% adherence. The estimated decrease in risk of failure was 0.45 (95%CI 0.28 to 0.70) at 50 to 74% adherence, 0.25 (CI 0.05 to 0.50) at 75 to 89% adherence, and 0.39 (CI 0.29 to 0.52) at 90 to 100% adherence. For adherence levels below 50%, estimated risk of failure decreased, but this was not statistically significant.

Conclusions:  The risk of virologic failure, for any given level of adherence above 50%, declines over time. While high level adherence is required to maximize the probability of durable viral suppression, the range of adherence capable of sustaining viral suppression increases with the duration of continuous viral suppression. These data suggest that once the virus is fully controlled, the level of drug exposure necessary to maintain such control declines; however, we cannot rule out selection bias as an alternative explanation. Our results may have implications for short-term interruption and induction/maintenance studies.