Reporting anything less than 100% adherence during viral suppression raised the odds of rebound by 50% at the next viral load measure in an analysis of 1986 participants in the SMART trial. The researchers believe their findings "highlight the importance of continued adherence counselling, even in people with VL [viral load] suppression."
Ample research demonstrates that adherence to antiretroviral therapy (ART) strongly predicts virologic outcomes. But SMART trial investigators noted a lack of robust data exploring associations between adherence during virologic suppression and risk of subsequent rebound. To address that question, they analyzed adherence and rebound rates with a self-reported adherence tool that predicted virologic outcomes in previous research.
The study involved SMART trial participants randomized to continuous ART (excluding participants randomized to CD4+-count-guided ART interruptions). Every four months participants completed a seven-day recall adherence survey in which they ranked their adherence to each component of their regimen as 100% or less than 100%. Researchers collected viral load pairs in which the second measure occurred at least four weeks, but not more than four months, after the first. Adherence could be reported as viral load pairs any time between four months before the first viral load measure up to seven days after that viral load. The researchers defined a suppressed viral load as ≤ 50 copies/mL and a rebound as a subsequent viral load >200 copies/mL. They used logistic regression to assess associations between adherence during viral suppression and a rebound in the next viral load measure.
The analysis considered 10,761 viral load pairs from 1986 SMART participants. Most viral load observations were from men (73%) and whites (51%). The United States contributed the highest proportion of viral load observations (46%), followed by Western Europe (33%). Participants reported suboptimal (less than 100%) adherence in 1220 viral load pairs (11.3%). Viral load rebounds happened in 507 viral load pairs (4.7%). Rebounds occurred in 8.6% of viral load pairs with suboptimal adherence during the first viral load and in 4.2% of pairs with optimal adherence during the first viral load (P < .0001). Suboptimal adherence had high negative predictive value (95.8%) and specificity (89.1%) for predicting viral load rebound and low positive predictive value (8.6%) and sensitivity (20.7%).
A logistic regression model adjusted for demographic and lifestyle factors determined that suboptimal adherence during viral suppression raised the odds of viral rebound 62% (odds ratio [OR] 1.62, 95% confidence interval [CI] 1.28 to 2.06, P = .0004). Further adjustment for HIV-specific variables determined that less than 100% adherence during suppression boosted odds of subsequent rebound 60% (OR 1.60, 95% CI 1.26 to 2.04, P = .0006). Additional adjustment for current regimen determined that suboptimal adherence during suppression raised the odds of viral rebound 51% (OR 1.51, 95% CI 1.19 to 1.92, P = .002). Secondary analyses yielded no evidence that the impact of adherence on viral rebound differed by type of antiretroviral regimen.
Among the 10,761 viral load pairs, there were 121 treatment interruptions (1.1%) between the first and second viral load in the pair and 525 treatment switches (4.9%) between the first and second viral load. Rebounds occurred after 47.9% of interruptions compared with 4.2% of viral load pairs not separated by an interruption (P < .0001). Rebounds occurred after 13.1% of treatment switches compared with 4.3% of viral load pairs not separated by a switch (P < .0001).
Although the adherence tool used in this analysis did not yield perfect predictions, the authors noted that it is "an inexpensive tool which is easy to utilize ... and does provide some predictive value." Thus, they recommend it as an appropriate tool for assessing the impact of adherence on viral load.