29 [95% confidence interval (CI) 0.92–1.80]. Rebound risks increased with decreasing levels of coverage: patients with 80–95% adherence had a 2.69% risk of rebound

(compared with 100% adherence: RR=1.62; 95% CI 1.23–2.14), patients with 60–80% adherence had a 3.15% risk of rebound (RR=1.90; 95% CI 1.39–2.61) and patients with adherence below 60% had a 3.26% risk of rebound (RR=1.97; 95% CI 1.40–2.78). When the percentage of drug coverage was analysed as a continuous variable (thus assuming that the true underlying relationship between adherence and the log risk ratio is linear) the risk of viral rebound decreased by 9% (RR=0.91; 95% CI 0.87–0.95; P=0.0001) per 10% higher coverage. After adjusting for potential confounding factors (variables shown in Table 2), low levels of drug coverage continued to be significantly associated with viral rebound: rates of viral rebound were increased by 51% (RR=1.51; 95% CI 1.14–1.99), EX 527 supplier 70% (RR=1.70; 95% CI 1.24–2.33) and 75% (RR=1.75;

95% CI 1.24–2.47) in patients who had drug coverage of 80–95, 60–80 and <60%, respectively (Fig. 2). When the drug coverage was analysed as a continuous variable, the risk of viral rebound decreased by 7% per 10% higher adherence (RR=0.93; 95% CI 0.88–0.98; P=0.004). Other Smad inhibitor independent predictors of viral rebound were shorter duration of VL suppression, higher number of previous virological failures, currently being on an unboosted PI regimen compared with an NNRTI-containing regimen, having experienced two or more treatment interruptions (while VL detectable at the time), having started HAART in the calendar period 1997–1999 compared with 2003–2006, and having a time-zero for the DCVL episode in the period 2002–2003 compared with 2006–2007. The results of the analysis stratified by most common current regimen (unboosted PI, boosted PI and NNRTI-based regimen) suggested that the risk of viral rebound at a particular level of adherence differed according to the regimen type received (Fig. 3). For example, at the lowest levels of adherence (≤60%), the risk of rebound was,

respectively, 5.24, 3.50 and 2.19%, for patients receiving unboosted until PI, boosted PI and NNRTI-based regimens, while among subjects who adhered completely these risks were 1.46, 1.89 and 1.47%, respectively. In sensitivity analyses, we considered the proportion of days covered by a prescription for at least one drug, instead of three, as our adherence measure and obtained similar results (data not shown). In addition, we considered the effect of modifying the definition of viral rebound from a threshold of 200 to 50 copies/mL. In this case, the overall risk of rebound was higher (5.36%) but the factors associated with rebound were generally similar. The estimated RR of VL rebound for a 10% higher coverage was 0.95 (95% CI 0.92–0.99), and after adjusting for the risk factors considered in the main analysis, the RR weakened marginally to 0.97 (95% CI 0.93–1.