As suggested by other researchers [13], there is an issue with comparing adherence in subjects on different ART regimens: different antiretroviral drugs have different pharmacokinetic and pharmacodynamic profiles, and thus different relationships with adherence, viral suppression and drug resistance [28,29,44]. For this reason we assessed the association between adherence and risk of viral rebound stratified by the type of current regimen. Analyses were performed using sas software (version 9.1; SAS Institute, Cary, NC, USA). All tests of significance used P<0.05 as the threshold of statistical significance. By 1 October 2007, a total

of 2547 patients on HAART with available antiretroviral drug prescriptions data and VL measurements were included within the Royal Free HIV Cohort. Of these, 2290 (90%) attained a VL measurement of ≤50 copies/mL at least once while receiving HAART, MEK inhibitor on or after the date of the first recorded prescription. Of these, 1632 buy KPT-330 patients contributed a total of 15 660 eligible DCVL episodes to this study [median 8 episodes; interquartile range (IQR) 4–14]. The characteristics of the patients included in the analysis at time-zero for each DCVL episode are shown in Table 1. In the majority of episodes, the patient was male (78%), white (68%), homosexual/bisexual (62%), started HAART in the period 1997–1999 (46%), had not experienced pre-HAART use of NRTIs (71%) or previous virological

failures (86%) and was currently on an NNRTI regimen (37%) or on a boosted-PI regimen (38%). Most (86%) patients had never interrupted ART, and had CD4 cell counts <200 cells/μL at the start of HAART (52%) and >350 cells/μL at time-zero (76%). The median time since start of HAART at time-zero was 2.7 years (IQR 1.3–4.6), and

the median time from time-zero to the subsequent VL measurement was 94 days (IQR 73–119). Of the 1632 patients, 346 (21.2%) experienced at least one VL rebound, with 376 rebound events overall in the HAS1 15 660 DCVL episodes (2.40%). Factors found to be associated with low drug coverage were: having started HAART in earlier years (before 2000) (P<0.0001), having experienced previous virological failures (P=0.04) and at least two treatment interruptions (P=0.02), currently being on an unboosted PI or NRTI-only regimen (P<0.0001), time-zero in 2002–2003 compared with 2006–2007 (P<0.0001), having a CD4 cell count at the start of HAART that was missing or >350 cells/μL (P=0.02) and having a CD4 cell count at time-zero <200 cells/μL (P<0.0001). The drug coverage was 100% for 32% of episodes and 95.1–99.9% for 16%. At the other extreme, it was below 60% for 10% of episodes. The risk of rebound was 2.13% in those with 95–99% coverage (i.e. those who had 95–99% drug coverage in the preceding 6-month period had a 2.13% chance of a detectable VL >200 copies/mL at the time of their next VL measurement), compared with 1.