Laboratory results were only returned to clinicians caring for CDM participants if there was a grade 4 toxicity or the treating physician had specifically requested them for clinical reasons: lymphocyte subset results were not returned for CDM participants. All causes of death and reported WHO stage 4 events were reviewed by an Endpoint Review DAPT purchase Committee (ERC) against criteria pre-specified in the protocol, blinded to treatment allocation and monitoring strategy; SAEs were also reviewed. The ERC adjudicated each WHO 4 event as ‘new’ (never occurred previously) or as a separate ‘recurrence’ of a previously resolved event.

Plasma HIV-1 RNA was retrospectively assayed on stored samples at 0, 4, 12, 24 and 48 weeks using the Roche Amplicor v1.5 assay (Roche Diagnostics, Basel, Switzerland) for baseline samples (lower limit NU7441 in vitro of detection 400 HIV-1 RNA copies/mL), and the Roche ultrasensitive assay subsequently (50 copies/mL). Exploratory analyses of virological, immunological and clinical (efficacy) outcomes to 48 weeks are reported. Results beyond 48 weeks are not included, because, as CD4 increases were greater in the nevirapine group (see ‘Results’), a greater proportion in the nevirapine group were randomized to STI (70; 23%)

or CT (47; 16%) than in the abacavir group (36; 12% and 53; 18%, respectively), making comparisons beyond 48 weeks complex. Clinical efficacy outcomes and subgroups considered here were those previously used for the final STI/CT analysis [6]. Trial entry

was the date of randomization. The log rank test and Cox proportional hazards models were used to compare the randomized groups for the time-to-event outcomes, censoring at 48 weeks after trial entry. All comparisons between the groups were as randomized (intent-to-treat), except that toxicity analyses were restricted to time on any ART plus 30 days. Comparisons of markers were based on observed values. A ‘missing=failure’ imputation was not used because this assumes all reasons for missing values 17-DMAG (Alvespimycin) HCl are failure-related: this is only one of several crude sensitivity analyses and is not necessarily conservative depending on the reasons (given in Table 2 footnote). Baseline values were those recorded nearest to but before and within 6 weeks of randomization; subsequently, the closest measurement to the scheduled assessment week within equally spaced windows was used. Changes in log10 HIV RNA including values below the lower limit of detection were estimated using normal interval regression [9]. All P-values reported are two-sided. All analyses presented were repeated with and without stratification for baseline CD4 cell count, centre and randomization to CDM vs. LCM to confirm that there were no major imbalances affecting results. stata 10.