When can HIV clinical trials detect treatment effects on drug resistance?

Methods of sampling patients for resistance testing, and statistical analyses of HIV drug resistance, have not been standardised in HIV clinical trials. We analysed methods of genotyping and rates of treatment-emergent drug resistance from 27 clinical trials identified from a MEDLINE search. Sample size calculations were conducted using NQUERY software, assuming 5% significance level, 80% power and 1:1 randomisation. The percentage of patients with treatment-emergent IAS-USA mutations after 96 weeks ranged from 1.8% to 9.1% for first-line 2NRTI/NNRTI treatments, 0.6% to 6.3% for first-line 2NRTI/PI/r treatments and 0.0% to 2.0% in switch trials of boosted PIs. The prevalence of drug resistance was higher in trials with no screening for drug resistance at baseline, where the HIV RNA cut-off for genotyping was >50 copies/mL, where patients were tested for drug resistance after discontinuation of treatment, and where follow-up times were 96 weeks or longer. HIV clinical trials could be designed to detect differences in the risk of HIV drug resistance between treatments, as an analysis supporting HIV RNA suppression as the primary endpoint. However, this would require a standardised approach, with intent-to-treat analyses, testing of all samples with HIV RNA>50 copies/mL and genotyping after drug discontinuation.