The three main components of long-term efficacy for a combination of antiretrovirals are: (i) the strength of the antiviral effect, (ii) toxicity profile and (iii) patient acceptability of the regimen. Intent-to-treat (ITT) analysis, where discontinuations and switches are considered failures [ITT, switch equals failure (ITT/S = F)], is a regulatory standard for analysing the efficacy of antiretrovirals. A review of all clinical trials published in FDA product labels was conducted, including all clinical trials of boosted protease inhibitor- or nucleoside reverse transcriptase inhibitor-based highly active antiretroviral therapy in treatment-naive patients, and all clinical trials of antiretrovirals in treatment-experienced patients. Clinical trials where the results are presented in the standard ITT/S = F method were included. For randomized clinical trials in treatment-naive patients, the majority of treatment discontinuations have been either for toxicity (32%) or patient refusal of treatment (41%), with only 27% of failure endpoints for virological reasons among recent clinical trials in naive patients. Therefore, there is the potential for the results from ITT/S = F analysis to be driven by non-virological endpoints - a new treatment can be classified as 'more efficacious' than control owing to fewer discontinuations due to adverse events or patient preference. In order to understand the intrinsic potency of the antiretroviral regimen under study, ITT analysis needs to be supplemented by standardized as-treated analyses, excluding withdrawals for toxicity or other reasons. To evaluate the efficacy of a treatment strategy or sequential treatment regimens, the 'ITT, switch included' (ITT/SI) method: where changes from the initial randomized treatment are not classified as treatment failure - can be used. However, interpretation of clinical trials using ITT/SI analysis is difficult and depends on the frequency of treatment switching in the different arms of a trial. Conclusions on efficacy from clinical trials can depend on the primary analysis used; most commonly, treatments could be significantly different by ITT/S=F analysis, but then interpreted as equivalent using the ITT/SI or as-treated methods.