Atazanavir plus ritonavir or efavirenz as part of a 3-drug regimen for initial treatment of HIV-1.

TO THE EDITOR: I read the recent article by Daar and colleagues (1) with great interest. A portion of the results of this important study has already been published (2), with the more recent results still raising comments and questions. As written in the article, 2 regimens will be considered equivalent if a 2-sided CI for the hazard ratio (HR) of virologic failure (VF) is entirely within the range of 0.71 to 1.40. The upper boundary of the HR of atazanavir plus ritonavir or efavirenz was 1.56 and 1.46 with abacavir– lamivudine or tenofovir disoproxil fumarate–emtricitabine, respectively. It seems that the main conclusion of the study is that the results of the trial are inconclusive, and the equivalence could not be demonstrated for the 2 paired comparisons (3). However, the Conclusion section in the abstract indicates that atazanavir plus ritonavir and efavirenz has similar antiviral activity. This could be very confusing after an attempt over several years to provide standardized guidelines, procedures, and design for equivalence and noninferiority trials (3, 4). Although the authors mentioned that criteria for equivalence were not met, a reader could understand that the trial has moved from an equivalence design to an unknown “similarity” design. Table 1 in the article indicates that a larger proportion of patients randomized in the atazanavir plus ritonavir groups were in the high viral load stratum at baseline compared with the efavirenz groups (27.7% vs. 22.5%; P 0.01). Considering that patients in the high viral load stratum had a higher risk for VF, the unbalanced viral load stratum would favor the efavirenz groups and could explain the slightly higher HR for atazanavir plus ritonavir observed in the article by Daar and colleagues (1). The large difference in the number of patients in the high viral load stratum at screening or baseline was already noted in the previous publication (2). This point is confusing and suggests a post hoc analysis based on the viral load at baseline. As pointed out in the Discussion, at week 96 results showed a lower rate of VF (11% to 17%) than expected (32%). It is not clear how this affects the probability of declaring equivalence because usually larger rates of failure provide larger CIs. In addition, the power of the study is higher than expected, although a posteriori power computation is usually improper (5), which increases the probability of accepting equivalence in an equivalence trial.