To the Editor—We thank Drs Ramírez and Diaz-Quijano for their comments on our study [1]. We agree with their statements about the importance of reporting the components of composite endpoints. As they have noted, our study reported 10 (2.0%) and 6 (1.2%) deaths among the 500 participants per arm randomized to continued co-trimoxazole or chloroquine prophylaxis, respectively. This imbalance comes despite the analysis of the composite endpoint of time to first World Health Organization (WHO) human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS) stage 3 or 4 event or death suggesting a reduction in disease burden among participants receiving co-trimoxazole relative to participants receiving chloroquine (9.6% vs 12.4% of participants with the composite primary endpoint, respectively). The very low mortality rate, and the difficulty in detecting statistically significant differences when events are so rare, highlight the critical need for carefully crafted composite endpoints.
A limitation of our study is that the analyses of all-cause mortality and the composite endpoints are not directly comparable. As noted in the Methods section, when participants experienced a WHO HIV/AIDS stage 3 or 4 event, they discontinued their assigned treatment and restarted co-trimoxazole prophylaxis for the remainder of the study. Hence, the analysis of all-cause mortality includes roughly 10% of chloroquine participants who experienced a stage 3 or 4 event after randomization and then crossed over to co-trimoxazole prophylaxis following their clinical event. Because of the ethical imperative to restart standard-of-care therapy, we cannot know the effect of chloroquine prophylaxis on all-cause mortality had these sickest participants remained on chloroquine.
Setting aside the lack of comparability of the analyses of mortality and the composite primary endpoint, our study suffers a limitation common to many other studies that are powered for a composite endpoint, but not for mortality itself. The inference on mortality from our study is imprecise, with an estimated risk ratio for death of 1.67 (95% confidence interval, 0.61–4.55). The wide confidence interval shows that these observations are consistent with chloroquine prophylaxis’s increasing, decreasing, or having no effect on overall mortality relative to co-trimoxazole.
We agree with Drs Ramírez and Diaz-Quijano’s caution that trials reporting heterogeneity in the effect of the components of the composite endpoint require critical review; however, we maintain that the inability to directly compare inferences from the mortality and composite endpoint data and the imprecision of the estimated mortality rates provide unreliable evidence of a differential effect on mortality and the incidence of WHO HIV/AIDS stage 3 and 4 events in this population.
Notes
Financial support. This work was supported by a grant from the National Institute of Allergy and Infectious Diseases, U.S. National Institutes of Health U01AI089342 with additional support from K24AI114996 to M. K. L.
Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.