We appreciate the thoughtful comments by Modjarrad  on our recent study, ‘Albendazole treatment of HIV-1 and helminth co-infection: a randomized, double-blind, placebo-controlled trial’ . Modjarrad  highlights several important issues that merit clarification, specifically concerning the analytical plan, merits of randomization, and species-specific outcome evaluations.
As outlined in our paper, we conducted a traditional randomized clinical trial. Individuals were enrolled at baseline, randomized to the intervention or placebo, and followed up at a prespecified standardized single time-point. CD4 cell counts and HIV-1 RNA levels at follow-up were compared between the two trial arms (albendazole and placebo) after controlling for baseline values. The mean initial CD4 cell counts and viral load by species are provided in the table of the manuscript by Walson et al. . This was not a cross-sectional observational study, and the trial was designed with a single follow-up time, so the methods of analysis mentioned by Modjarrad  would not be applicable.
Modjarrad  erroneously suggests that the analyses were conducted a posteriori. As noted in the methods section of our paper, the statistical analysis plan was determined a priori. The decision to include a modified intent-to-treat analysis of follow-up CD4 cell counts and HIV-1 RNA levels as well as the decision to stratify results by helminth species were a-priori decisions. In fact, a post-hoc analysis of changes in CD4 cell count and HIV-1 RNA between the randomization arms suggests a significant benefit of deworming on CD4 cell counts in all helminth-infected individuals (decrease in CD4 cell count of 72 cells/μl in the placebo arm compared with a decrease of 26 cells/μl in the treatment arm; P = 0.043) as well as in the Ascaris lumbricoides infected cohort (decrease in CD4 cell count of 99 cells/μl in the placebo arm compared with an increase of 9 cells/μl in the treatment arm; P = 0.005). Because we had not elected a priori to include ‘change in CD4’ or ‘change in RNA’ as an outcome, we did not include these findings in our paper and presented only data derived from the a-priori statistical analysis plan.
Modjarrad  seems to confuse issues of randomized trial design with analyses of multiple outcomes. It is important to recognize the inherent superiority of the randomized controlled trial over observational studies. Prior observational studies of helminth infection in HIV-1 infected individuals have been limited by potential confounding. These studies have compared helminth-infected and helminth-uninfected individuals or have included historical controls. Significant differences exist at baseline in rates of disease progression, immune activation, and natural variation in the rate of CD4 cell count decline and plasma HIV-1 RNA levels between these comparison groups. The strength of the randomized controlled trial study design is precisely that differences due to natural variation are randomly distributed between the comparison groups. If our analysis resulted in a type II error (not finding an effect when one truly exists), this would most likely be due to limitations in sample size (statistical power) and not due to inherent differences between the groups.
Finally, as we noted in our conclusions, we concur with Modjarrad  that this study raises additional questions concerning the species-specific nature of the effects observed and that additional well designed randomized trials should be conducted to answer these questions.
1. Modjarrad K. Which helminth coinfections really affect HIV disease progression? AIDS
2. Walson JL, Otieno PA, Mbuchi M, Richardson BA, Lohman-Payne B, et al
. Albendazole treatment of HIV-1 and helminth co-infection: a randomized, double blind, placebo-controlled trial. AIDS 2008; 22:1601–1609.