Research ArticlesMedicine

Vaginal bacteria modify HIV tenofovir microbicide efficacy in African women

See allHide authors and affiliations

Science  02 Jun 2017:
Vol. 356, Issue 6341, pp. 938-945
DOI: 10.1126/science.aai9383

eLetters is an online forum for ongoing peer review. Submission of eLetters are open to all. eLetters are not edited, proofread, or indexed.  Please read our Terms of Service before submitting your own eLetter.

Compose eLetter

Plain text

  • Plain text
    No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Author Information
First or given name, e.g. 'Peter'.
Your last, or family, name, e.g. 'MacMoody'.
Your email address, e.g.
Your role and/or occupation, e.g. 'Orthopedic Surgeon'.
Your organization or institution (if applicable), e.g. 'Royal Free Hospital'.
Statement of Competing Interests
Publication Date - String

This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.

Enter the characters shown in the image.

Vertical Tabs

  • RE: Statistical Evaluation of Effect Modification
    • Adam D. Burgener, National HIV and Retrovirology Labs and University of Manitoba, Winnipeg, Canada
    • Other Contributors:
      • Anneke Grobler, Clinical Epidemiology and Biostatistics Unit, Murdoch Children’s Research Institute, Melbourne, Australia
      • Nichole R. Klatt, University of Washington, Seattle, USA

    The CAPRISA 004 trial was originally designed to test the efficacy of a tenofovir microbicide gel to prevent HIV acquisition in women. Thus, it is expected there will be power limitations to examine other interactions rather than main effects, including that of the vaginal microbiome which was not originally considered in the trial. Dr. Moulton emphasizes an important point regarding interaction analysis. However, given the combination of efficacy differences between microbiome groups and supporting mechanistic data, we are confident in the interpretation that the vaginal microbiome has a substantial effect on tenofovir efficacy and HIV acquisition.

    We comprehensively discussed many caveats of this study in the manuscript, including several interaction analyses and potential confounders, and were upfront about small sample size for sub-analyses. While we did not include a formal interaction analysis of the subgroups, agreeing with Dr. Moulton’s analysis, when testing the hypothesis examining interaction effects between tenofovir gel assignment and Lactobacillus-dominant vs. non-Lactobacillus dominant microbiomes, provides an interaction p-value of 0.17. Considering those with high reported adherence (>50%) in each group, the p-value for interaction is 0.15. However, given this is a secondary analysis, the test for interaction has low power with an alpha range of 0.10 to 0.20, within a range utilized for detecting interactions other than those of the main e...

    Show More
    Competing Interests: None declared.
  • RE: Statistical Evaluation of Effect Modification
    • Lawrence H. Moulton, Professor, Depts. of International Health and (joint) Biostatistics, Johns Hopkins Bloomberg School of Public Health

    In their Research Article “Vaginal bacteria modify HIV tenofovir microbicide efficacy in African women” (2 June, p. 938), N.R. Klatt and colleagues present a great deal of evidence for the purported effect modification mentioned in the title. Nonetheless, they do not present a statistical evaluation to support their thesis. Although the reported point estimates of tenofovir efficacy for HIV prevention are 61% and 18% among Lactobacillus-dominant and non-Lactobacillus-dominant women, respectively, these estimates are rather variable, being based on only 31 infections in each study arm. A statistical test of interaction yields a p-value of 0.17 with respect to the null hypothesis of no difference between the two subgroups. Indeed, the two 95% confidence intervals for the rate ratios encompass each other’s point estimate—even if one subgroup’s estimate were based on a million infections, the variability of the other would render the two efficacies not statistically significantly different at the 0.05 level. It may be best to modify the term “modify” in the article title by the word “may.”

    It may be that the authors were misled by the fact one efficacy was statistically significant while the other was not. It is a common misinterpretation of Neyman-Pearson testing to conclude from this that the efficacies are statistically different from each other. Because there is much less power to detect an interaction than a main effect, some researchers test interactions at the α=...

    Show More
    Competing Interests: None declared.