Copyright 2007 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2007
In our previous study,1 data were presented from a randomized, double-blind, placebo-controlled trial indicating that participants treated over 9 months with high selenium yeast demonstrated an increase in serum selenium concentration and an associated suppression of HIV-1 viral load compared with placebo-treated individuals. We appreciate the opportunity to respond to concerns regarding our chosen analytic approach and other issues. All variables used were evaluated for normality and transformed when needed. The intention-to-treat (ITT) approach implemented within a structural equation modeling (SEM) framework used full-information maximum likelihood estimation. Therefore, all randomized participants were included in the analysis, a prerequisite in clinical trial evaluation. The model assessed by the analysis fit the data because the treatment effect on viral load was completely explained by the change in serum selenium concentration. Most clinical trials are unable to measure the blood concentration of the treatment under examination and rely on treatment adherence measures as proxies. These trials, therefore, cannot separate direct from indirect effects. We included a measure of serum selenium concentration obtained on a monthly basis over the intervention period. Therefore, the chosen SEM approach afforded the opportunity to simultaneously test the direct effect of randomized allocation to group on subsequent changes in serum selenium level and then its indirect effect on HIV-1 viral load at treatment month 9, while controlling for pretreatment HIV-1 viral load. In so doing, we have evaluated the treatment effect by using the optimal method to assess the central question, which is whether the increase in selenium concentration by taking a single daily 200-μg selenium dose is effective in influencing HIV-1 disease severity.
Hurwitz BE, Llabre MM. Selenium Effects on HIV RNA and CD4 Cell Counts—Reply. Arch Intern Med. 2007;167(14):1557. doi:10.1001/archinte.167.14.1557