Re: Randomized Double-Blind Factorial Trial of Three Treatments to Reduce the Prevalence of Precancerous Gastric Lesions

Extract

We congratulate You et al. ( 1 ) for the successful completion of their chemoprevention trial. Their results have clear implications for gastric cancer prevention. As emphasized by the authors, all available reports conclude that curing Helicobacter pylori infection is a promising strategy to control gastric cancer. Previously, we supplied the authors with raw data from our Colombian chemoprevention trial at 6 years of follow-up ( 2 , 3 ) . Because our published conclusions, based on the same data, differed considerably from theirs ( 1 ) , we would like to comment on the discrepant interpretations of our results.

Although an analysis based on main effects is customary for a factorial design, it assumes that there are no negative interactions among the interventions given (i.e., it assumes that all the interventions would have an effect in the same direction). Our analysis at 6 years showed that there was a statistically significant negative interaction between H. pylori treatment and beta-carotene supplementation. Basing the analysis on the main effects ( 1 ) in this case is not appropriate because some subjects who received H. pylori treatment showed improvement, but those who received H. pylori treatment and beta-carotene supplementation did worse than those on placebo. Only the full factorial analysis accomplished through polytomous and binary multivariate logistic regression models using data collected over time from the same individual demonstrates that after taking into account the negative effect of beta-carotene supplementation, patients who underwent anti– H. pylori therapy have statistically significant more regression of precancerous lesions than those on placebo.