Abstract
The analysis by Knip et al (Nov. 11 issue),1 which suggests that dietary intervention through infant weaning to hydrolyzed formula has a long-lasting effect on markers of beta-cell autoimmunity, is flawed in that it has overestimated the evidence of benefit because of failure to adequately use the intention-to-treat principle.2 First, children from whom follow-up samples were not obtained were excluded from the primary analysis of seroconversion to autoantibody positivity. Of the children who were excluded, diabetes (and presumably autoantibody positivity before clinical disease) developed in three in the treatment group versus none in the control group. Correct inclusion of these 3 children in an intention-to-treat protocol would minimize evidence of treatment effect, with the development of beta-cell autoimmunity in 11 of 113 infants randomly assigned to hydrolyzed formula (9.7%) and in 17 of 109 infants assigned to standard formula (15.6%). A second breach of the principle occurred by adjusting for the duration of exposure to the study formula. In so doing, the authors were able to invalidly “improve” the P value regarding the hazard of development of two or more markers of beta-cell autoimmunity between treatment groups from 0.12 to 0.07.