[Statistical analysis in the evaluation of controlled diffusion of iodine in the water in developing countries]

This report describes the methodology and findings of a novel statistical technique for evaluation of the efficacy of the Rhodifuse Iode system in prevention of endemic goiter in Mali. The system involves continuous release of iodine in ground water used for drinking. Study was carried out in four villages for one year. The iodine release system was used in three villages. The fourth village served as the control. The incidence of goiter graded using the criteria of the WHO was assessed in each village according to sex. Statistical analysis consisted of correlating goiter grade with four predictors, i.e., village, sex, iodine release, and time. Since goiter grade is a dependent variable, its law of probability was modeled in function of the predictors. The Cat Mot procedure included in the SAS software package allowed both definition of the law of probability of grade of goiter and its transformation in function of predictor. The generalized linear model was obtained by either the generalized least square method or greatest likelihood method. The Proc Catmod procedure was then used to generalize analysis of variance in case of a nominal or ordinal, binary or polytomic response. The results of this novel statistical technique suggested that the Rhodifuse Iode system was effective.