Sample selection and outcome evaluation in primary prevention

The practical implementation of evidence-based prevention programmes is a major challenge that requires a sound scientific basis. This paper refers to statistical aspects that are relevant in the planning phase of a primary prevention programme. First, a stepwise approach using graphical models and CHAID decision trees is presented to select subgroups of the population suitable for targeted intervention. Second, adequate statistical methods for sample size determination and for outcome evaluation to be used in the planning phase of a study are described. The IDEFICS project is presented as an example to illustrate the way leading from the aetiology of a disease to the design of a primary prevention programme, its implementation and its evaluation. The practical implications of the presented approaches and their methodological caveats are discussed.