Criteria have been proposed for genetic testing of breast and ovarian cancer susceptibility genes BRCA1 and BRCA2. Using simulations, this study evaluates the efficiency (sensitivity, positive predictive value [PPV] and specificity) of the various criteria used in France. The efficiency of the criteria published in 1998, which are largely used, is not optimal. We show that some extensions of these criteria provide an increase in sensitivity with a low decrease in specificity and PPV. The study shows that scoring systems (Manchester, Eisinger) have similar efficiency that may be improved. In this aim, we propose a new scoring system that takes into account unaffected individuals and kinship coefficients between family members. This system increases sensitivity without affecting PPV and specificity. Finally, we propose a two-step procedure with a large screening by the physician for recommending genetic counselling, followed by a more stringent selection by the geneticist for prescribing genetic testing. This procedure would result in an increase of genetic counselling activity but would allow the identification of almost 80% of mutation carriers among affected individuals, with a mutation detection rate of 15% and a specificity of 88%.