To establish an efficient, reliable and easy to apply risk assessment tool to select families with breast and/or ovarian cancer patients for BRCA mutation testing, using available probability models. In a retrospective study of 263 families with breast and/or ovarian cancer patients, the utility of the Frank (Myriad), Gilpin (family history assessment tool) and Evans (Manchester) model was analysed, to select 49 BRCA mutation-positive families. For various cutoff levels and combinations, the sensitivity and specificity were calculated and compared. The best combinations were subsequently validated in additional sets of families. Comparable sensitivity and specificity were obtained with the Gilpin and Evans models. They appeared to be complementary to the Frank model. To obtain an optimal sensitivity, five 'additional criteria' were introduced that are specific for the selection of small or uninformative families. The optimal selection is made by the combination 'Frank >or=16% or Evans2 >or=12 or one of five additional criteria'. The efficiency of the selection of families for mutation testing of BRCA1 and BRCA2 can be optimised by using a combination of available easy to apply risk assessment models.