Can the Gail model increase the predictive value of a positive mammogram in a European population screening setting? Results from a Spanish cohort.

AIMS OF THE STUDY

The Gail Model (GM) is the most well-known model to assess the individual risk of breast cancer (BC). Although its discriminatory accuracy is low in the clinical context, its usefulness in the screening setting is not well known. The aim of this study is to assess the utility of the GM in a European screening program.

METHODS

Retrospective cohort study of 2200 reassessed women with information on the GM available in a BC screening program in Barcelona, Spain. The 5 year-risk of BC applying the GM right after the screening mammogram was compared first with the actual woman's risk of BC in the same screening round and second with the BC risk during the next 5 years.

RESULTS

The curves of BC Gail risk overlapped for women with and without BC, both in the same screening episode as well as 5 years afterward. Overall sensitivity and specificity in the same screening episode were 22.3 and 86.5%, respectively, and 46.2 and 72.1% 5 years afterward. ROC curves were barely over the diagonal and the concordance statistics were 0.59 and 0.61, respectively.

CONCLUSION

The GM has very low accuracy among women with a positive mammogram result, predicting BC both in the concomitant episode and 5 years later. Our results do not encourage the use of the GM in the screening context to aid the referral decision or the type of procedures after a positive mammogram or to identify women at high risk among those with a false-positive outcome.