Predictive accuracy of the breast cancer genetic risk model based on eight common genetic variants: The BACkSIDE study

Breast cancer (BC) development is caused by the interaction of environmental and genetic factors. At least 90 susceptible genetic variants with different population penetration and incidence have been associated with BC. This paper therefore analysed the individual discrimination power of 8 low penetrant common genetic variants and calculated the predictive accuracy of the genetic risk model. The study enrolled 171 women with developed breast cancer (57.06 +/- 11.60 years) and 146 control subjects (50.24 +/- 10.69 years). The genotyping was performed by high resolution melting method (HRM) and confirmed by Sanger sequencing, and the Random Forest algorithm provided the ROC curve with AUC values. Significant association with BC was confirmed in 2 SNPs: rs2981582 FGFR2 and rs889312 MAP3K1, and the odds ratios of homozygotes with two risk alleles in both SNP's were higher than in heterozygotes with one mutant allele, as follows: FGFR2 TT: 1.953 (95% CI 1.014-3.834, p = 0.049), CT 1.771 (95% CI 1.088-2.899, p = 0.026) and MAP3K1 CC 2.894 (95% CI 1.028-9.566, p = 0.048), AC 1.760 (95% CI 1.108-2.813, p = 0.019). FGFR2 had the best discrimination ability, followed by MAP3K1 and CASP8. Discriminative accuracy of the genetic risk model distinguishing the breast cancer patients and controls explained by AUC was 0.728, with 70.6% sensitivity and 65.1% specificity. Our study results therefore confirmed polygenic breast cancer inheritance with important involvement of FGFR2, MAP3K1, LSP1 and CASP8 gene variants.