Can multiple SNP testing in BRCA2 and BRCA1 female carriers be used to improve risk prediction models in conjunction with clinical assessment?

Table 3 Average (st. dev.) 1 - c-index performance results of cox proportional hazards and random survival forest models as estimated by collating out-of-bag distributions from 100 bootstrap runs

Model	Average (st.dev.) 1 - c-index
	BRCA1	BRCA2
(i) GPS	0.468 (0.037)*	0.417 (0.021)*
(ii) GPS + Clin./Demogr.	0.221 (0.019)	0.215 (0.018)
(iii) SNPs	0.504 (0.026)*	0.469 (0.028)*
(iv) SNPs + Clin./Demogr.	0.238 (0.019)	0.241 (0.019)
(v) Clin./Demogr.	0.222 (0.019)	0.217 (0.018)
(vi) all variables (by AIC)	0.236 (0.018)	0.232 (0.019)
Random survival forest (all variables)	0.243 (0.019)	0.230 (0.019)

*p > 0.0001 as compared to model (ii) by an adjusted t-test. Values in bold show the best performance.

ISSN: 1472-6947