Background: SNP risk information can potentially improve the accuracy of breast cancer risk prediction. We aim to review and assess the performance of SNP-enhanced risk prediction models.
Methods: Studies that reported area under the ROC curve (AUC) and/or net reclassification improvement (NRI) for both traditional and SNP-enhanced risk models were identified. Meta-analyses were conducted to compare across all models and within similar baseline risk models.
Results: Twenty-six of 406 studies were included. Pooled estimate of AUC improvement is 0.044 [95% confidence interval (CI), 0.038-0.049] for all 38 models, while estimates by baseline models ranged from 0.033 (95% CI, 0.025-0.041) for BCRAT to 0.053 (95% CI, 0.018-0.087) for partial BCRAT. There was no observable trend between AUC improvement and number of SNPs. One study found that the NRI was significantly larger when only intermediate-risk women were included. Two other studies showed that majority of the risk reclassification occurred in intermediate-risk women.
Conclusions: Addition of SNP risk information may be more beneficial for women with intermediate risk.
Impact: Screening could be a two-step process where a questionnaire is first used to identify intermediate-risk individuals, followed by SNP testing for these women only.
©2018 American Association for Cancer Research.