Prediction of anthracycline-related cardiomyopathy in long-term childhood cancer survivors: The power of survivor-specific polygenic risk score—A COG-ALTE03N1 study.

10005 Background: Cardiomyopathy is a major dose-limiting toxicity of anthracyclines in the treatment of childhood cancer. Inter-patient variability in cardiomyopathy risk suggests an underlying genetic susceptibility, providing an opportunity to identify survivors at high risk and enable personalized interventions. We evaluated whether addition of a curated survivor-specific PRS for anthracycline-related cardiomyopathy (PRS-S) or general population PRSs for heart failure (PRS-G) to a clinical model improved cardiomyopathy risk classification in CCS. Methods: In a matched case-control study of CCS (COG-ALTE03N1: matched on cancer diagnosis and year of diagnosis; controls followed at least as long as cases), anthracycline-exposed non-Hispanic White CCS were randomly split (80/20) into training (cases: n=106; controls: n=180) and test (cases: n=27; controls: n=44) sets. From >100 published genetic variants associated with anthracycline-related cardiomyopathy, we selected a prioritized list of 8 independently validated variants with demonstrated biologic plausibility (reduced drug efflux [ ABCC2 ] , altered drug metabolism [ SLC22A17, SLC28A3, UGT1A6 ] , reduced antioxidant capacity [ HAS3 ] , impaired TOP2B-mediated DNA damage response [ RARG ] , alternative splicing of TNNT2 , reducing cardiac pumping efficiency [ CELF4 ], and cardiac profibrotic pathway [ ROBO2] ) to construct a standardized weighted survivor-specific PRS (PRS-S). Area under the receiver operating characteristic curve (AUC) was used to compare the performance of a clinical model (age at cancer diagnosis, sex, anthracycline dose, chest radiation (Y/N)) v combined models incorporating either PRS-S or PRS-Gs. All models included PRS × anthracycline dose interaction terms and adjusted for matching variables. Risk scores derived from training model were applied to the test set for risk classification. Results: Median age at cancer diagnosis was 7.1y for cases and 8.6y for controls. Cases received higher anthracycline doses (median: 330 v 240 mg/m 2 ) and were more likely to have received chest radiation (24.8% v 19.2%). In the test set, PRS-S combined model significantly outperformed the clinical model (AUC clin+PRS8 =0.70, 95%CI=0.57-0.82 v AUC clin =0.63, 95%CI=0.50-0.76, p=0.04); 85% of all cases were classified as high risk and the adjusted negative predictive value was 95.1%. Importantly, none of the 11 PRS-G combined models (AUCs: 0.60-0.65) out-performed the clinical model. Conclusions: Integrating a survivor-specific PRS derived from a parsimonious set of biologically plausible genetic variants with readily-available clinical factors informs cardiomyopathy risk classification in CCS and provides an opportunity to develop personalized screening recommendations and behavioral/therapeutic interventions.