Predictors of germline DNA damage repair gene mutations (gDDRm) in patients (pts) with metastatic castration-resistant prostate cancer (CRPC).

Abstract

159 Background: gDDRm are present in 6-12% of pts with CRPC and have implications for prognosis, treatment outcomes and familial screening. This study's aim was to identify clinical/pathologic characteristics associated with a higher risk of a pt harboring a gDDRm. Methods: 631 consecutive pts with metastatic prostate cancer were screened for germline mutations in 22 DNA repair genes (including ATM, BRCA1/2, MSH2/6, FANC/ERCC members). Clinical/pathologic characteristics (age at diagnosis, visceral metastases (mets) at time of CRPC, Gleason score, intraductal/cribriform histology (I/C) at diagnosis, family history (FHx) of prostate, breast, ovarian, or pancreatic cancer, time from androgen deprivation therapy initiation to CRPC) were compared between gDDRm+ and gDDRm- cases. A multivariate logistic model of gDDRm status was constructed using purposeful selection of covariates where clinical judgement was employed in addition to statistical significance. A weighted scoring system was created by multiplying each risk factor (RF) by its estimated β coefficient. A pt’s total score was calculated by the sum of each weighted RF. Results: gDDRm+ were identified in 38/631 pts (6.0%, BRCA2 = 28, ATM = 4, BRCA1 = 2, PALB2 = 2, MSH2 = 1, ERCC3 = 1). 29 gDDRm+ and 128 gDDRm- cases with complete clinical/pathologic information were included in the model. Multivariate modeling revealed that visceral mets, FHx and I/C were statistically significantly associated with gDDRm+; age ≤ 65 was included in the model because it was considered clinically relevant (Table). Based on the model results, the weights of each RF are: visceral mets (2), FHx (1), I/C (2), age ≤ 65 (1). Using a total score cut-off ≥ 1, the model achieves a sensitivity, specificity, positive predictive value, and negative predictive value of 97%, 31%, 24%, and 98%, respectively. Conclusions: In this cohort, visceral mets, FHx and I/C were independently associated with a higher risk of harboring a gDDRm. The predictive model may aid in selecting pts for gDDRm screening. Model validation is required. [Table: see text]