Development and validation of a nomogram (APGRC) to predict the presence of germline DNA damage repair pathogenic variants in Asian patients with prostate cancer

Dear Editor, Prostate cancer (PCa) is a prominent global cancer among men, estimated at 1.3 million diagnoses and 359 thousand deaths in 2018,1 with a strong hereditary component. DNA damage repair (DDR) genes play a vital role in maintaining genetic stability and germline mutations in them can result in high-risk biological behaviours2 and influence the response to targeted therapies like platinum treatments, poly-ADP ribose polymerase (PARP) inhibitors, and pembrolizumab.3 Identifying germline DDR pathogenic variants (PVs) carriers in PCa patients has significant implications for personalised treatment, risk reduction and familial testing. However, current genetic testing guidelines are primarily based on Western populations, with a lack of representation from Asians. Thus, we aimed to develop the APGRC (Asian PCa Germline Risk Calculator), a risk assessment nomogram for predicting the probability of carrying germline PVs in 14 PCa predisposition DDR genes. The development cohort included 2,052 unselected PCa patients from four Chinese cancer centers. Stepwise logistic regression determined predictors for the development of the APGRC nomogram, validated in an independent cohort of 743 patients (347 Singaporean patients and 396 Japanese patients4) (Figure S1). Relevant details of model development and validation were provided in Supporting information. The characteristics of the study participants were summarised in Table S1. The Japan cohort (N = 549) exhibited the most aggressive characteristics, with 406 (74.5%) participants having metastatic PCa. The Singapore cohort (N = 920) demonstrated predominantly early features, with 583 (63.5%) participants having localized low- or intermediate-risk PCa, while the development cohort exhibited an intermediate disease profile (Table S1). In total, 162 (7.9%), 27 (2.9%) and 35 (6.4%) participants in the development cohort, Singapore cohort and Japan cohort, respectively, carried germline PVs in the 14 PCa predisposition DDR genes (Table 1, Figure S2). Germline pathogenic/likely PVs were detailed in Table S2. The results of univariate logistic analysis indicated a number of predictors of germline DDR PVs. The carriers were found to be more likely to have a younger age at diagnosis (63.5 vs. 67.0 years old, P < .001), higher prostate specific antigen (PSA) levels (100.0 vs. 61.8 ng/mL, P = 0.006), and higher Gleason scores (52.5% vs. 44.1% in > 8 group, P = .040) compared to non-carriers (Table 1). Additionally, carriers were more likely to have a personal or family history of cancer, excluding PCa (Table 1). The carriers also displayed more aggressive disease characteristics, such as a higher rate of metastatic disease (70.1% vs. 57.6%, P = .002) and later stage of disease (Table 1). Based on the results of the univariate logistic analysis, stepwise logistic regression showed that the most significant predictors of the presence of germline DDR PVs were age at diagnosis (OR .95 [95% CI .93 to .98]; P < .001), personal history of Lynch syndrome-related cancers (OR 4.63 [95% CI 2.03–9.70]; P < .001), PSA at diagnosis (log, OR 1.34 [95% CI 1.02–1.76]; P = .037), stage (Regional/Metastatic vs. Localised, OR 2.11 [95% CI 1.35–3.39]; P = .001), family history of breast, pancreatic, or ovarian cancers in first-degree relatives (OR 18.66 [95% CI 8.24–34.12]; P < .001), and family history of Lynch syndrome-related cancers in first-, second-, or third- degree relatives (OR 4.12 [95% CI 2.39–6.89]; P < .001) (Table 2). A nomogram prediction model, named APGRC, was developed based on the results of the multivariable stepwise logistic regression (Figure 1). The performance of APGRC was compared to current guidelines5-10 (Table S3) in the development cohort, and it was found to have the highest area under the curve (AUC) value (.706 [95% CI .664–.748]), while other guidelines had lower AUC values ranging from .518 to .590 (Figure 2A, Table S4). The APGRC prediction model exhibited high concordance with actual probabilities, as demonstrated by the results of the predicted risk analysis (Figure 2B). Furthermore, the well-calibrated nature of the model was verified through the Hosmer–Lemeshow test (P = .706). The decision curve analysis (DCA) of the development cohort revealed that using APGRC to identify patients for germline testing above a 5% probability threshold resulted in a favourable clinical impact (Figure 2C). The reduction rate in germline genetic testing and the corresponding missing rate of germline DDR PV carriers across varying threshold probabilities determined by APGRC were summarised in Table S5. Notably, at a 4.0% threshold, APGRC achieved a reduction of 11.3% (231/2 152) in germline genetic testing, missing only 1.2% (2/162) of carriers in the development cohort. Similarly, at a 3.0% threshold, APGRC exhibited a reduction of 14.1% (105/743) in germline genetic testing, albeit with a slightly higher missing rate of 4.9% (2/41) among germline DDR PV carriers in the validation cohort. Furthermore, APGRC has demonstrated commendable validation outcomes within both the early-stage disease population, as exemplified by the Singapore cohort (.654 [95% CI.517–.792]), and the late-stage disease population, as represented by the Japanese cohort (.653 [95% CI.518–.789]) (Figure 2D). The similar AUC values (.706 [95% CI .664–.748] vs. .648 [95% CI .553–.743]) in the combined validation cohort and the development cohort (Figure 2D) underscore APGRC's consistent discriminatory capacity. It is worth mentioning that none of the existing guidelines take age at diagnosis into account as a criterion (Table S3). To evaluate the impact of age at diagnosis on the performance of APGRC, we excluded this factor from the model. The results showed a decline in APGRC's performance, with an AUC of .666 in predicting the presence of germline DDR PVs in the developing cohort (Figure S3B, highlighting age at diagnosis as a crucial predictive factor for the presence of germline DDR PVs. In conclusion, the APGRC model represents a valuable tool in the prediction of the presence of germline DDR PVs in Asian patients with PCa. This is the first study to integrate various important clinical variables in a comprehensive and tailored manner for the Asian population. By doing so, APGRC can support clinical decision-making by physicians and patients and pave the way to improve precision oncology in an underrepresented patient population. We appreciate the support provided by the Medical Science Data Center in Shanghai Medical College of Fudan University. Melvin L.K. Chua reports personal fees from Astellas, Janssen, Bayer, Pfizer, MSD, Varian, IQVIA, Telix Pharmaceuticals, AstraZeneca, personal fees and non-financial support from BeiGene, non-financial support from Decipher Biosciences, non-financial support from MedLever, consults for immunoSCAPE Inc., and is a co-inventor of the patent of a High Sensitivity Lateral Flow Immunoassay For Detection of Analyte in Sample (10202107837T), Singapore and serves on the Board of Directors of Digital Life Line Pte Ltd that owns the licensing agreement of the patent, outside the submitted work. The other authors declare that they have no competing interests. the National Natural Science Foundation of China (82172621, YZ); Chinese Anti-Cancer Association-Hengrui PARP Inhibitor Tumor Research Fund (Phase I, YZ); Bethune Urology Tumor Special Research Fund (mnzl202004, YZ); Program of Shanghai Academic Research Leader (23XD1420600, YZ); Shanghai Medical Innovation Research Special Project (21Y11904300, YZ); the General Program of Beijing Xisike Clinical Oncology Research Foundation (Y-2019AZMS-0012, YZ); the National Medical Research Council Singapore Clinician Scientist Award (NMRC/CSA-INV/0027/2018, CSAINV20nov-0021, MC); the Duke-NUS Oncology Academic Program Goh Foundation Proton Research Programme (MC); NCCS Cancer Fund (MC); the Kua Hong Pak Head and Neck Cancer Research Programme (MC) All data generated or analysed during this study are included in this published article and its supplementary files. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.