Skip to main content
VA Author Manuscripts logoLink to VA Author Manuscripts
. Author manuscript; available in PMC: 2026 May 19.
Published in final edited form as: Nat Cancer. 2026 Jan 26;7(2):352–367. doi: 10.1038/s43018-025-01103-0

Genomic risk model to implement precision prostate cancer screening in clinical care: the ProGRESS Study

Jason L Vassy 1,2,3,*, Anna M Dornisch 4,*, Roshan Karunamuni 4,5, Michael Gatzen 6, Christopher J Kachulis 6,7, Niall J Lennon 6,7, Charles A Brunette 1,2, Morgan E Danowski 1, Richard L Hauger 8,9, Isla P Garraway 10,11, Adam S Kibel 12, Kyung M Lee 13, Julie A Lynch 13, Kara N Maxwell 14,15, Dmitry Ratner 1, Brent S Rose 4,5, Craig C Teerlink 16,17, George J Xu 4,5, Sean E Hofherr 6, Katherine A Lafferty 6,7, Katie Larkin 6,7, Edyta Malolepsza 6,7, Candace J Patterson 6,7, Diana M Toledo 6,7, Jenny L Donovan 18, Freddie C Hamdy 19, Richard M Martin 20,21, David E Neal 19,22,23, Emma L Turner 20, Ole A Andreassen 24, Anders M Dale 25,26,27,28, Ian G Mills 19, Aswin Abraham 29, Jyotsna Batra 30,31,32,33, Judith Clements 31,32,33, Olivier Cussenot 34,35, Cezary Cybulski 36, Rosalind A Eeles 37,38, Jay H Fowke 39, Eli Marie Grindedal 40, Henrik Grönberg 41, Robert J Hamilton 42,43, Jasmine Lim 44, Yong-Jie Lu 45, Robert J MacInnis 46,47, Christiane Maier 48,49, Lorelei A Mucci 50, Luc Multigner 51, Susan L Neuhausen 52, Sune F Nielsen 53,54, Marie-Élise Parent 55,56, Jong Y Park 57, Gyorgy Petrovics 58,59, Anna Plym 41,50, Azad Razack 44, Barry S Rosenstein 60, Johanna Schleutker 61,62, Karina Dalsgaard Sørensen 63,64, Paul A Townsend 65,66, Ruth C Travis 67, Ana Vega 68,69,70, Catharine ML West 71, Fredrik Wiklund 41, Wei Zheng 72; Profile Steering Committee, IMPACT Study Steering Committee and Collaborators, PRACTICAL Consortium, VA Million Veteran Program, Tyler M Seibert 4,5,26,73,74
PMCID: PMC13181739  NIHMSID: NIHMS2159576  PMID: 41588240

Abstract

Precision healthcare aims to tailor disease prevention and early detection to individual risk. Prostate cancer screening may benefit from genomics-informed approaches. We developed and validated the P-CARE model—a prostate cancer risk prediction tool combining a polygenic score, family history, and genetic ancestry—using data from over 585,000 male participants in the Million Veteran Program. The model was externally validated in diverse cohorts and implemented via a blended genome-exome assay for clinical use. Here we show that the P-CARE model identifies clinically meaningful gradients of prostate cancer risk among men, with higher scores associated with increased risk of any, metastatic, and fatal prostate cancer. The model is now being used in a clinical trial of precision prostate cancer screening. This work demonstrates the potential for genomics-enabled health systems to improve prostate cancer screening and prevention in men. ClinicalTrials.gov registration: NCT05926102.


Preventive healthcare is moving from a one-size-fits-all approach to more personalized, risk-adapted strategies. Individual risk prediction is an important step for developing tailored strategies for disease prevention and early detection. Risk prediction models can now incorporate larger and more complex arrays of clinical, genetic, environmental, and other risk factors from more diverse populations.1 Genomics specifically is increasingly demonstrating its potential to inform risk stratification for several diseases.2-5 However, much of this potential clinical utility for disease prevention remains theoretical, absent prospective intervention studies demonstrating improved patient outcomes.

Healthcare systems linked to genomic biobanks thus have the opportunity both to generate knowledge about the clinical validity of genomic risk prediction and to demonstrate the clinical utility of implementing that knowledge in care.6 These genomics-enabled learning healthcare systems can leverage knowledge-generating infrastructure to determine whether genomics and other novel risk predictors improve the effectiveness of disease screening and prevention within the healthcare system. The result is not only improved care for patients within that system but also potentially generalizable knowledge for patients in other settings.

Prostate cancer screening is one clinical context where a genomics-enabled learning healthcare system approach might be particularly beneficial. Prostate cancer is one of the most heritable cancers, and recent genomic discoveries have characterized the rare and common genetic variation underlying much of this heritability.7,8 At the same time, clinical guidelines differ on which patient populations are most likely to experience net benefit from prostate cancer screening, including Black men or those with a family history of the disease.9-11 Universal screening with prostate-specific antigen (PSA) reduces prostate cancer mortality but can also overdiagnose indolent disease and lead to unnecessary procedures and treatments.12-15 The result is significant variation in prostate cancer screening practices.16,17 Clinical consensus is even less developed on whether genotype should play a role in prostate cancer risk stratification, despite the discovery of robust associations between risk and both single rare variants and polygenic scores.8

Given this context, genomics-enabled learning health systems can lead the development of genomics-tailored approaches to prostate cancer screening and then evaluate the effectiveness of those approaches in clinical care. This evidence generation is an important step towards the development of clinical guidelines. Here, we describe the development, validation, and clinical implementation of a genomics-informed prostate cancer risk model, developed to enable a randomized clinical trial of precision prostate cancer screening in a large national healthcare system (Clinicaltrials.gov NCT05926102).

RESULTS

Overview of risk prediction model development

Figure 1 illustrates our discovery-to-implementation approach. The Prostate CAncer integrated Risk Evaluation (P-CARE) model was developed and validated to enhance genomic risk assessment for prostate cancer. Using data from the Million Veteran Program (MVP), a large biobank linked to the U.S. Veterans Health Administration, we refined a prostate cancer polygenic score and integrated it with family history and genetic ancestry to create P-CARE. This model was externally validated in four diverse prostate cancer cohorts from the PRACTICAL Consortium. To facilitate clinical implementation, we developed a blended genome-exome (BGE) assay to assess both P-CARE and rare monogenic variants associated with prostate cancer risk. The assay is now being deployed in a randomized clinical trial (ProGRESS, NCT05926102) to evaluate genomics-informed prostate cancer screening in a real-world healthcare setting.

Figure 1: Translating prostate cancer genomic risk discovery to clinical trial implementation.

Figure 1:

1) A prostate cancer polygenic score (PHS601) is trained in the Million Veteran Program (MVP) biobank of the Veterans Health Administration (VA) using known prostate cancer and other prostate trait-associated loci. 2) The Prostate CAncer Risk and Evaluation (P-CARE) model is developed in MVP from PHS601, genetic principal components, and prostate cancer family history. 3) Both PHS601 and P-CARE are replicated in external multiancestry datasets from the Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome (PRACTICAL) consortium. 4) A blended genome-exome (BGE) platform is validated for the P-CARE model, including imputation, analytic validation of PHS601 against whole genome sequencing (WGS) and clinical laboratory validation of P-CARE in All of Us Research Program data. 5) BGE platform is validated for gene panel annotation, filtering, and analysis of rare variants in guideline-informed prostate cancer-associated genes. 6) Clinical P-CARE and rare variant reports with summary recommendations are developed. 7) Clinical laboratory analysis and reporting pipeline is implemented in a pragmatic clinical trial of precision prostate cancer screening across the VA.

Association of polygenic score with prostate cancer outcomes

We assessed whether a polygenic hazard score (PHS601) was significantly associated with prostate cancer risk, metastasis, and mortality in both MVP and PRACTICAL as well as across ancestry groups. We hypothesized that PHS601 would show a strong, consistent association with these outcomes across diverse populations.

The final model included 601 of the 707 unique candidate variants evaluated (Supplementary Data). The resulting polygenic hazard score (PHS601) was associated with age at diagnosis of prostate cancer, metastatic prostate cancer, and prostate cancer death in MVP (Table 1). Among the overall MVP cohort, the HR per standard deviation increase in PHS601 for prostate cancer, metastatic prostate cancer, and prostate cancer death were 2.02 (95% CI 1.97-2.07), 2.07 (95% CI 1.95-2.17), and 1.96 (95% CI 1.75-2.18), respectively. The associations between PHS601 and prostate cancer outcomes were similar in each ancestry-stratified analysis with >100 events in MVP and within each ancestry-specific PRACTICAL dataset (Table 1). Among the East Asian subgroup in MVP, which had small case numbers, associations with metastatic and fatal prostate cancer were not statistically significant but had consistent directions of effect; the association with clinically significant disease was statistically significant in the Asian cohort in PRACTICAL (HR 2.11 95% CI 1.90-2.39). Among the American subgroup in MVP, the association with fatal prostate cancer was not significant but had a consistent direction of effect (HR 2.22, 95% CI 0.98-4.25).

Table 1: Association of polygenic score with prostate cancer outcomes in MVP and PRACTICAL cohorts.

Association of PHS601 with any, metastatic, and fatal prostate cancer in MVP (total and genetic ancestry-stratified groups) and with any, clinically significant, and fatal prostate cancer in four PRACTICAL consortium datasets. Results with less than 50 events per subset were excluded given the unstable nature of the hazard ratio estimates. Abbreviations: CI, confidence interval; COSM, Cohort of Swedish Men; HR, hazard ratio; MVP, Million Veteran Program; PHS, polygenic hazard score; PRACTICAL, Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome; ProtecT, Prostate Testing for Cancer and Treatment.

Clinical endpoint N Event, n HR (95% CI)
HRSD HR80/20 HR20/50 HR80/50 HR95/50
MVP development and validation
Any prostate cancer
 All 585,418 68,618 2.02 (1.97 - 2.07) 6.27 (5.85 - 6.72) 0.43 (0.41 - 0.45) 2.72 (2.62 - 2.82) 4.06 (3.86 - 4.29)
 African 105,014 16,178 1.94 (1.84 - 2.06) 5.81 (5.06 - 6.79) 0.48 (0.45 - 0.52) 2.81 (2.59 - 3.08) 3.74 (3.37 - 4.23)
 European 420,722 48,178 2.04 (1.97 - 2.11) 5.74 (5.26 - 6.20) 0.43 (0.41 - 0.45) 2.46 (2.35 - 2.55) 3.78 (3.54 - 4.03)
 American 50,590 3,775 2.03 (1.83 - 2.26) 5.55 (4.24 - 7.07) 0.44 (0.39 - 0.50) 2.44 (2.13 - 2.79) 3.72 (3.06 - 4.52)
 East Asian 9,092 487 2.10 (1.59 - 2.81) 5.81 (2.90 - 10.65) 0.44 (0.32 - 0.60) 2.41 (1.72 - 3.36) 3.83 (2.31 - 6.04)
Metastatic prostate cancer
 All 585,418 6,606 2.07 (1.95 - 2.17) 6.69 (5.70 - 7.62) 0.42 (0.40 - 0.45) 2.81 (2.58 - 3.01) 4.27 (3.78 - 4.71)
 African 105,014 1,726 1.90 (1.65 - 2.19) 5.60 (3.81 - 7.84) 0.50 (0.43 - 0.58) 2.73 (2.18 - 3.39) 3.62 (2.70 - 4.77)
 European 420,722 4,467 1.96 (1.80 - 2.19) 5.20 (4.19 - 6.83) 0.45 (0.39 - 0.50) 2.33 (2.08 - 2.68) 3.50 (2.97 - 4.29)
 American 50,590 369 2.07 (1.42 - 2.64) 6.02 (2.33 - 10.47) 0.44 (0.32 - 0.67) 2.51 (1.55 - 3.40) 3.92 (1.91 - 6.08)
 East Asian 9,092 44
Fatal prostate cancer
 All 585,418 1709 1.96 (1.75 - 2.18) 5.81 (4.31 - 7.65) 0.45 (0.40 - 0.51) 2.60 (2.22 - 3.03) 3.82 (3.06 - 4.72)
 African 105,014 365 1.62 (1.16 - 2.12) 3.81 (1.49 - 7.26) 0.61 (0.44 - 0.85) 2.15 (1.27 - 3.22) 2.68 (1.35 - 4.45)
 European 420,722 1250 1.92 (1.60 - 2.20) 4.99 (3.18 - 6.86) 0.47 (0.39 - 0.57) 2.27 (1.81 - 2.69) 3.39 (2.41 - 4.32)
 American 50,590 87 2.22 (0.98 - 4.25) 8.45 (0.95 - 32.23) 0.48 (0.19 - 1.03) 2.78 (0.97 - 6.14) 4.81 (0.96 - 14.32)
 East Asian 9,092 8
PRACTICAL replication
Any prostate cancer
 COSM 3,415 2,298 2.27 (2.11 - 2.46) 9.18 (6.66 - 12.93) 0.36 (0.31 - 0.42) 3.34 (2.97 - 3.98) 5.35 (4.26 - 6.92)
 ProtecT 6,411 1,583 1.87 (1.78 - 2.01) 5.67 (4.75 - 6.73) 0.44 (0.40 - 0.48) 2.78 (2.47- 3.05) 3.78 (3.31 - 4.29)
 African 6,253 3,240 1.84 (1.72 - 1.98) 8.55 (6.64 - 11.08) 0.41 (0.36 - 0.46) 3.47 (2.93 - 3.99) 4.49 (3.71 - 5.37)
 Asian 2,320 1,164 2.15 (1.92 - 2.38) 8.80 (6.73 - 11.09) 0.35 (0.30 - 0.39) 3.02 (2.63 - 3.39) 5.26 (4.24 - 6.48)
Clinically significant prostate cancer
 COSM 3,415 1,487 2.30 (2.09 - 2.53) 9.35 (7.32 - 12.20) 0.36 (0.31 - 0.41) 2.81 (2.58 - 3.01) 4.27 (3.78 - 4.71)
 ProtecT 6,411 628 2.02 (1.88 - 2.21) 7.02 (5.65 - 8.42) 0.40 (0.36 - 0.44) 2.73 (2.18 - 3.39) 4.45 (3.76 - 5.11)
 African 6,253 1,424 1.85 (1.69 - 2.02) 8.61 (6.50 - 11.61) 0.41 (0.35 - 0.47) 2.51 (1.55 - 3.40) 3.92 (1.91 - 6.08)
 Asian 2,320 716 2.11 (1.90 - 2.39) 7.88 (5.94 - 10.68) 0.36 (0.32 - 0.42) 2.85 (2.44 - 3.32) 4.83 (3.84 - 6.09)
Fatal prostate cancer
 COSM 3,415 278 1.91 (1.65 - 2.28) 5.88 (3.51 - 9.19) 0.45 (0.36 - 0.56) 2.58 (1.97 - 3.32) 3.82 (2.59 - 5.35)

Association of P-CARE with prostate cancer outcomes

We evaluated whether P-CARE, which integrates a polygenic risk score with genetic ancestry and family history, improves prostate cancer risk stratification and correlates with disease severity. Family history was independently significant for prostate cancer risk stratification (Supplemental Table 1) and inclusion of genetic ancestry improved performance of our previous PHS,18 so both were included in the model a priori. As hypothesized, the P-CARE model that integrated PHS601, genetic ancestry, and family history described a strong gradient of risk for any, clinically significant, metastatic, and fatal prostate cancer across MVP and PRACTICAL datasets (Table 2, Supplemental Tables 2-5). Among the overall MVP cohort, the HR per standard deviation increase in P-CARE for prostate cancer, metastatic prostate cancer, and prostate cancer death were 2.04 (95% CI 1.99-2.08), 2.05 (95% CI 1.93-2.16), and 1.95 (95% CI 1.76-2.15), respectively. Across the MVP and PRACTICAL datasets, compared to men with median P-CARE values, men in the lowest P-CARE quintile had HR 0.35-0.46 for the 4 prostate cancer outcomes (HR20/50), while men in the highest P-CARE quintile had HR 2.48-4.03 (HR80/50, Table 2). The direction and magnitude of association between P-CARE and the prostate cancer outcomes were similar in analyses of subgroups defined by genetic ancestry (Supplemental Table 6) and, alternatively, by self-reported race and ethnicity (Supplemental Table 7), in each subgroup with adequate case counts. As additional validation, time-dependent AUC analysis, sensitivity and specificity of defined P-CARE risk-category thresholds, and random forest survival modeling confirmed consistent model discrimination and robustness (Supplemental Tables 8-10)

Table 2: Association of P-CARE model with prostate cancer outcomes in MVP and PRACTICAL cohorts.

Association of P-CARE model with any prostate cancer, clinically significant prostate cancer, metastatic prostate cancer, and fatal prostate cancer in MVP and four PRACTICAL consortium datasets. As described, P-CARE model consists of PHS601, first-degree family history of prostate cancer, and genetic principal components. Abbreviations: CI, confidence interval; COSM, Cohort of Swedish Men; HR, hazard ratio; MVP, Million Veteran Program; P-CARE, Prostate CA Risk and Evaluation; PRACTICAL, Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome; ProtecT, Prostate Testing for Cancer and Treatment.

Clinical endpoint N HR (95% CI)
HRSD HR80/20 HR20/50 HR80/50 HR95/50
Any prostate cancer
 MVP 585,418 2.04 (1.99 - 2.08) 6.33 (5.95 - 6.71) 0.43 (0.42 - 0.45) 2.75 (2.66 - 2.84) 4.09 (3.89 - 4.29)
 COSM 3,415 2.33 (2.14 - 2.58) 9.40 (6.88 - 13.20) 0.37 (0.31 - 0.42) 3.43 (2.91 - 4.08) 5.45 (4.31 - 7.00)
 ProtecT 6,411 1.87 (1.77 - 2.01) 5.53 (4.60 - 6.66) 0.45 (0.41 - 0.49) 2.48 (2.27 - 2.73) 3.71 (3.23 - 4.23)
 African 6,253 2.00 (1.86 - 2.16) 9.68 (7.55 - 12.49) 0.40 (0.36 - 0.46) 3.88 (3.35 - 4.42) 5.16 (4.35 - 6.03)
 Asian 2,320 2.17 (1.95 - 2.42) 8.74 (6.60 - 11.28) 0.35 (0.30 - 0.40) 3.06 (2.65 - 3.46) 5.23 (4.09 - 6.47)
Clinically significant prostate cancer
 COSM 3,415 2.34 (2.12 - 2.55) 9.39 (7.25 - 12.27) 0.37 (0.32 - 0.42) 3.43 (2.98 - 3.99) 5.47 (4.47 - 6.76)
 ProtecT 6,411 2.01 (1.87 - 2.18) 6.81 (5.42 - 8.20) 0.40 (0.36 - 0.44) 2.77 (2.48 - 3.04) 4.35 (3.71 - 4.12)
 African 6,253 2.04 (1.86 - 2.23) 10.30 (7.66 - 13.47) 0.39 (0.36 - 0.44) 4.03 (3.39 - 4.86) 5.41 (4.38 - 6.77)
 Asian 2,320 2.11 (1.91 - 2.35) 7.60 (5.70 - 10.15) 0.38 (0.33 - 0.43) 2.84 (2.42 - 3.31) 4.68 (3.71 - 5.73)
Metastatic prostate cancer
 MVP 585,418 2.05 (1.93 - 2.16) 6.50 (5.51 - 7.38) 0.43 (0.40 - 0.46) 2.78 (2.54 - 2.99) 4.17 (3.68 - 4.59)
Fatal prostate cancer
 MVP 585,418 1.95 (1.76 - 2.15) 5.71 (4.33 - 7.30) 0.45 (0.41 - 0.52) 2.59 (2.22 - 2.97) 3.77 (3.05 - 4.57)
 COSM 3,415 1.95 (1.66 - 2.37) 5.95 (3.61 - 9.29) 0.46 (0.37 - 0.56) 2.65 (2.03 - 3.41) 3.87 (2.69 - 5.49)

Within the ProtecT dataset, the PPV of a PSA value ≥3 ng/mL for clinically significant prostate cancer was 0.13 (95% CI 0.12-0.14) in the overall dataset and 0.19 (95% CI 0.16-0.21) and 0.23 (0.17-0.28) in the subsets in the top 20% and top 5% of P-CARE values, respectively (Figure 2, stratified by PSA level in Extended Data Figure 1). The percentage of true positive cases within the ProtecT dataset that fall into high P-CARE categories is shown in Extended Data Figure 2.

Figure 2. Positive predictive value of PSA in ProtecT by P-CARE values.

Figure 2.

Illustrated are mean PPV (95% CI) for a PSA value ≥3 ng/mL for clinically significant prostate cancer among three groups of men in the ProtecT study (n=6,411): all men (regardless of P-CARE value), men in the top 20% of P-CARE values (P-CARE80), and men in the top 5% of P-CARE values (P-CARE95). Abbreviations: CI, confidence interval; P-CARE, Prostate CA Risk and Evaluation; PPV, positive predictive value; ProtecT, Prostate Testing for Cancer and Treatment; PSA, prostate-specific antigen.

We defined P-CARE risk categories by hazard ratio thresholds (HR 0.75 and HR 1.5 for metastatic prostate cancer) and evaluated both cumulative incidence and risk-equivalent age for any, metastatic, and fatal prostate cancer (Table 3). Overall, the model categorized 25.1%, 37.3%, and 37.6% of MVP participants as low-, average-, and high-risk, respectively. The model categorized 68.7% of participants with positive family history as high-risk and only 5.6% as low-risk. Among participants self-reporting Black or African-American race, only 2.8% were categorized as low risk. Figure 3 shows cumulative prostate cancer incidence curves in MVP both by P-CARE percentile groups and by P-CARE risk category. As shown in Table 4, by age 80, men in the high-risk P-CARE group had a cumulative risk of any, metastatic, and fatal prostate cancer of 37.4%, 4.4%, and 0.8%, respectively. The expected age of any and metastatic prostate cancer occurred 5 years earlier in the high-risk group compared to the men in the standard risk; specifically, a man in the high-risk group reached a prostate cancer detection risk equivalent to the 55-year standard at an age of 50 years and a metastatic prostate cancer risk equivalent to the 70-year standard at an age of 63.5 years (Supplemental Table 11).

Table 3: Characteristics of P-CARE risk categories for metastatic prostate cancer among 585,418 MVP participants.

P-CARE risk categories are defined by thresholds of HR 0.75 and HR 1.5 for metastatic prostate cancer. Participants with a hazard ratio for metastatic prostate cancer <0.75 and >1.5 are defined as low- and high-risk, respectively. MVP, Million Veteran Program; P-CARE, Prostate CA Risk and Evaluation.

P-CARE risk category, N (%)
N Low risk (HR<0.75) Average risk (HR 0.75-1.5) High risk (HR>1.5)
Total 585,418 146,826 (25.08) 218,530 (37.33) 220,062 (37.59)
Positive family history 28,358 1,595 (5.62) 7,270 (25.63) 19,493 (68.73)
Genetic ancestry groups
 African 105,014 2,607 (2.48) 19,314 (18.39) 83,093 (79.12)
 European 420,722 128,096 (30.44) 173,774 (41.30) 118,852 (28.24)
 American 50,590 12,842 (25.38) 21,518 (42.53) 16,230 (32.08)
 East Asian 9,092 3,281 (36.08) 3,924 (43.15) 1,887 (20.75)
Self-reported race/ethnicity groups
 American Indian or Alaska Native 5507 1,346 (24.44) 2,236 (40.60) 1,925 (34.95)
 Asian 6210 2,403 (38.69) 2,684 (43.22) 1,123 (18.08)
 Black or African American 101,920 2,812 (2.75) 18,986 (18.62) 80,122 (78.61)
 Hispanic White 26,037 6,871 (26.38) 11,067 (42.50) 8,099 (31.10)
 Native Hawaiian or Pacific Islander 3042 755 (24.81) 1,259 (41.38) 1,028 (33.79)
 Non-Hispanic White 418,387 126,633 (30.26) 172,661 (41.26) 119,093 (28.46)
 Other 8077 1,938 (23.99) 3,303 (40.89) 2,836 (35.11)
 Unknown 16,238 4,068 (25.05) 6,334 (39.00) 5,836 (35.94)

Figure 3: Prostate cancer cause-specific cumulative incidence in MVP by P-CARE strata.

Figure 3:

Cause specific cumulative incidence within MVP for (A) prostate cancer, (B) metastatic prostate cancer, and (C) fatal prostate cancer. The left column shows incidence for each endpoint by P-CARE percentile group: 0-20th, 30-70th, 80-100th, and 95-100th. The right column shows incidence for each endpoint by P-CARE risk category: high, average, and low risk. Abbreviations: MVP, Million Veteran Program; P-CARE, Prostate CA Risk and Evaluation.

Table 4: Prostate cancer cause-specific cumulative incidence in MVP by P-CARE category.

Abbreviations: HR, hazard ratio; MVP, Million Veteran Program; P-CARE, Prostate CA Risk and Evaluation.

Clinical endpoint Cumulative incidence (%)
Low risk (HR<0.75) Average risk (HR 0.75-1.5) High risk (HR>1.5)
Prostate cancer
 By age 70 4.02 8.38 21.22
 By age 80 9.17 17.68 37.43
 By age 90 14.11 25.29 47.87
Metastatic prostate cancer
 By age 70 0.18 0.46 1.41
 By age 80 0.77 1.64 4.38
 By age 90 1.96 4.29 9.34
Fatal prostate cancer
 By age 70 0.02 0.06 0.21
 By age 80 0.13 0.34 0.82
 By age 90 0.81 1.58 3.61

Development and validation of clinical laboratory assay for genetic prostate cancer risk

We then used the BGE platform to develop a clinical laboratory assay for inherited prostate cancer risk by combining P-CARE, which evaluates polygenic risk, with targeted testing for 12 genes known to be associated with hereditary prostate cancer. First, both PHS601 and the integrated P-CARE model were again externally validated in the All of Us (AoU) cohort, demonstrating strong associations with prostate cancer across diverse ancestry groups. Within the AoU dataset, the PHS601 was associated with prostate cancer with an odds ratio per standard deviation of 1.91 (95% CI: 1.85-1.98). In the same dataset, for the full P-CARE model (PHS601 plus genetic principal components and family history) we found an odds ratio of 2.41 (95% CI: 2.25-2.60) for individuals in the high-risk category to be diagnosed with prostate cancer, compared to individuals classified as average risk. Similarly, individuals in the low-risk category show an odds ratio of 0.48 (95% CI: 0.44-0.54), compared to individuals classified as average risk. Notably, this strong association holds across different ancestries. (Extended Data Figure 3).

Next, the accuracy and reliability of the BGE assay in detecting both polygenic and rare monogenic variants associated with prostate cancer risk were evaluated against known reference samples; the BGE platform produced nearly identical results for polygenic risk scores and ancestry estimates, with Pearson correlations exceeding r>0.998 for PHS601 and r>0.999 for both principal components. For the 12 genes related to hereditary prostate cancer risk, the assay met quality thresholds for coverage and variant detection in 11 of 12 genes. The one exception was PMS2, a technically challenging gene due to its similarity to a nearby pseudogene, which can interfere with accurate sequencing. Within the PMS2 gene, exon 13, 14 and 15 were undercovered in a subset of samples, with 80% and 20% of samples missing full coverage in those regions, respectively. Of the 18 samples assessed for monogenic rare variants, all 18 variants of interest were successfully detected, including 7 single-nucleotide variants (SNVs), 5 insertions/deletions (indels) and 6 copy number variants (CNVs). However, three of the CNVs were classified as low quality based on pre-specified thresholds for clinical reporting QUAL ≥ 50 for duplications, QUAL ≥ 100 for heterozygous deletions, and QUAL ≥ 400 for homozygous deletions) and would have not been clinically reported in a real-world setting. . This is consistent with known limitations in detecting small CNVs involving fewer than three exons. Despite this, the platform showed excellent technical performance with 100% precision across repeated tests, both within and between sequencing runs.

Clinical P-CARE and monogenic reports

Here we describe the implementation of the P-CARE and monogenic risk reports in clinical use, linking them to personalized screening recommendations in the ProGRESS trial. An example of the resulting laboratory report package is shown in Supplementary Information. The cover page summarizes the results of both the monogenic and P-CARE analyses and provides an overall risk category for the individual based on these results. An individual with a pathogenic or likely pathogenic variant in one of the 12 prostate cancer-associated genes is categorized as high-risk, regardless of P-CARE results. Individuals without such a variant are categorized as low-, average-, or high-risk according to their P-CARE result, with thresholds at HR=0.75 and HR=1.5, as described in the Methods. The cover page also links these risk categories to tailored prostate cancer screening recommendations for the individual. After this cover page summary, separate P-CARE and rare variant reports provide further detail about these individual result types, including information about P-CARE model development and validation, technical descriptions of the analyses performed, relevant gene and disease information, and literature references. These reports are now being used in the national ProGRESS randomized clinical trial, in which 5,000 prostate cancer screen-eligible VA patients are randomly assigned to usual care versus precision screening recommendations informed by P-CARE and rare variants.

DISCUSSION

We used genomic, clinical, and survey data from a large national biobank to develop a genomics-informed prostate cancer prediction model consisting of family history, genetic principal components, and an updated polygenic score of 601 prostate trait-associated loci. Patients in the lowest and highest 20% of values under this model have 0.4-fold and 2.7-fold risk of prostate cancer, respectively, compared to those with median values; replication in external multiancestry cohorts confirmed these associations. Men at highest risk of developing advanced prostate cancer are most likely to benefit from screening; the P-CARE model is associated with risk of all, clinically significant, metastatic, and fatal prostate cancer. When low and high risk were defined as HR<0.75 and HR>1.5, respectively, for metastatic prostate cancer, the cumulative incidence of metastatic prostate cancer by age 80 in the biobank was 0.8% in the low-risk group and 4.4% in the high-risk group.

Unlike our previous PHS (PHS290), both P-CARE and PHS601 have relatively similar performance at a population level to discriminate prostate cancer risk; family history and agnostic genetic ancestry have less prognostic value in the current multivariable model than in our prior model. However, the effect of family history is substantial for individuals so inclusion of family history could make a difference at an individual level in clinical decision making. While ancestry appears to be mostly accounted for by PHS601, we opted not to exclude this post-hoc. We then developed and validated a clinical assay on a cost-efficient BGE platform for both the prediction model and rare pathogenic variants in known prostate cancer genes. This assay and associated clinical reports are now enabling a clinical trial of precision prostate cancer screening among patients receiving care from the national healthcare system from which the biobank data were derived. This approach illustrates the power of genomics-enabled learning health systems to generate translatable discoveries for implementation in preventive healthcare.

We designed the P-CARE model and ongoing prostate cancer screening trial to examine how the routine collection and interpretation of genomic data in preventive care might improve upon existing screening practices in a large integrated health system. Prostate cancer is highly prevalent, but despite randomized controlled trial evidence that screening with PSA testing can reduce prostate cancer mortality,14,19 guidelines vary by organization and country11 on how to balance the benefits of screening (early detection and treatment, resulting in lower incidence of advanced and lethal disease) and its potential risks (overdiagnosis of apparently indolent disease and morbidity from unnecessary procedures and treatments). As a result, screening practices are highly variable.16,17,20-23 Better models are needed to distinguish men most likely to benefit from screening from those for whom its risks might outweigh its benefits. A learning health system approach is ideal to improve prostate cancer screening for a few reasons. First, risk prediction models that inform the net benefit of cancer screening depend in large part on model calibration within a population; relative and absolute risk estimates derived from a healthcare system-linked biobank are thus particularly informative for patients receiving care in that system. In particular, age is a critical factor not only in the risk of advanced prostate cancer but also for the competing health risks that might make prostate cancer early detection less important;24 our time-to-event analysis and age-specific cumulative incidence curves account for age and allow physicians to balance these with age-related competing risks for a given individual to guide age-based screening decisions. Second, the effect sizes of polygenic scores themselves, including for prostate cancer, can vary between biobanks.25,26 Third, the net benefit of prostate cancer screening in a population is highly dependent on the downstream diagnostic and therapeutic management of elevated PSA values and abnormal prostate biopsy results;27,28 nesting the evaluation of a new screening paradigm within its target healthcare delivery system helps ensure that system-specific clinical practice patterns are included.

Our approach also seeks to address controversies in prostate cancer screening that are intimately intertwined with health disparities. In the United States, Black men are more likely to be both diagnosed with and die from prostate cancer.29 Possible causal factors include genetic, environmental, and social determinants of health, including structural factors including access to screening and other healthcare.30-32 Black men are highlighted in prostate cancer guidelines as a group whose high risk merits earlier screening.9,10 This recommendation is appropriate to address racial disparities in prostate cancer outcomes. However, at the same time, the use of race in medical decision-making can inappropriately ascribe to biology effects that arise from a complex social construct confounded by myriad social determinants of health; it also ignores the complex multiracial and multiancestry backgrounds of individuals in modern healthcare system populations. We thus set out to develop a prostate cancer risk prediction model that did not include discrete race or genetic ancestry categories, favoring instead principal components as a continuous measure of genetic variation. At the same time, we confirmed that the resulting model performed well across categories of socially defined populations (race and ethnicity groups in MVP and external cohorts). Initial genome studies predominantly included individuals with European ancestry, but more recent work has improved genetic discovery and risk stratification in more diverse populations, including African ancestry.18,33-37 The P-CARE model extends this work, confirming that most Black men, but not all, have high risk. While the model does not fully disentangle the confounded associations between genetic ancestry and social determinants of prostate cancer risk, it represents an advance towards a more equitable, tailored approach to risk stratification and screening that does not treat race as a biological construct.

Family history of prostate cancer and certain rare genetic variants are also known prostate cancer risk factors, independent of ancestry and polygenic score.18,33,34,37 We designed the P-CARE model to build upon, not replace, these clinical risk factors, similar to breast cancer screening models.38,39 Rare variants in several genes, including BRCA2 and MSH2, are known to increase prostate cancer risk and thus have separate screening guidelines for carriers.20 Carrier status of these variants is presently unknown for the vast majority of prostate cancer screen-eligible patients and yet might play a more prominent role in preventive care in a future when genomic testing is more commonplace. Despite aggregate analyses suggesting that polygenic scores can modify the effects of these rare variants,40-42 we determined that these modified associations are not yet robust enough for individual variant-level clinical reporting and should not supersede NCCN guidelines for the clinical management of rare variants. We therefore chose a genomic analysis platform that could detect and interpret these important rare variants and will report them according to established clinical guidelines to participants. By combining high-coverage exome and low-coverage whole genome sequencing data, the novel BGE technology provides a cost-efficient, scalable and accurate platform for implementing the P-CARE model in clinical care. In the ProGRESS trial (ClinicalTrials.gov ID NCT05926102) participants and their healthcare providers are now receiving clinical reports with P-CARE results and the results of rare variant analysis, enabling an evaluation of a precision screening approach on prostate cancer in the U.S. Veterans Health Administration.

Prior modeling studies suggest that the use of polygenic scores can improve the cost-effectiveness of prostate cancer screening pathways with and without MRI;43,44 the ProGRESS trial will provide additional empiric data to determine the costs and cost-effectiveness of such an approach in real-world implementation. Even if the cost-effectiveness of a polygenic approach to screening is marginal for single cancers,45,46 multiplex platforms such as the BGE enable both monogenic variant screening and polygenic risk stratification with one test. Modeling studies already suggest that population genomic screening for a select number of monogenic diseases is cost-effective.47,48 More complex models incorporating polygenic risk for multiple diseases are needed, but it is plausible that as the costs of genomic testing decrease, the incremental cost-effectiveness of adding polygenic approaches to genomic medicine programs will be favorable, depending on patient population, healthcare setting, and country.

Our work has some limitations. Despite the strengths of our learning healthcare system approach described above, this system-specific model may not generalize to other settings with different population risks and screening practices. Model replication in the diverse PRACTICAL and All of Us datasets mitigates this concern, but other healthcare systems should examine model calibration in their own data before implementation. In addition, while the inclusion of family history, polygenic score, genetic principal components, and rare variants improves upon existing clinical prostate cancer screening approaches, the P-CARE model cannot disentangle the effects of genetic predisposition from environmental exposures and other social determinants that shape prostate cancer risk. Ongoing and future work should examine how to model and include other important risk factors in a clinically implementable risk stratification tool, including the consideration of other machine learning-based prediction approaches.29,49,50 Finally, BGE has many benefits including genome level variant information for PRS as well as an exome backbone for monogenic reporting, but there are limitations that come with an exome based approach that a purpose built capture panel may overcome, including lower sensitivity around complex regions of genes like PMS2 and reduced sensitivity of small copy number variants below 3 exons in size.

In summary, a healthcare system-linked biobank has enabled the development, replication, and clinical laboratory validation of an updated prostate cancer risk model, now implemented in a clinical trial of precision prostate cancer screening. This approach exemplifies the power of genomics-enabled learning health systems to accelerate the discovery and translation of precision technologies to improve population health outcomes.

METHODS

Study overview

The VA Central IRB approved this study (IRBNet 1735869 and 1735136). As described in detail below, we used data from a large biobank linked to a national healthcare system to update a prior prostate cancer polygenic score.18 We then developed and cross-validated a prostate cancer prediction model based on the combination of that score and family prostate cancer history, now termed the Prostate CAncer integrated Risk Evaluation (P-CARE) model. We further validated the P-CARE model in four external prostate cancer cohort datasets prior to the development and validation of a clinical blended genome-exome (BGE) assay both for the P-CARE model and also for rare prostate cancer-associated monogenic variants. This assay is now being implemented in a randomized clinical trial of genomics-informed prostate cancer screening in a new cohort of patients from the national healthcare system in which the P-CARE model was first developed [the Prostate Cancer, Genetic Risk, and Equitable Screening Study (ProGRESS), ClinicalTrials.gov ID NCT05926102].

Participants and phenotype definitions

Genotype and phenotype data were analyzed from the following cohorts33,51,52 also summarized in Supplemental Tables 12 and 13.

Million Veteran Program

Data from the Million Veteran Program (MVP) were used to update a previous prediction model18 to develop the new P-CARE model. MVP is a mega-biobank linked to the national Veterans Health Administration healthcare system of the US Department of Veterans Affairs (VA).51 Participants provide biospecimens, consent to research access to their VA health records, and complete surveys about family health history, health behaviors, military and environmental exposures, and other health-related factors. For the present analyses, we used data from 585,418 male MVP participants to develop and cross-validate the P-CARE model. All study participants provided blood samples for DNA extraction and genotyping using a custom Affymetrix Axiom biobank array containing 723,305 variants, enriched for low-frequency variants in African and Hispanic populations.53 Family history was defined as the presence or absence of paternal history of prostate cancer, as reported on the MVP survey. Prostate cancer diagnosis, age at diagnosis, and date of last follow-up were retrieved from the VA Corporate Data Warehouse based on International Classification of Diseases (ICD) diagnosis codes and VA Central Cancer Registry data.18,54 Age at diagnosis of metastasis (nodal and/or distant, regardless of whether metastases were detected at diagnosis or at recurrence) was determined via a validated natural language processing tool developed in the VA system.18,55 Cause and date of death were obtained from the National Death Index. Fatal prostate cancer was defined by ICD9 code 185 or ICD10 code C61 as the underlying cause of death.

PRACTICAL Consortium

Data from 4 external cohorts from the Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome (PRACTICAL) Consortium were used to externally validate the P-CARE model. Data from 18,457 men previously genotyped via OncoArray or iCOGs arrays56,57 were divided into four datasets, as described in prior studies evaluating polygenic scores: 1) men of African ancestry (n=6,253); 2) men of Asian ancestry (n=2,320); 3) the Cohort of Swedish Men (COSM) population-based cohort with long-term outcomes (n=3,415); and 4) the population-based Prostate Testing for Cancer and Treatment (ProtecT) screening trial (n=6,411).33 Family history was defined as the presence or absence of a first-degree relative with a prostate cancer diagnosis. Clinically significant prostate cancer was defined as any case with Gleason score ≥7, PSA ≥10 ng/mL, T3-T4 stage, nodal metastases, or distant metastases.33 The COSM dataset additionally had age at prostate cancer death,58 and the ProtecT dataset had prostate biopsy results for both cases and controls with screening PSA ≥3 ng/mL.59,60

All of Us Research Program

Data from the v7 release of the All of Us (AoU) Research Program were used as an additional external validation cohort. Excluding samples flagged for failing QC criteria, for being related, or for lack of available electronic health record data, 74,331 samples with short-read WGS data and male sex assigned at birth were analyzed. Samples were classified as cases (n=4,473) and controls (n=69,858) based on the presence or absence of “Malignant neoplasm of prostate” or “Personal history of malignant neoplasm of prostate” in the AoU electronic health record data. Family history was determined based on AoU survey data as positive (responses “Father,” “Sibling,” or “Son” to the question “Including yourself, who in your family has had prostate cancer?”; n=3,034) or negative otherwise (no response or different response to survey question; n=71,297). The model validity was evaluated within and across AoU-provided predicted genetic ancestries (African/African American, n=16,733; American Admixed/Latino, n=10,769; East Asian, n=1,436; European, n=43,917; Middle Eastern, n=346; South Asian, n=1,130) (Supplemental Table 14).

Candidate variants and training for polygenic score

We considered variants previously identified from the following sources for potential inclusion in an updated polygenic score for the P-CARE model: 290 variants from a prior score, 613 variants identified as prostate cancer susceptibility loci in a multi-ancestry genome-wide association studies, 23 variants identified as susceptibility loci for benign elevation of prostate-specific antigen (PSA) or benign prostatic hypertrophy (BPH), 9 variants identified as prostate cancer susceptibility loci in men of African ancestry in a genome-wide meta-analysis, and 128 variants identified as susceptibility loci for prostate cancer in a genome-wide multi-ancestry meta-analysis.33,35,36,61,62 A machine-learning, least absolute shrinkage and selection operator (LASSO)-regularized Cox proportional hazards model approach was used in the MVP dataset to select the final variants for the polygenic score and estimate weights, using the R (v.4.4) “glmnet” package (v.4.1.8).63-65 To develop the polygenic score, age at any prostate cancer diagnosis in MVP was the time to event, as this gives the most statistical power; controls were censored at age of last follow-up. First, we identified pairs of variants with highly correlated genotype (defined as r2>0.95) and used univariable Cox models to exclude the variant from each pair with weakest univariable association. Next, all remaining candidate variants were evaluated for inclusion in the new polygenic score using a Cox model with genotype allele counts of candidate variants and the first five FastPop principal components as predictor variables. Genetic principal components were estimated using 2,309 ancestry informative markers from FastPop.66 Loadings for the first 5 principal components were estimated in the 1000 Genomes Phase 3 dataset.67 The final form of the LASSO model was estimated using the lambda value that minimized the mean cross-validated error.68

We then used Cox proportional hazards models to evaluate the association of the new polygenic score with age at diagnosis of prostate cancer, age at diagnosis of nodal and/or distant metastatic prostate cancer, and age at prostate cancer death within the MVP dataset overall and in analyses stratified by continental population ancestry group, as in prior work.18,37,58,63,69-72 Similarly, Cox models were used to evaluate the association between the new score and age at diagnosis of any prostate cancer, clinically significant prostate cancer, and fatal prostate cancer (in the COSM dataset) in the PRACTICAL cohort.18

P-CARE model development and validation

The resulting polygenic score was then carried forward for use in the development of an integrated clinical prediction model within MVP. We developed a Cox model for age at prostate cancer diagnosis as a function of the polygenic score, modeled as a continuous variable; family history of prostate cancer (Supplemental Table 1), modeled as a binary variable indicating presence or absence of at least one first-degree relative with prostate cancer; and population structure, modeled using the first two genetic principal components (PCs). Prior analyses showed that the first two PCs are sufficient to capture genetic variation for prostate cancer risk stratification compared to 5-10 PCs.37 Individuals not meeting the endpoint of interest were censored at last follow-up. Training on metastatic disease did not give improved results, due to lower event rates.

The resulting P-CARE model was then validated internally within the MVP dataset and externally within the 4 PRACTICAL datasets. Where available, we evaluated the association of the P-CARE model with age of diagnosis of any prostate cancer, clinically significant prostate cancer, metastatic prostate cancer, and fatal prostate cancer. As in our prior work,18,33,34,37,58,63,70-72 we estimated illustrative effect sizes using hazard ratios (HRs) and 10 iterations of 10-fold cross validation, calculated to make the following comparisons: HR80/20, men in the highest 20% versus lowest 20%; HR95/50, men in the highest 5% versus those with median values; and HR20/50, men in the lowest 20% vs those with median values. Within the MVP dataset, we generated cumulative incidence curves for each prostate cancer endpoint by P-CARE percentile groups, as in prior work.63,70 We additionally generated cumulative incidence curves by P-CARE risk categories defined by risk of metastatic disease, given its morbidity and mortality and to counter the criticism that current prostate cancer screening approaches over-detect indolent disease.12-15 The high risk category was defined as an overall P-CARE HR>1.5 for metastatic prostate cancer and the low risk category was defined as HR<0.75 (consistent with routine clinical prediction tools for other diseases, such as breast cancer, diabetes, and cardiovascular disease73-75); all other risk values were considered average risk. The ages at which different P-CARE percentiles and P-CARE risk groups reached an equivalent cumulative risk of any and metastatic/fatal prostate cancer as that of the average risk man at 55 and 70 years old, respectively, were also determined. Because ProtecT systematically collected prostate biopsies, this dataset offered the opportunity to correlate PSA values with likelihood of clinically significant prostate cancer. Within the ProtecT dataset, we calculated the positive predictive value (PPV) of a PSA value ≥3 ng/mL for clinically significant prostate cancer on biopsy among participants in the top 5th (PPV95) and top 20th P-CARE percentile (PPV80).33,60

Clinical laboratory assay development and validation

The P-CARE model was then carried forward to develop a clinical laboratory assay (Broad Clinical Labs, Burlington, MA, USA) to enable precision prostate cancer screening informed by both the model and relevant rare variants, given their importance in prostate cancer risk.

Blended genome exome assay

We constructed the assay on a novel blended genome exome (BGE) platform76, which achieves cost-efficiency for detecting rare and common variants by combining 2-3x whole genome sequencing (WGS) with 60-90x exome sequencing in a single sequenced sample. The BGE platform has achieved >99% concordance with 30x genome sequencing data for both exome and genome short variants.76 Short variant calling was performed over the high coverage exome target regions using the Illumina DRAGEN Bio-IT platform version 4.2.7. Genotypes and dosage information over the whole genome were obtained from sequencing data through GLIMPSE2 imputation77 using the gnomAD HGDP and 1000 Genomes callset.78 Copy number variation was detected over the exome target regions using GATK-gCNV.79

Analytic and clinical laboratory validation of polygenic score and P-CARE model

The analytic validity of the BGE platform for the polygenic score was assessed by comparing 60 clinical samples with previously-identified variants; reference samples from Coriell Institute for Medical Research with curated reference variant data sets maintained by the National Institute of Standards and Technology; and samples with known SNVs, indels, and CNVs from a combination of previous in-production clinical samples, previous eMERGE studies, previous CAP proficiency testing samples, Coriell samples, and the Coriell Ancestral Panel. For each of these samples, representing 6 genetic ancestry groups (Admixed American, African, Non-Finnish European, East Asian, South-East Asian, Ashkenazi), we generated both BGE and WGS data and calculated the polygenic score and genetic principal components. Additional evidence of clinical validity for both the polygenic score and the P-CARE model was obtained using 74,331 samples from the All of Us (AoU) Research Program. Polygenic score and genetic principal components were calculated from the WGS genotypes provided by AoU. Individuals were classified as cases and controls based on the AoU electronic health record data. P-CARE values were calculated for each AoU participant using polygenic score, the first two genetic principal components, first-degree family history of prostate cancer, and the MVP-derived coefficients. To determine the association between P-CARE and prostate cancer case status in AoU, we calculated odds ratios for an individual to be diagnosed with prostate cancer in the low and high P-CARE categories, relative to the average category, using logistic regression models controlling for age.

Rare variant selection, validation, and interpretation

We identified known prostate cancer-associated genes for which the National Comprehensive Cancer Network has issued clinical management recommendations.20,80,81 This gene list informed the filtering for an in silico gene panel for rare variant analysis. The ability of the BGE to identify pathogenic or likely pathogenic variants in these genes was evaluated by assessing the overall technical performance of 12 genes related to hereditary prostate cancer risk (BRCA1, BRCA2, ATM, PALB2, CHEK2, HOXB13, MLH1, MSH2, MSH6, PMS2, TP53, and EPCAM) and identification of known variants from previous clinical testing (SNVs, small InDels, CNVs) within these genes in 18 clinical samples. Technical performance of these genes was assessed by determining the percentage of undercovered bases within a panel gene. A base is considered covered if it satisfies the following: coverage >20X, base quality >20, and mapping quality >20. This coverage analysis was performed with two sample fraction thresholds: ≥80% and ≥20%. We determined the sensitivity for the detection of rare monogenic variants if the variant of interest was identified in the variant call file and would meet quality and prioritization metrics to be flagged for manual review by our tertiary analysis platform. Additionally, inter and intra run precision was assessed by running samples in triplicate across different runs and within the same run, respectively. We developed a workflow to classify, review, and prioritize variants in a tertiary analysis platform (Fabric Genomics, Oakland, CA, USA) prior to in-house clinical interpretation and reporting of pathogenic and likely pathogenic variants by a team of board-certified geneticists.

Clinical report development

After clinical laboratory validation of the P-CARE and rare variant pipelines, we developed a laboratory report package suitable for the clinical implementation of these results, consistent in format and content with other clinical genetic test reports and with our prior work.4,5,82 As described in the Results, the report package consisted of separate laboratory reports for the P-CARE and rare variant results and a summary report synthesizing the result types and providing prostate cancer screening recommendations for the patient and provider.

Statistics and reproducibility

This study was designed to develop, validate, and clinically implement a genomic risk model for prostate cancer screening using large, diverse biobank-linked cohorts. Statistical analyses were conducted using Cox proportional hazards models to evaluate associations between polygenic scores, family history, genetic principal components, and prostate cancer outcomes. Model development included internal cross-validation and external replication in multiple cohorts, with effect sizes estimated using hazard ratios and cumulative incidence curves.

Sample sizes were determined by the availability of eligible participants in the Million Veteran Program (MVP), PRACTICAL Consortium, and All of Us Research Program datasets. No statistical method was used to predetermine sample size. No data were excluded from the analyses unless flagged for failing quality control criteria, being from related individuals, or lacking available electronic health record data, as described below. Reproducibility was assessed through internal cross-validation (10 iterations of 10-fold cross-validation) and external validation in independent cohorts. Analytic validity of the BGE platform was confirmed by comparison with reference samples and repeated testing across sequencing runs. All statistical analyses were performed using R (v4.4) and relevant packages. The statistical analyses in this study primarily utilized Cox proportional hazards models and related approaches. Data distribution, including normality and equal variances, were formally tested and met model assumptions. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Extended Data

Extended Data Fig. 1.

Extended Data Fig. 1

Positive predictive value of PSA in ProtecT by P-CARE values stratified by PSA values

Extended Data Fig. 2.

Extended Data Fig. 2

Percentage of true positive cases in ProtecT across P-CARE categories

Extended Data Fig. 3.

Extended Data Fig. 3

Odds of prostate cancer in All of Us Research Program by P-CARE category

Supplementary Material

Genetic loci
Supplemental Tables 1-14
Suppelmentary Information: Template of laboratory report package for ProGRESS clinical trial, and MVP Consortium Members

ACKNOWLEDGEMENTS

This research was funded by the U.S. Department of Veterans Affairs Office of Research and Development (I01 CX001727 to RLH and I01 CX002635 to JLV), which played no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript. It was supported using resources and facilities of the Department of Veterans Affairs (VA) Informatics and Computing Infrastructure (VINCI) ORD 24-VINCI-01, including writing support from Kathryn Pridgen, under the research priority to Put VA Data to Work for Veterans (VA ORD 24-D4V). Funding for salaries includes: Department of Veterans Affairs (VISN22 Veterans Center of Excellence for Stress and Mental Health to RLH), VA Office of Research and Development (1I01CX002709, 1I01CX002622 to KNM), National Institutes of Health (R01AG050595 to RLH, K08CA215312 to KNM), the Department of Defense (DOD/CDMRP PC220521 to TMS), the Prostate Cancer Foundation (23CHAL12 to TMS, 20YOUN02 to KNM, 22CHAL02 to IPG, BSR, KNM), the Burroughs Wellcome Foundation (#1017184 to KNM), Basser Center for BRCA (KNM). The authors thank the Million Veteran Program (MVP) staff, researchers, and volunteers, who have contributed to MVP, and especially who previously served their country in the military and now generously agreed to enroll in the study (see mvp.va.gov for more information). The underlying work was based on data from the Million Veteran Program, Office of Research and Development, Veterans Health Administration, and was supported by the Veterans Administration MVP award #000. This publication does not represent the views of the Department of Veterans Affairs or the United States Government.

CAP: The CAP trial was funded by grants C11043/A4286, C18281/A8145, C18281/A11326, C18281/A15064; and C18281/A24432 from Cancer Research UK. The UK Department of Health, National Institute of Health Research provided partial funding.

ProtecT: The ProtecT trial was funded by project grants 96/20/06 and 96/20/99 from the UK National Institute for Health Research, Health Technology Assessment Programme. RMM is a National Institute for Health Research Senior Investigator (NIHR202411). RMM is supported by a Cancer Research UK 25 (C18281/A29019) programme grant (the Integrative Cancer Epidemiology Programme). RMM is also supported by the NIHR Bristol Biomedical Research Centre which is funded by the NIHR (BRC-1215-20011) and is a partnership between University Hospitals Bristol and Weston NHS Foundation Trust and the University of Bristol. Department of Health and Social Care disclaimer: The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care. AV is supported by Spanish Instituto de Salud Carlos III (ISCIII) funding, an initiative of the Spanish Ministry of Economy and Innovation partially supported by European Regional Development FEDER Funds (PI22/00589, INT24/00023, DTS24/00083, PI25/00744); and by the AECC (PRYES211091VEGA). ASK is supported by National Institutes of Health, Grant/Award Numbers: U01 - U01CA268810.

CRUK and PRACTICAL Consortium: This work was supported by the Canadian Institutes of Health Research, European Commission's Seventh Framework Programme grant agreement n° 223175 (HEALTH-F2-2009-223175), Cancer Research UK Grants C5047/A7357, C1287/A10118, C1287/A16563, C5047/A3354, C5047/A10692, C16913/A6135, and The National Institute of Health (NIH) Cancer Post-Cancer GWAS initiative grant: No. 1 U19 CA 148537-01 (the GAME-ON initiative). We would also like to thank the following for funding support: The Institute of Cancer Research and The Everyman Campaign, The Prostate Cancer Research Foundation, Prostate Research Campaign UK (now PCUK), The Orchid Cancer Appeal, Rosetrees Trust, The National Cancer Research Network UK, The National Cancer Research Institute (NCRI) UK. We are grateful for support of NIHR funding to the NIHR Biomedical Research Centre at The Institute of Cancer Research, The Royal Marsden NHS Foundation Trust, and Manchester NIHR Biomedical Research Centre. The Prostate Cancer Program of Cancer Council Victoria also acknowledge grant support from The National Health and Medical Research Council, Australia (126402, 209057, 251533, , 396414, 450104, 504700, 504702, 504715, 623204, 940394, 614296,), VicHealth, Cancer Council Victoria, The Prostate Cancer Foundation of Australia, The Whitten Foundation, PricewaterhouseCoopers, and Tattersall’s. EAO, DMK, and EMK acknowledge the Intramural Program of the National Human Genome Research Institute for their support. Genotyping of the OncoArray was funded by the US National Institutes of Health (NIH) [U19 CA 148537 for ELucidating Loci Involved in Prostate cancer SuscEptibility (ELLIPSE) project and X01HG007492 to the Center for Inherited Disease Research (CIDR) under contract number HHSN268201200008I]. Additional analytic support was provided by NIH NCI U01 CA188392 (PI: Schumacher). Research reported in this publication also received support from the National Cancer Institute of the National Institutes of Health under Award Numbers U10 CA37429 (CD Blanke), and UM1 CA182883 (CM Tangen/IM Thompson). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Funding for the iCOGS infrastructure came from: the European Community's Seventh Framework Programme under grant agreement n° 223175 (HEALTH-F2-2009-223175) (COGS), Cancer Research UK (C1287/A10118, C1287/A 10710, C12292/A11174, C1281/A12014, C5047/A8384, C5047/A15007, C5047/A10692, C8197/A16565), the National Institutes of Health (CA128978) and Post-Cancer GWAS initiative (1U19 CA148537, 1U19 CA148065 and 1U19 CA148112 - the GAME-ON initiative), the Department of Defence (W81XWH-10-1-0341), the Canadian Institutes of Health Research (CIHR) for the CIHR Team in Familial Risks of Breast Cancer, Komen Foundation for the Cure, the Breast Cancer Research Foundation, and the Ovarian Cancer Research Fund. PROtEuS received funding from the Canadian Cancer Society, the Cancer Research Society, the Fonds de Recherche du Québec–Santé, the Ministère du Développement Économique, de l’Innovation et de l’Exportation du Québec and the Canada Research Chairs Program

BPC3: The BPC3 was supported by the U.S. National Institutes of Health, National Cancer Institute (cooperative agreements U01-CA98233 to D.J.H., U01-CA98710 to S.M.G., U01-CA98216 to E.R., and U01-CA98758 to B.E.H., and Intramural Research Program of NIH/National Cancer Institute, Division of Cancer Epidemiology and Genetics).

CAPS: CAPS GWAS study was supported by the Cancer Risk Prediction Center (CRisP; www.crispcenter.org), a Linneus Centre (Contract ID 70867902) financed by the Swedish Research Council, (grant no K2010-70X-20430-04-3), the Swedish Cancer Foundation (grant no 09-0677), the Hedlund Foundation, the Soederberg Foundation, the Enqvist Foundation, ALF funds from the Stockholm County Council. Stiftelsen Johanna Hagstrand och Sigfrid Linner's Minne, Karlsson's Fund for urological and surgical research.

PEGASUS: PEGASUS was supported by the Intramural Research Program, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health.

COMPETING INTERESTS STATEMENT

NL has received speaking honoraria from Illumina Inc and is an advisory board member for FYR Diagnostics and Everygene; NL has received research collaborative funding (for work unrelated to this publication) from Illumina Inc and PacBio Inc. JAL, KML, and CTC report grants from Alnylam Pharmaceuticals, Inc., Astellas Pharma, Inc., AstraZeneca Pharmaceuticals LP, Biodesix, Inc, Celgene Corporation, Cerner Enviza, GSK PLC, IQVIA Inc., Janssen Pharmaceuticals, Inc., Novartis International AG, Parexel International Corporation through the University of Utah or Western Institute for Veteran Research outside the submitted work. ASK reports funding (for work unrelated to this publication) from Janssen, Pfizer, Profound, Bristol Myers Squibb, and Merck. SLD reports grants from AstraZeneca Pharmaceuticals, Biodesix, Myriad Genetic Laboratories, Parexel, Moderna, GlaxoSmithKline, Cerner Enviza, Janssen Research & Development, Celgene, Novartis Pharmaceuticals, IQVIA, Astellas Pharma, and Alnylam Pharmaceuticals. RAE reports speaking honoraria from GU-ASCO, Janssen, University of Chicago, and Dana Farber Cancer Institute, educational honorarium from Bayer and Ipsen, being a member of external expert committee to AstraZeneca UK and Member of Active Surveillance Movember Committee and is a member of the Scientific Advisory Board of Our Future Health; she additionally undertakes private practice as a sole trader at The Royal Marsden NHS Foundation Trust and 90 Sloane Street SW1X 9PQ and 280 Kings Road SW3 4NX, London, UK. LAM reports research funding from AstraZeneca to Harvard University; she holds equity in Convergent Therapeutics. TMS reports honoraria from Varian Medical Systems, WebMD, GE Healthcare, and Janssen; he has an equity interest in CorTechs Labs, Inc. and serves on its Scientific Advisory Board; he receives research funding from GE Healthcare through the University of California San Diego. These companies might potentially benefit from the research results. The terms of this arrangement have been reviewed and approved by the University of California San Diego in accordance with its conflict-of-interest policies. The other authors have no disclosures.

CONSORTIA

Profile Steering Committee

Rosalind A. Eeles1, Elizabeth K Bancroft1,2, Eva McGrowder1, Zsofia Kote-Jarai1, Mark N. Brook1, Elizabeth C Page1, Jana McHugh1, Holly Ni Raghallaigh1, Denzil James2, Pardeep Kumar2, Steve Hazell2, Aslam Sohaib2, Alexander Dias1, Christos Mikropoulos1, Netty Kinsella2, Declan Cahill2

1 Oncogenetics Team, Genetics and Epidemiology, The Institute of Cancer Research, 123 Old Brompton Road, London, UK2 The Royal Marsden NHS Foundation Trust, London, UK

IMPACT Study Steering Committee and Collaborators

Rosalind A. Eeles1, Elizabeth K Bancroft1,2, Elizabeth C Page1, Mark N. Brook1, Zsofia Kote-Jarai1, D. Gareth Evans3, Geoffrey J. Lindeman4,5,6, Paul James4, Lucy Side7, Karina Rønlund8, Brian T. Helfand9, Cezary Cybulski10, Kai-Ren Ong11, Monica Salinas12, Jackie Cook13, Kara Maxwell14, Manuel R. Teixeira15, Rosemarie Davidson16, Annelie Liljegren17, Ashraf Azzabi18, Marc Tischkowitz19, Ruth Cleaver20, Julian Barwell21, Neil K. Aaronson22, Audrey Ardern-Jones2, Chris H. Bangma23, Elena Castro24, David Dearnaley25, Diana M. Eccles26,7, Alison Falconer27, Henrik Gronberg28, Freddie C. Hamdy29,30, Vincent Khoo2,31,25,6, Hans Lilja32,33, Jan Lubinski10, Christos Mikropoulos34, Anita Mitra35, Judith Offman36, Gad Rennert37, Mohnish Suri38

1 Oncogenetics Team, Genetics and Epidemiology, The Institute of Cancer Research, 123 Old Brompton Road, London, UK

2 Cancer Genetics Unit & Academic Urology Unit, The Royal Marsden NHS Foundation Trust, London, UK

3 Genomic medicine, Division of Evolution infection and genomic sciences, University of Manchester, Manchester, UK

4 Parkville Familial Cancer Centre, Peter MacCallum Cancer Centre and The Royal Melbourne Hospital,, Melbourne, VIC, Australia

5 Cancer Biology and Stem Cells Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, VIC, Australia

6 Department of Medicine, The University of Melbourne, Parkville, VIC, Australia

7 Wessex Clinical Genetics Service, Princess Anne Hospital, Southampton, UK

8 Department of Clinical Genetics, University Hospital of Southern Denmark, Vejle Hospital, Vejle, Denmark

9 John and Carol Walter Center for Urological Health, Division of Urology, Endeavor Health NorthShore University HealthSystem, Evanston, IL, USA

10 Department of Genetics and Pathology, International Hereditary Cancer Center, Pomeranian Medical University in Szczecin, Szczecin, Poland

11 Department of Clinical Genetics, Birmingham Women's Hospital, Birmingham, UK

12 Hereditary Cancer Program, Catalan Institute of Oncology, ICO-IDIBELL (Bellvitge Biomedical Research Institute), 08908 Barcelona, CIBERONC, Madrid, Spain

13 Sheffield Clinical Genetics Service, Sheffield Children's NHS Foundation Trust, Sheffield, UK

14 Basser Research Center, University of Pennsylvania, Philadelphia, PA, USA

15 Department of Laboratory Genetics and IPO Porto Research Center, Portuguese Oncology Institute (IPO Porto), Porto, Portugal

16 West of Scotland Genetic Service, Queen Elizabeth University Hospital, Glasgow, UK

17 Karolinska University Hospital and Karolinska Institutet, Stockholm, Sweden

18 Northern Genetics Service, Newcastle upon Tyne Hospitals, UK

19 East Anglian Medical Genetics Service, Cambridge University Hospitals NHS Trust, Cambridge, UK

20 Department of Clinical Genetics, Royal Devon and Exeter Hospital, Exeter, UK

21 University of Leicester, Leicester, UK

22 Division of Psychosocial Research and Epidemiology, Netherlands Cancer Institute, Amsterdam, The Netherlands

23 Department of urology, Erasmus University Medical Center, Dr. Molewaterplein 40, Rotterdam, The Netherlands

24 Hospital Universitario 12 de Octubre, Madrid, Spain

25 Division of Radiotherapy and Imaging, The Institute of Cancer Research, 123 Old Brompton Road, London, UK

26 Cancer Sciences, The University of Southampton, Southampton, UK

27 Imperial College Healthcare NHS Trust, London, UK

28 Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden

29 Churchill Hospital, Oxford, UK

30 Nuffield Department of Surgical Sciences, University of Oxford, Oxford, UK

31 St George’s Hospital, Tooting, London, UK

32 Department of Translational Medicine, Lund University, Malmö, Sweden

33 Departments of Pathology and Laboratory Medicine, Surgery and Medicine, Memorial Sloan-Kettering Cancer Center, New York, NY, USA

34 Royal Surrey Hospital, Guildford, UK

35 University College London Hospitals NHS Foundation Trust, London, UK

36 Centre for Cancer Screening, Prevention and Early Diagnosis, Wolfson Institute of Population Health, Queen Mary University of London, Charterhouse Square, London, UK

37 Technion-Israel Institute of Technology, Haifa, Israel

38 Nottingham City Hospital, Nottingham, UK

PRACTICAL Consortium

Rosalind A. Eeles1,2, Christopher A. Haiman3, Zsofia Kote-Jarai1, Fredrick R. Schumacher4,5, Sara Benlloch6,1, Ali Amin Al Olama6,7, Kenneth R. Muir8, Sonja I. Berndt9, David V. Conti3, Fredrik Wiklund10, Stephen Chanock9, Ying Wang11, Catherine M. Tangen12, Jyotsna Batra13,14,15, Judith A. Clements13,14,15, APCB BioResource (Australian Prostate Cancer BioResource)16,17, Henrik Grönberg10, Nora Pashayan18,19, Johanna Schleutker20,21, Demetrius Albanes9, Stephanie J. Weinstein9, Alicja Wolk22, Catharine M. L. West23, Lorelei A. Mucci24, Géraldine Cancel-Tassin25,26, Stella Koutros9, Karina Dalsgaard Sørensen27,28, Eli Marie Grindedal29, David E. Neal30,31,32, Freddie C. Hamdy33,34, Jenny L. Donovan35, Ruth C. Travis36, Robert J. Hamilton37,38, Sue Ann Ingles39, Barry S. Rosenstein40, Yong-Jie Lu41, Graham G. Giles42,43,44, Robert J. MacInnis42,43, Adam S. Kibel45, Ana Vega46,47,48, Manolis Kogevinas49,50,51,52, Kathryn L. Penney53, Jong Y. Park54, Janet L. Stanford55,56, Cezary Cybulski57, Børge G. Nordestgaard58,59, Sune F. Nielsen58,59, Hermann Brenner60,61, Christiane Maier62, Jeri Kim63, Esther M. John64, Manuel R. Teixeira65,66,67, Susan L. Neuhausen68, Kim De Ruyck69, Azad Razack70, Lisa F. Newcomb55,71, Davor Lessel72, Radka Kaneva73, Nawaid Usmani74,75, Frank Claessens76, Paul A. Townsend77,78, Jose Esteban Castelao79, Monique J. Roobol80, Florence Menegaux81, Kay-Tee Khaw82, Lisa Cannon-Albright83,84, Hardev Pandha78, Stephen N. Thibodeau85, David J. Hunter86, Peter Kraft87, William J. Blot88,89, Elio Riboli90

1The Institute of Cancer Research, London, SM2 5NG, UK

2Royal Marsden NHS Foundation Trust, London, SW3 6JJ, UK

3Center for Genetic Epidemiology, Department of Preventive Medicine, Keck School of Medicine, University of Southern California/Norris Comprehensive Cancer Center, Los Angeles, CA 90015, USA

4Department of Population and Quantitative Health Sciences, Case Western Reserve University, Cleveland, OH 44106-7219, USA

5Seidman Cancer Center, University Hospitals, Cleveland, OH 44106, USA.

6Centre for Cancer Genetic Epidemiology, Department of Public Health and Primary Care, University of Cambridge, Strangeways Research Laboratory, Cambridge CB1 8RN, UK

7University of Cambridge, Department of Clinical Neurosciences, Stroke Research Group, R3, Box 83, Cambridge Biomedical Campus, Cambridge CB2 0QQ, UK

8Division of Population Health, Health Services Research and Primary Care, University of Manchester, Oxford Road, Manchester, M13 9PL, UK

9Division of Cancer Epidemiology and Genetics, National Cancer Institute, NIH, Bethesda, Maryland, 20892, USA

10Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, SE-171 77 Stockholm, Sweden

11Department of Population Science, American Cancer Society, 250 Williams Street, Atlanta, GA 30303, USA

12SWOG Statistical Center, Fred Hutchinson Cancer Research Center, Seattle, WA 98109, USA

13School of Biomedical Sciences, Faculty of Health, Queensland University of Technology, Brisbane, Queensland, Australia

14Centre for Genomics and Personalised Health, Queensland University of Technology, Brisbane, Queensland, Australia

15Translational Research Institute, QUT, Woolloongabba, Brisbane, Queensland, Australia

16Centre for Genomics and Personalised Health, Queensland University of Technology, Brisbane; Prostate Cancer Research Program, Monash University, Melbourne; Dame Roma Mitchell Cancer Centre, University of Adelaide, Adelaide; Chris O'Brien Lifehouse and The Kinghorn Cancer Centre, Sydney, Australia

17Translational Research Institute, Brisbane, Queensland 4102, Australia

18Department of Public Health & Primary Care, Strangeways Research Laboratory, Worts Causeway, Cambridge, CB1 8RN

19Department of Applied Health Research, University College London, London, WC1E 7HB, UK

20Institute of Biomedicine, University of Turku, Finland

21Department of Medical Genetics, Genomics, Laboratory Division, Turku University Hospital, PO Box 52, 20521 Turku, Finland.

22Institute of Environmental Medicine, Karolinska Institutet, 177 77 Stockholm, Sweden

23Division of Cancer Sciences, University of Manchester, Manchester Academic Health Science Centre, Radiotherapy Related Research, The Christie Hospital NHS Foundation Trust, Manchester, M13 9PL UK

24Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, MA 02115, USA

25CeRePP, Tenon Hospital, F-75020 Paris, France.

26Sorbonne Universite, GRC n°5 , AP-HP, Tenon Hospital, 4 rue de la Chine, F-75020 Paris, France

27Department of Molecular Medicine, Aarhus University Hospital, Palle Juul-Jensen Boulevard 99, 8200 Aarhus N, Denmark

28Department of Clinical Medicine, Aarhus University, DK-8200 Aarhus N

29Department of Medical Genetics, Oslo University Hospital, 0424 Oslo, Norway

30Nuffield Department of Surgical Sciences, University of Oxford, Room 6603, Level 6, John Radcliffe Hospital, Headley Way, Headington, Oxford, OX3 9DU, UK

31University of Cambridge, Department of Oncology, Box 279, Addenbrooke's Hospital, Hills Road, Cambridge CB2 0QQ, UK

32Cancer Research UK, Cambridge Research Institute, Li Ka Shing Centre, Cambridge, CB2 0RE, UK

33Nuffield Department of Surgical Sciences, University of Oxford, Oxford, OX1 2JD, UK

34Faculty of Medical Science, University of Oxford, John Radcliffe Hospital, Oxford, UK

35Population Health Sciences, Bristol Medical School, University of Bristol, BS8 2PS, UK

36Cancer Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Oxford, OX3 7LF, UK

37Dept. of Surgical Oncology, Princess Margaret Cancer Centre, Toronto ON M5G 2M9, Canada

38Dept. of Surgery (Urology), University of Toronto, Canada

39Department of Preventive Medicine, Keck School of Medicine, University of Southern California/Norris Comprehensive Cancer Center, Los Angeles, CA 90015, USA

40Department of Radiation Oncology and Department of Genetics and Genomic Sciences, Box 1236, Icahn School of Medicine at Mount Sinai, One Gustave L. Levy Place, New York, NY 10029, USA

41Centre for Cancer Biomarker and Biotherapeutics, Barts Cancer Institute, Queen Mary University of London, John Vane Science Centre, Charterhouse Square, London, EC1M 6BQ, UK

42Cancer Epidemiology Division, Cancer Council Victoria, 200 Victoria Parade, East Melbourne, VIC, 3002, Australia43Centre for Epidemiology and Biostatistics, Melbourne School of Population and Global Health, The University of Melbourne, Grattan Street, Parkville, VIC 3010, Australia

44Precision Medicine, School of Clinical Sciences at Monash Health, Monash University, Clayton, Victoria 3168, Australia

45Division of Urologic Surgery, Brigham and Womens Hospital, 75 Francis Street, Boston, MA 02115, USA

46Fundación Pública Galega Medicina Xenómica, Santiago de Compostela, 15706, Spain.

47Instituto de Investigación Sanitaria de Santiago de Compostela, Santiago de Compostela, 15706, Spain.

48Centro de Investigación en Red de Enfermedades Raras (CIBERER), Spain

49ISGlobal, Barcelona, Spain

50IMIM (Hospital del Mar Medical Research Institute), Barcelona, Spain

51Universitat Pompeu Fabra (UPF), Barcelona, Spain

52CIBER Epidemiología y Salud Pública (CIBERESP), 28029 Madrid, Spain

53Channing Division of Network Medicine, Department of Medicine, Brigham and Women's Hospital/Harvard Medical School, Boston, MA 02115, USA

54Department of Cancer Epidemiology, Moffitt Cancer Center, 12902 Magnolia Drive, Tampa, FL 33612, USA

55Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, Washington, 98109-1024, USA

56Department of Epidemiology, School of Public Health, University of Washington, Seattle, Washington 98195, USA

57International Hereditary Cancer Center, Department of Genetics and Pathology, Pomeranian Medical University, 70-115 Szczecin, Poland

58Faculty of Health and Medical Sciences, University of Copenhagen, 2200 Copenhagen, Denmark

59Department of Clinical Biochemistry, Herlev and Gentofte Hospital, Copenhagen University Hospital, Herlev, 2200 Copenhagen, Denmark

60Division of Clinical Epidemiology and Aging Research, German Cancer Research Center (DKFZ), D-69120, Heidelberg, Germany

61German Cancer Consortium (DKTK), German Cancer Research Center (DKFZ), D-69120 Heidelberg, Germany

62Humangenetik Tuebingen, Paul-Ehrlich-Str 23, D-72076 Tuebingen, Germany

63The University of Texas M. D. Anderson Cancer Center, Department of Genitourinary Medical Oncology, 1515 Holcombe Blvd., Houston, TX 77030, USA

64Departments of Epidemiology & Population Health and of Medicine, Division of Oncology, Stanford Cancer Institute, Stanford University School of Medicine, Stanford, CA 94304 USA

65Department of Laboratory Genetics, Portuguese Oncology Institute of Porto (IPO Porto) / Porto Comprehensive Cancer Center, Porto, Portugal

66Cancer Genetics Group, IPO Porto Research Center (CI-IPOP) / RISE@CI-IPOP (Health Research Network), Portuguese Oncology Institute of Porto (IPO Porto) / Porto Comprehensive Cancer Center, Porto, Portugal

67School of Medicine and Biomedical Sciences (ICBAS), University of Porto, Porto, Portugal

68Department of Population Sciences, Beckman Research Institute of the City of Hope, 1500 East Duarte Road, Duarte, CA 91010

69Ghent University, Faculty of Medicine and Health Sciences, Basic Medical Sciences, Proeftuinstraat 86, B-9000 Gent

70Department of Surgery, Faculty of Medicine, University of Malaya, 50603 Kuala Lumpur, Malaysia

71Department of Urology, University of Washington, 1959 NE Pacific Street, Box 356510, Seattle, WA 98195, USA

72Institute of Human Genetics, University of Regensburg, and Institute of Clinical Human Genetics, University Hospital Regensburg, Franz-Josef-Strauss-Allee 11, D-93053 Regensburg

73Molecular Medicine Center, Department of Medical Chemistry and Biochemistry, Medical University of Sofia, Sofia, 2 Zdrave Str., 1431 Sofia, Bulgaria

74Department of Oncology, Cross Cancer Institute, University of Alberta, 11560 University Avenue, Edmonton, Alberta, Canada T6G 1Z2

75Division of Radiation Oncology, Cross Cancer Institute, 11560 University Avenue, Edmonton, Alberta, Canada T6G 1Z2

76Molecular Endocrinology Laboratory, Department of Cellular and Molecular Medicine, KU Leuven, BE-3000, Belgium

77Division of Cancer Sciences, Manchester Cancer Research Centre, Faculty of Biology, Medicine and Health, Manchester Academic Health Science Centre, NIHR Manchester Biomedical Research Centre, Health Innovation Manchester, Univeristy of Manchester, M13 9WL

78The University of Surrey, Guildford, Surrey, GU2 7XH, UK

79Genetic Oncology Unit, CHUVI Hospital, Complexo Hospitalario Universitario de Vigo, Instituto de Investigación Biomédica Galicia Sur (IISGS), 36204, Vigo (Pontevedra), Spain

80Department of Urology, Erasmus University Medical Center, Cancer Institute, 3015 GD Rotterdam, The Netherlands

81"Exposome and Heredity", CESP (UMR 1018), Faculté de Médecine, Université Paris-Saclay, Inserm, Gustave Roussy, Villejuif

82Clinical Gerontology Unit, University of Cambridge, Cambridge, CB2 2QQ, UK

83Division of Epidemiology, Department of Internal Medicine, University of Utah School of Medicine, Salt Lake City, Utah 84132, USA

84George E. Wahlen Department of Veterans Affairs Medical Center, Salt Lake City, Utah 84148, USA

85Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester, MN 55905, USA

86Nuffield Department of Population Health, University of Oxford, United Kingdom

87Program in Genetic Epidemiology and Statistical Genetics, Department of Epidemiology, Harvard School of Public Health, Boston, MA, USA

88Division of Epidemiology, Department of Medicine, Vanderbilt University Medical Center, 2525 West End Avenue, Suite 800, Nashville, TN 37232 USA.

89International Epidemiology Institute, Rockville, MD 20850, USA

90Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, SW7 2AZ, UK

VA Million Veteran Program

Sumitra Muralidhar1, Jennifer Moser1, Jennifer E. Deen1, Philip S. Tsao2, J. Michael Gaziano3, Adriana Hung4, Dave Oslin5, Deepak Voora6, Jessica V. Brewer3, Mary T. Brophy3, Kelly Cho3, Lori Churby2, Jacob T. Kean7, Saiju Pyarajan3, Robert Ringer8, Luis E. Selva3, Shahpoor (Alex) Shayan3, Brady Stephens9, Stacey B. Whitbourne3

1US Department of Veterans Affairs, Washington, DC

2VA Palo Alto Health Care System, Palo Alto, CA

3VA Boston Healthcare System, Boston, MA

4VA Tennessee Valley Healthcare System, Nashville, TN

5Philadelphia VA Medical Center, Philadelphia, PA

6Durham VA Medical Center, Durham, NC

7VA Salt Lake City Health Care System, Salt Lake City, UT

8New Mexico VA Health Care System, Albuquerque, NM

9Canandaigua VA Medical Center, Canandaigua, NY

DATA AVAILABILITY

The data generated from our analyses are included in the text, tables, figures, and supplemental information. The genetic loci included in the polygenic score and their effect sizes are included in the supplemental information. Source data for Figures 2 and 3 and Extended Data Figures 1 and 2 have been provided as Source Data files. All other data supporting the findings of this study are available from the corresponding author on reasonable request. It is not possible for the authors to share individual-level data from the Million Veteran Program (MVP) due to constraints stipulated in the informed consent. Anyone wishing to gain access to this data should inquire directly to MVP (MVPLOI@va.gov). Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome (PRACTICAL) Consortium data are available upon request to the Data Access Committee (http://practical.icr.ac.uk/blog). Data from the All of Us Research Program are accessible through the Researcher Workbench to researchers with an approved Data Use and Registration Agreement.

CODE AVAILABILITY

The code used for analyses is available at https://github.com/precimed/MVP-PCa-PHS.

REFERENCES

  • 1.Khan SS et al. Development and Validation of the American Heart Association’s PREVENT Equations. Circulation 149, 430–449 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Huntley C et al. Utility of polygenic risk scores in UK cancer screening: a modelling analysis. Lancet Oncol. S1470204523001560 (2023) doi: 10.1016/S1470-2045(23)00156-0. [DOI] [PubMed] [Google Scholar]
  • 3.Barkas F. et al. Advancements in risk stratification and management strategies in primary cardiovascular prevention. Atherosclerosis 395, 117579 (2024). [DOI] [PubMed] [Google Scholar]
  • 4.Hao L et al. Development of a clinical polygenic risk score assay and reporting workflow. Nat. Med 28, 1006–1013 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Lennon NJ et al. Selection, optimization and validation of ten chronic disease polygenic risk scores for clinical implementation in diverse US populations. Nat. Med 1–8 (2024) doi: 10.1038/s41591-024-02796-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Genomics-Enabled Learning Health Care Systems: Gathering and Using Genomic Information to Improve Patient Care and Research: Workshop Summary. (National Academies Press, Washington, D.C., 2015). doi: 10.17226/21707. [DOI] [PubMed] [Google Scholar]
  • 7.Mucci LA et al. Familial Risk and Heritability of Cancer Among Twins in Nordic Countries. JAMA 315, 68–76 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Hall R, Bancroft E, Pashayan N, Kote-Jarai Z & Eeles RA Genetics of prostate cancer: a review of latest evidence. J. Med. Genet. 61, 915–926 (2024). [DOI] [PubMed] [Google Scholar]
  • 9.Grossman DC et al. Screening for Prostate Cancer: US Preventive Services Task Force Recommendation Statement. Jama 319, 1901–1913 (2018). [DOI] [PubMed] [Google Scholar]
  • 10.Garraway IP et al. Prostate Cancer Foundation Screening Guidelines for Black Men in the United States. NEJM Evid. 3, EVIDoa2300289 (2024). [DOI] [PubMed] [Google Scholar]
  • 11.Jackson SD et al. Screening asymptomatic men for prostate cancer: A comparison of international guidelines on prostate-specific antigen testing. J Med Screen 29, 268–271 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Loeb S. et al. Overdiagnosis and overtreatment of prostate cancer. Eur Urol 65, 1046–55 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Ilic D. et al. Prostate cancer screening with prostate-specific antigen (PSA) test: a systematic review and meta-analysis. BMJ 362, k3519 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Hugosson J. et al. A 16-yr Follow-up of the European Randomized study of Screening for Prostate Cancer. Eur Urol 76, 43–51 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Paschen U. et al. Assessment of prostate-specific antigen screening: an evidence-based report by the German Institute for Quality and Efficiency in Health Care. BJU Int 129, 280–289 (2022). [DOI] [PubMed] [Google Scholar]
  • 16.Stone BV et al. The Association of County-level Prostate-specific Antigen Screening with Metastatic Prostate Cancer and Prostate Cancer Mortality. Eur. Urol. Oncol 7, 563–569 (2024). [DOI] [PubMed] [Google Scholar]
  • 17.Iyer HS et al. Access to Prostate-Specific Antigen Testing and Mortality Among Men With Prostate Cancer. JAMA Netw. Open 7, e2414582 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Pagadala MS et al. Polygenic risk of any, metastatic, and fatal prostate cancer in the Million Veteran Program. J Natl Cancer Inst 115, 190–199 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Martin RM et al. Prostate-Specific Antigen Screening and 15-Year Prostate Cancer Mortality: A Secondary Analysis of the CAP Randomized Clinical Trial. JAMA https://doi.org/10.1001/jama.2024.4011 (2024) doi: 10.1001/jama.2024.4011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®) Prostate Cancer Early Detection Version 2.2024. [Google Scholar]
  • 21.Wei JT et al. Early Detection of Prostate Cancer: AUA/SUO Guideline Part I: Prostate Cancer Screening. J Urol 210, 46–53 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Parker C. et al. Prostate cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 31, 1119–1134 (2020). [DOI] [PubMed] [Google Scholar]
  • 23.Mottet N. et al. EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer-2020 Update. Part 1: Screening, Diagnosis, and Local Treatment with Curative Intent. Eur Urol 79, 243–262 (2021). [DOI] [PubMed] [Google Scholar]
  • 24.Huynh-Le M-P et al. Age dependence of modern clinical risk groups for localized prostate cancer-A population-based study. Cancer 126, 1691–1699 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Hou K. et al. Calibrated prediction intervals for polygenic scores across diverse contexts. Nat. Genet. 1–11 (2024) doi: 10.1038/s41588-024-01792-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Moreno-Grau S. et al. Polygenic risk score portability for common diseases across genetically diverse populations. Hum. Genomics 18, 93 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Roth JA, Gulati R, Gore JL, Cooperberg MR & Etzioni R Economic Analysis of Prostate-Specific Antigen Screening and Selective Treatment Strategies. JAMA Oncol. 2, 890–898 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Shoag JE, Nyame YA, Gulati R, Etzioni R & Hu JC Reconsidering the Trade-offs of Prostate Cancer Screening. N. Engl. J. Med 382, 2465–2468 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Bergengren O. et al. 2022 Update on Prostate Cancer Epidemiology and Risk Factors-A Systematic Review. Eur Urol 84, 191–206 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Das H & Rodriguez R Health Care Disparities in Urologic Oncology: A Systematic Review. Urology 136, 9–18 (2020). [DOI] [PubMed] [Google Scholar]
  • 31.Riviere P. et al. Survival of African American and non-Hispanic white men with prostate cancer in an equal-access health care system. Cancer 126, 1683–1690 (2020). [DOI] [PubMed] [Google Scholar]
  • 32.Dess RT et al. Association of Black Race With Prostate Cancer-Specific and Other-Cause Mortality. JAMA Oncol 5, 975–983 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Huynh-Le M-P et al. Prostate cancer risk stratification improvement across multiple ancestries with new polygenic hazard score. Prostate Cancer Prostatic Dis. 1–7 (2022) doi: 10.1038/s41391-022-00497-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Karunamuni RA et al. Performance of African-ancestry-specific polygenic hazard score varies according to local ancestry in 8q24. Prostate Cancer Prostatic Dis 25, 229–237 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Wang A et al. Characterizing prostate cancer risk through multi-ancestry genome-wide discovery of 187 novel risk variants. Nat. Genet. https://doi.org/10.1038/s41588-023-01534-4 (2023) doi: 10.1038/s41588-023-01534-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Kachuri L. et al. Genetically adjusted PSA levels for prostate cancer screening. Nat Med 29, 1412–1423 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Huynh-Le M-P et al. Polygenic hazard score is associated with prostate cancer in multi-ethnic populations. Nat. Commun 12, 1236 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Mavaddat N. et al. Incorporating Alternative Polygenic Risk Scores into the BOADICEA Breast Cancer Risk Prediction Model. Cancer Epidemiol. Biomark. Prev. Publ. Am. Assoc. Cancer Res. Cosponsored Am. Soc. Prev. Oncol 32, 422–427 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Yang X. et al. Prospective validation of the BOADICEA multifactorial breast cancer risk prediction model in a large prospective cohort study. J. Med. Genet 59, 1196–1205 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Darst BF et al. Combined Effect of a Polygenic Risk Score and Rare Genetic Variants on Prostate Cancer Risk. Eur. Urol 80, 134–138 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Hughley RW et al. Polygenic Risk Score Modifies Prostate Cancer Risk of Pathogenic Variants in Men of African Ancestry. Cancer Res. Commun. https://doi.org/10.1158/2767-9764.CRC-23-0022 (2023) doi: 10.1158/2767-9764.CRC-23-0022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Kang JH et al. Polygenic risk and rare variant gene clustering enhance cancer risk stratification for breast and prostate cancers. Commun. Biol 7, 1289 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Callender T. et al. Polygenic risk-tailored screening for prostate cancer: A benefit–harm and cost-effectiveness modelling study. PLOS Med. 16, e1002998 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Callender T, Emberton M, Morris S, Pharoah PDP & Pashayan N Benefit, Harm, and Cost-effectiveness Associated With Magnetic Resonance Imaging Before Biopsy in Age-based and Risk-stratified Screening for Prostate Cancer. JAMA Netw. Open 4, e2037657 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Dixon P, Keeney E, Taylor J, Wordsworth S & Martin R Can polygenic risk scores contribute to cost-effective cancer screening? A systematic review. Genet. Med 24, 1604–1617 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Jiang S. et al. Cost-effectiveness of population-wide genomic screening for Lynch Syndrome and polygenic risk scores to inform colorectal cancer screening. Genet. Med. Off. J. Am. Coll. Med. Genet 27, 101285 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Guzauskas GF et al. Population Genomic Screening for Three Common Hereditary Conditions : A Cost-Effectiveness Analysis. Ann. Intern. Med 176, 585–595 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Lacaze P et al. Combined population genomic screening for three high-risk conditions in Australia: a modelling study. EClinicalMedicine 66, 102297 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Wongvibulsin S, Wu KC & Zeger SL Clinical risk prediction with random forests for survival, longitudinal, and multivariate (RF-SLAM) data analysis. BMC Med. Res. Methodol 20, 1 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

METHODS-ONLY REFERENCES

  • 50.Loef B. et al. Using random forest to identify longitudinal predictors of health in a 30-year cohort study. Sci. Rep 12, 10372 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Gaziano JM et al. Million Veteran Program: A mega-biobank to study genetic influences on health and disease. J Clin Epidemiol 70, 214–223 (2016). [DOI] [PubMed] [Google Scholar]
  • 52.Bick AG et al. Genomic data in the All of Us Research Program. Nature 627, 340–346 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Hunter-Zinck H. et al. Genotyping Array Design and Data Quality Control in the Million Veteran Program. Am. J. Hum. Genet 106, 535–548 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Pagadala MS et al. Agent orange exposure and prostate cancer risk in the million veteran program. Acta Oncol. Stockh. Swed 63, 373–378 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Alba PR et al. Ascertainment of Veterans With Metastatic Prostate Cancer in Electronic Health Records: Demonstrating the Case for Natural Language Processing. JCO Clin. Cancer Inform 1005–1014 (2021) doi: 10.1200/CCI.21.00030. [DOI] [PubMed] [Google Scholar]
  • 56.Amos CI et al. The OncoArray Consortium: A Network for Understanding the Genetic Architecture of Common Cancers. Cancer Epidemiol. Biomark. Prev. Publ. Am. Assoc. Cancer Res. Cosponsored Am. Soc. Prev. Oncol 26, 126–135 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Eeles RA et al. Identification of 23 new prostate cancer susceptibility loci using the iCOGS custom genotyping array. Nat. Genet 45, 385–391, 391e1-2 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Huynh-Le M-P et al. Common genetic and clinical risk factors: association with fatal prostate cancer in the Cohort of Swedish Men. Prostate Cancer Prostatic Dis. 24, 845–851 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Discacciati A et al. Coffee consumption and risk of localized, advanced and fatal prostate cancer: a population-based prospective study. Ann. Oncol 24, 1912–1918 (2013). [DOI] [PubMed] [Google Scholar]
  • 60.Hamdy FC et al. Fifteen-Year Outcomes after Monitoring, Surgery, or Radiotherapy for Prostate Cancer. N. Engl. J. Med 388, 1547–1558 (2023). [DOI] [PubMed] [Google Scholar]
  • 61.Gudmundsson J. et al. Genome-wide associations for benign prostatic hyperplasia reveal a genetic correlation with serum levels of PSA. Nat Commun 9, 4568 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Chen F. et al. Evidence of Novel Susceptibility Variants for Prostate Cancer and a Multiancestry Polygenic Risk Score Associated with Aggressive Disease in Men of African Ancestry. Eur. Urol 84, 13–21 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 63.Karunamuni RA et al. Additional SNPs improve risk stratification of a polygenic hazard score for prostate cancer. Prostate Cancer Prostatic Dis. 24, 532–541 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 64.Tibshirani R Regression Shrinkage and Selection via the Lasso. J. R. Stat. Soc. Ser. B Methodol 58, 267–288 (1996). [Google Scholar]
  • 65.Tibshirani R The lasso method for variable selection in the Cox model. Stat Med 16, 385–95 (1997). [DOI] [PubMed] [Google Scholar]
  • 66.Li Y. et al. FastPop: a rapid principal component derived method to infer intercontinental ancestry using genetic data. BMC Bioinformatics 17, 122 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 67.Auton A. et al. A global reference for human genetic variation. Nature 526, 68–74 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 68.Friedman JH, Hastie T & Tibshirani R Regularization Paths for Generalized Linear Models via Coordinate Descent. J. Stat. Softw 33, 1–22 (2010). [PMC free article] [PubMed] [Google Scholar]
  • 69.Wendt FR et al. Modeling the longitudinal changes of ancestry diversity in the Million Veteran Program. Hum. Genomics 17, 46 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 70.Huynh-Le M-P et al. A Genetic Risk Score to Personalize Prostate Cancer Screening, Applied to Population Data. Cancer Epidemiol. Biomarkers Prev. 29, 1731–1738 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 71.Seibert TM et al. Polygenic hazard score to guide screening for aggressive prostate cancer: development and validation in large scale cohorts. BMJ 360, j5757 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 72.Karunamuni RA et al. African-specific improvement of a polygenic hazard score for age at diagnosis of prostate cancer. Int. J. Cancer 148, 99–105 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 73.Yeh HC, Duncan BB, Schmidt MI, Wang NY & Brancati FL Smoking, smoking cessation, and risk for type 2 diabetes mellitus: a cohort study. Ann Intern Med 152, 10–7 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 74.Brentnall AR, Cuzick J, Buist DSM & Bowles EJA Long-term Accuracy of Breast Cancer Risk Assessment Combining Classic Risk Factors and Breast Density. JAMA Oncol 4, e180174 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 75.Wang TJ et al. Plasma natriuretic peptide levels and the risk of cardiovascular events and death. N Engl J Med 350, 655–63 (2004). [DOI] [PubMed] [Google Scholar]
  • 76.DeFelice M et al. Blended Genome Exome (BGE) as a Cost Efficient Alternative to Deep Whole Genomes or Arrays. 2024.04.03.587209 Preprint at 10.1101/2024.04.03.587209 (2024). [DOI] [Google Scholar]
  • 77.Rubinacci S, Hofmeister RJ, Sousa da Mota B & Delaneau O Imputation of low-coverage sequencing data from 150,119 UK Biobank genomes. Nat. Genet 55, 1088–1090 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 78.Tiao G & Goodrich J gnomAD v3.1 New Content, Methods, Annotations, and Data Availability. https://gnomad.broadinstitute.org/news/2020-10-gnomad-v3-1-new-content-methods-annotations-and-data-availability/#the-gnomad-hgdp-and-1000-genomes-callset (2023). [Google Scholar]
  • 79.Babadi M. et al. GATK-gCNV enables the discovery of rare copy number variants from exome sequencing data. Nat. Genet 55, 1589–1597 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 80.NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®) Genetic/Familial High-Risk Assessment: Colorectal Version 2.2023. [Google Scholar]
  • 81.NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®) Genetic/Familial High-Risk Assessment: Breast, Ovarian, and Pancreatic Version 3.2024. [Google Scholar]
  • 82.Farmer GD, Gray H, Chandratillake G, Raymond FL & Freeman ALJ Recommendations for designing genetic test reports to be understood by patients and non-specialists. Eur. J. Hum. Genet. EJHG 28, 885–895 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Genetic loci
Supplemental Tables 1-14
Suppelmentary Information: Template of laboratory report package for ProGRESS clinical trial, and MVP Consortium Members

Data Availability Statement

The data generated from our analyses are included in the text, tables, figures, and supplemental information. The genetic loci included in the polygenic score and their effect sizes are included in the supplemental information. Source data for Figures 2 and 3 and Extended Data Figures 1 and 2 have been provided as Source Data files. All other data supporting the findings of this study are available from the corresponding author on reasonable request. It is not possible for the authors to share individual-level data from the Million Veteran Program (MVP) due to constraints stipulated in the informed consent. Anyone wishing to gain access to this data should inquire directly to MVP (MVPLOI@va.gov). Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome (PRACTICAL) Consortium data are available upon request to the Data Access Committee (http://practical.icr.ac.uk/blog). Data from the All of Us Research Program are accessible through the Researcher Workbench to researchers with an approved Data Use and Registration Agreement.

The code used for analyses is available at https://github.com/precimed/MVP-PCa-PHS.

RESOURCES