Investigating the Role of Neighborhood Socioeconomic Status and Germline Genetics on Prostate Cancer Risk

Background: Genetic factors play an important role in prostate cancer (PCa) development with polygenic risk scores (PRS) predicting disease risk across genetic ancestries. However, there are few convincing modifiable factors for PCa and little is known about their potential interaction with genetic risk. We analyzed incident PCa cases (n=6,155) and controls (n=98,257) of European and African ancestry from the UK Biobank (UKB) cohort to evaluate the role of neighborhood socioeconomic status (nSES)–and how it may interact with PRS–on PCa risk.

We evaluated a multi-ancestry PCa PRS containing 269 genetic variants to understand the association of germline genetics with PCa in UKB. Using the English Indices of Deprivation, a set of validated metrics that quantify lack of resources within geographical areas, we performed logistic regression to investigate the main effects and interactions between nSES deprivation, PCa PRS, and PCa.

The PCa PRS was strongly associated with PCa (OR=2.04; 95%CI=2.00-2.09; P<0.001). Additionally, nSES deprivation indices were inversely associated with PCa: employment (OR=0.91; 95%CI=0.86-0.96; P<0.001), education (OR=0.94; 95%CI=0.83-0.98; P<0.001), health (OR=0.91; 95%CI=0.86-0.96; P<0.001), and income (OR=0.91; 95%CI=0.86-0.96; P<0.001). The PRS effects showed little heterogeneity across nSES deprivation indices, except for the Townsend Index (P=0.03)

We reaffirmed genetics as a risk factor for PCa and identified nSES deprivation domains that influence PCa detection and are potentially correlated with environmental exposures that are a risk factor for PCa. These findings also suggest that nSES and genetic risk factors for PCa act independently. 2 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint

Prostate cancer (PCa) is the second most common cancer in men in the United States, with the highest incidence rates observed in African American men (1). In addition to older age, there is evidence for a strong genetic component in PCa etiology, including family history and germline susceptibility loci (2). Genome-wide association studies (GWAS) have identified 451 independent risk variants for PCa, and these findings have been used to develop a polygenic risk score (PRS). Although the most recent PRS for PCa has improved risk prediction across ancestry groups, performance in African and East Asian populations continues to lag behind European ancestry (3). Individual and neighborhood socioeconomic status (SES) has been explored as a potential risk factor for PCa (4–6), however, studies of neighborhood SES (nSES) have produced mixed results (7–9). Some find a positive relationship between nSES and PCa risk, theorizing that advantaged populations are more likely to undergo PSA screening or engage with the healthcare system (9). Other research has found the opposite effect with lower nSES groups being at higher risk, suggesting a target for public health interventions (8,10). Gene-by-environment (GxE) interactions are thought to play an important role in complex trait variation across populations (11). In addition to interactions between environmental factors and specific risk variants, there is accumulating evidence that PRS performance is context-dependent and may be modulated by SES factors (12). In this study, we investigated the main effects and interactions of PCa PRS and nSES using the UK Biobank (UKB), a recently published PCa PRS, and the English Indices of Deprivation.

Study Population The United Kingdom Biobank (UKB) is a prospective study of 503,317 participants aged 40–69 years across England, Wales and Scotland (13). Between 2006 and 2010, the UKB collected genotyping data, linkage to national cancer registries, and demographic and questionnaire data at 22 assessment centers. 3 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint All quality control and phenotyping of data used in our analyses was described and performed in previous studies (14). We restricted our analyses to men with concordant self-reported and genetically inferred sex, available genotyping data, known values for all deprivation indices, and known cancer status histories. Additionally, we excluded men with a cancer diagnosis before assessment were or a post-assessment cancer diagnosis other than PCa or non-melanoma skin cancer. Controls included men without any cancer diagnosis (excluding non-melanoma skin cancer). Analyses were performed separately in populations groups defined based on self-reported ethnicity and genetic principal component (PCs) values within five standard deviations of the UK10K and 1000 Genomes phase 3 reference panel group means. Using this criteria, we analyzed two separate populations throughout this study that we will refer to as the European (5,960 cases and 93,990 controls) and African (109 cases and 1,226 controls) ancestries (14,15). Indices of Deprivation & Townsend Index The English Indices of Deprivation are a series of metrics calculated by the UK Government that assess relative deprivation of geographic areas across seven domains: Income Deprivation; Employment Deprivation; Health Deprivation and Disability; Education, Skills and Training Deprivation; Barriers to Housing and Services; Living Environment Deprivation; and Crime (16). Domain-specific indices are combined into a single Index of Multiple Deprivation (IMD) using predefined weights: Income=22.5%; Employment=22.5%; Health and Disability=13.5%; Education, Skills, and Training=13.5%; Housing and Services=9.3%; Environment=9.3%; Crime=9.3% (17).. The Townsend Index is another nSES deprivation metric composed of four area-based indicators: percentage of households without a car, overcrowded households, households not owner-occupied, and persons unemployed, that are standardized and summed (18). For all UKB participants, values of deprivation indices were assigned based on their postcode and year of assessment. For detailed descriptions, see Technical Report and Section 3.2 of (17,18), respectively. Prostate Cancer PRS 4 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint The PCa PRS used in this analysis consists of 269 risk variants identified using multi-ancestry fine-mapping and summary statistics from a GWAS of 107,247 cases and 127,006 controls across three genetic ancestries (19,20). UKB participants were not included in the GWAS meta-analysis used for training this PRS model. For each individual, i, PRS i= , where gim is the genotype dosage of variant m of M total 𝑚=1 𝑀 ∑ 𝑤 𝑚𝑔𝑖𝑚 variants for individual i and wm is a variant-specific weight (logOR scale). Associations with PCa were estimated separately in each ancestral population using logistic regression with adjustment for age at assessment and the top 10 genetic ancestry principal components (PCs). We estimated odds ratios (OR) per standard deviation (SD) increase in PRS 269, standardized within each ancestry group. We also modeled the PRS as a discrete variable with cut-points based on deciles with the 40-60% group as the reference category. Our primary analysis uses PRS 269, which was not trained on the largest PCa GWAS to date but avoids overlap with UKB. We also present sensitivity analyses using two of the most recent PCa PRS models for which UKB was part of the training data (3): one model with 451 variants, and another developed using PRS-CSx (21). Modeling Neighborhood Socioeconomic Status We used logistic regression to assess the potential association between nSES deprivation, as measured by the English Indices of Deprivation and the Townsend Index, and PCa. Age at assessment was included as a covariate for all models. Additionally, to investigate if any associations between nSES deprivation and PCa are mediated by PSA screening and healthcare utilization (22), we used the “Ever had a PSA test” variable in UKB and calculated a proxy for healthcare utilization by summing the number of general practitioner visits and hospital inpatient episodes in the year prior to assessment (23). To evaluate the relationship between nSES deprivation, PSA screening, and PCa, we performed logistic regression between nSES deprivation and UKB’s “Ever had a PSA test” variable before modeling the association between nSES deprivation and PCa while adjusting for PSA screening and age. We repeated this analysis for 5 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint healthcare utilization by modeling the association between nSES deprivation and number of clinical visits before analyzing the association between nSES deprivation and PCa while adjusting for number of clinical visits and age. Lastly, we calculated the correlation between “Ever had a PSA test”, number of clinical visits, and all nSES variables before estimating the association between nSES deprivation and PCa using PSA screening, number of clinical visits, and age as covariates. Finally, to further investigate the potential impact of high nSES on increasing PSA screening and prostate cancer, we performed a stratified analysis based on self-reported screening status. We then performed a difference of means z-test for each nSES variable between the ever had a PSA test and never had a PSA test group. Due to small sample sizes in the other ancestry groups, we performed this analysis on only individuals of European ancestry. Evaluation of Genetic Effect Modification We investigated whether the PRS 269 associations with PCa were modified by nSES deprivation in two ways. First, we fitted logistic regression models that included the main effects and an interaction between PRS 269 and nSES to estimate their effects on PCa risk. We assessed modification by looking at whether the interaction term was associated with PCa. Second, we evaluated the standardized PRS 269 effect within quartiles of each nSES index. We computed Cochran’s Q to assess heterogeneity in the effect size of the PRS across quartiles. This analysis was limited to men of European ancestry due to smaller sample sizes in the African ancestry (109 cases and 1,226 controls.

Participants Characteristics This study included a total of 6,069 PCa cases and 95,216 cancer-free controls (Table 1). Most participants were of European ancestry (5,960 cases and 93,990 controls), compared to African (109 cases and 1,226 controls) ancestry. In these populations, 30% and 24% of participants reported having had at least one PSA test prior to recruitment. For both ancestries, the neighborhood deprivation scores, 6 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint including the index of multiple deprivation (IMD) and Townsend Index, were lower in the cases compared to the controls. In contrast, the median number of clinical visits the year before assessment was higher for cases than controls in both European (Controls:6; Cases:7) and African (Controls:7; Cases:9) ancestries. Relationship between PRS and Prostate Cancer We reaffirmed the association between PRS 269 and PCa using our specific sample (19). We found PRS 269 to be associated with PCa in both the European (OR=2.04; 95%CI=2.00-2.09; P<0.001) and African (OR=1.35; 95%CI=1.16-1.58; P<0.001) ancestries. We also evaluated this association in two more recently developed PCa PRS models–one with 451 variants and the other using PRS-CSx, to investigate the change in the PRS and PCa association fueled by models from a larger more-diverse GWAS (3). When we did this, we found associations of similar magnitude across all three models for both populations (Supp. Table 1). Even though the secondary PRS models were developed from a larger GWAS than PRS 269, they included UKB as part of their training data. Because this posed a risk of being overfit to our sample, PRS269 was used for all future analyses. When we investigated if the relationship between PRS 269 and PCa risk was linear, we found that the association increased across PRS deciles for men of European ancestry, but not men of African ancestry. Men in the top decile of the PRS distribution had a significantly increased risk of PCa compared to those with average PRS (40-60%) in both the European (OR=3.62; 95%CI=3.38-3.89; P<0.001) and African (OR=1.9;95%CI=1.13-3.20; P=0.02) ancestries. The association between PRS269 and PCa was statistically significant for all deciles in the European ancestry models, ranging from OR=0.26 to OR=3.62. However, in men of African ancestry, only the second highest decile of the PRS distribution had a significantly increased risk compared to the average (OR=1.76; 95%CI=1.04-3.00; P=0.04) (Table 2). The associations in the African ancestry models were also less precise than those in the European models likely due to smaller samples, especially cases, in the lower decile groups. 7 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint Relationship between nSES and Prostate Cancer For each ancestry, we assessed the relationship between nSES deprivation and PCa using a minimally adjusted model. In the European ancestry, six of the nine nSES deprivation indices had inverse associations with prostate cancer: Education (OR=0.92; 95%CI=0.90-0.95; P<0.001), Employment (OR=0.91; 95%CI=0.88-0.93; P<0.001), Health (OR=0.91; 95%CI=0.88-0.93; P<0.001), Income (OR=0.91; 95%CI=0.89-0.94; P<0.001), IMD (OR=0.92; 95%CI=0.89-0.95; P<0.001), and Townsend Index (OR=0.93; 95%CI=0.90-0.96; P<0.001) (all ORs for a SD unit increase in the deprivation index) (Figure 1a, Supp. Table 2). These results signify that higher deprivation and lower neighborhood resources are associated with lower PCa risk. In the European ancestry, the Housing/Services index (OR=1.04; 95%CI=1.01-1.06; P<0.001) was the only nSES domain with a positive association, and we found no associations between the Crime and Environment indices in this group. In men of African ancestry, none of the nSES deprivation indices were associated with PCa (Figure 1b, Supp. Table 2). PSA Screening and Healthcare Utilization Mediation Differences in PSA screening and healthcare utilization may be mediating observed associations between nSES and PCa (22). Similar to the relationship between nSES deprivation and PCa, the Housing/Services index (OR=1.13; 95%CI =1.12-1.15; P<0.001) was positively associated with PSA screening in men of European ancestry. In this ancestry, the remaining eight indices were negatively associated with screening, with effects ranging from OR=0.93 (Housing and Education) to OR=0.97 (Environment). In the African ancestry models, the IMD, Income, Employment, Health, Education, and Townsend indices, were also negatively associated with ever having a PSA test. The remaining indices, including Housing/Services, were not associated with PSA screening (Supp. Table 3). Inversely, all deprivation indices, except for Housing/Services (Beta=-0.33; 95%CI=-0.4--0.25; P<0.001), were positively associated with the number of clinical visits in men of European ancestry. The IMD (Beta=1.07; 95%CI=0.21-1.06; P=0.02), Health (Beta=1.06; 95%CI=0.10-2.02; P=0.03), Education (Beta=0.81; 8 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint 95%CI=0.06-1.56; P=0.03), and Environment (Beta=1.00; 95%CI=0.16-1.84; P=0.02) indices were also positively associated with number of clinical visits in men of African ancestry (Supp. Table 4). When we calculated Pearson's correlation between PSA screening, clinical visits, and the nSES indices, PSA screening was inversely correlated with all nSES indices while clinical visits were positively correlated with them. The exception in both cases was the Housing/Services index that had a correlation in the opposite direction (Supp. Table 5). When included as covariates in the model between nSES deprivation and PCa, “Ever PSA Screened” was associated with PCa across models, while the number of clinical visits was not in any model in either ancestry. Additionally, the effect size of the nSES variables remained consistent after adjusting for the number of clinical visits and PSA screening status, including the statistically significant European ancestry associations: Education (OR=0.94; 95%CI=0.89-0.98; P=0.006), Employment (OR=0.91; 95%CI=0.86-0.96; P<0.001), Health (OR=0.91; 95%CI= 0.86-0.96; P<0.001), Income (OR=0.91; 95%CI=0.86-0.96; P<0.001), IMD (OR=0.92; 95%CI=0.87-0.96; P=0.001), and Townsend Index (OR=0.94; 95%CI=0.89-0.98; P=0.009) (Supp. Table 6-8). Since PSA screening was associated with PCa when modeled with nSES, we undertook an additional analysis stratified by screening to reduce residual confounding remaining in the logistic regression model. In men of European ancestry, the association between nSES deprivation and prostate cancer remained consistent across screening status. For the never-screened individuals, the associations between nSES and PCa varied between OR=0.91 (IMD) to OR=0.94 (Townsend). We found no associations using the Crime and Environment index, and while not statistically significant, the Housing/Services index was the only variable with a positive association with PCa (OR=1.03; 95%CI=0.99-1.07; P=0.12)). In terms of the all screened group, IMD (OR=0.95; 95%CI=0.91-0.99; P=0.17), Income (OR=0.94; 95%CI=0.90-0.98; P=0.007), Employment (OR=0.94; 95%CI=0.90-0.98; P=0.003), Education (OR=0.95; 95%CI=0.92-1.00; P=0.32), and Townsend (OR=0.93; 95%CI=0.90-0.97; P=0.001) indices were also inversely associated with PCa. The nSES variable with the largest difference in effects was the Crime index 9 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint (ORnever-screened=0.97; 95%CI=0.93-1.01; P=0.10 & OR screened=1.01; 95%CI=0.97-1.06; P=0.43) (Figure 2, Supp. Table 9). PRS-by-nSES Effect Modification When analyzing a potential GxE interaction between PRS 269 and nSES deprivation, we found no effect modification across all indices for both the European (P>0.14) and African (P>0.56) ancestries (Supp. Table 10). To more thoroughly analyze if there were no GxE interactions, we compared the PRS 269 association across nSES quartiles in men of European ancestry. When we did this, we found heterogeneity in the PRS 269 associations across Townsend index quartiles (P=0.03) with associations ranging between OR=2.13 (Deprivation Quartile 1) to OR=1.93 (Deprivation Quartile 2) (Table 3). This analysis wasn’t performed in men of African ancestry due to the reduced power that would occur from stratifying an already small number of cases.

We investigated potential main effects and interactions between a well-established risk factor, germline genetics, and a less well-studied exposure, nSES as measured through deprivation, and PCa. Previous studies analyzing nSES and PCa reported conflicting results with some finding a positive association (7,9,24) and others finding a negative association (8,10,25). These variations are likely due to differences in study population, methodology, country of study, healthcare systems, and utilization of different measures of nSES deprivation and covariates. And while the association between nSES and PCa incidence is unclear, studies consistently found nSES deprivation positively associated with both PCa stage and mortality (4,6,9). Our findings align with the subset of studies, performed in the United States, Taiwan, and Scotland that find increasing neighborhood resources (or decreasing deprivation) increases PCa risk, specifically in the domains of education, employment, health, and income (9,24,25). These studies suggest that higher nSES groups are more likely to be PSA screened or utilize the healthcare system (26). While PSA screening was positively associated with 10 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint PCa, healthcare utilization was not. Moreover, nSES was associated with PCa after adjusting for self-reported screening at baseline. One possible explanation for the nSES association is that while PSA screening may lead to overdiagnosis, there may be environmental exposures more common in resourced neighborhoods that increase PCa incidence (2). More research investigating environmental factors associated with nSES and PCa risk is needed to further investigate this. The only nSES deprivation domain positively associated with PCa was the Housing/Services index, implying that men in neighborhoods with less access to housing and services were more likely to be diagnosed with PCa. The mechanism underlying this association, and its opposite effect from other domains, is unknown. It is likely connected to our finding that the Housing/Services index is uncorrelated with the other indices, except for the Environment and Townsend indices that include housing deprivation in their calculation. When we remove housing-relevant indicators, the Housing/Services index is calculated using the average distance to key services, such as intensive medical care, supermarkets, schooling, and post offices. The association between Housing/Services and PCa is likely driven by services deprivation, particularly factors associated with distance to medical care and schooling. The associations between nSES deprivation and PCa were also only seen in the European ancestry group. Due to the smaller number of African ancestry men, it is difficult to ascertain if there is no effect in these groups or if we were underpowered to identify these associations. We also reaffirm the relationship between PCa PRS and PCa risk in a sample of European ancestry men in UKB using a 269 variant PRS model. As previously shown, the PRS association with PCa increased over PRS deciles (14,27,28). These associations and trends were diminished in the African ancestry. Attenuated PRS associations for non-European ancestries, including African ancestries, have been consistently replicated across PCa PRS models (3,19,29,30). Current explanations include differences in linkage disequilibrium patterns, allelic effect size, and causal allele frequency whose effects are exacerbated by smaller African ancestry sample sizes in the GWAS that generate PRS models (19). The lack of PRS associations in 11 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint men of African ancestry is likely attributed to the smaller sample sizes of this population in the UKB. Gene-environment (GxE) interactions are thought to be driven by biological processes or spurious associations, and are of increasing interest for improving genetic discovery and explaining missing heritability and population heterogeneity (31). When we investigated the potential role of gene-nSES interactions to explain PCa heritability and heterogeneity in previous studies, we identified an interaction between PRS269 and the Townsend index. The lack of an interaction for the other index measuring overall deprivation, the Index of Multiple Deprivation, may signify that the interaction with Townsend index is a spurious association. While GxE interactions are challenging to detect due to limited statistical power, the absence of interactions for the other indices may reflect that nSES and PRS act independently. It could also reflect differences in factors affecting PCa development versus those affecting PCa detection. There may be environmental factors that modify genetic susceptibility to PCa, but more work must be done to identify variables that are risk factors themselves for PCa. Future work should also incorporate novel frameworks and techniques to facilitate the discovery of these interactions and explain the remaining variation in PCa risk (31,32). Several key limitations prevent us from interpreting our results more broadly. First, the limited number of African ancestry individuals in UKB, combined with the known decrease in PRS performance for ancestries underrepresented in the training data (30,33), limited our ability to detect effect modification and identify PRS association trends in men of African ancestry. Future studies require larger cohorts that are more diverse in ancestry, personal identity, and socioeconomic background (34). Second, we lacked in-depth cancer stage and screening information to determine the magnitude of our PCa associations and PSA screening effect. For example, our PSA screening variable was a binary self-reported measurement taken at assessment. To avoid measurement error, reporting bias, and other confounding, a more-precise, repeated, and validated variable was required. Lastly, nSES indices are broad measures of deprivation assigned at assessment without considering how long they lived in these neighborhoods. This could lead to exposure misclassification for nSES in the etiologically relevant time frame prior to PCa diagnosis. In addition, individual SES, 12 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint which varies within geographic regions, can add unmeasured confounding to the nSES relationship studied in this research (24,35). Given available data, it was not possible to adjust for this confounding across indices measuring domain-specific deprivation and overall deprivation in a principled manner. While previous research has adjusted for individual SES and identified associations with nSES (8,9), future work should account for an individual’s SES in specific socioeconomic domains and length of exposure in a specific neighborhood. In conclusion, we identified an inverse relationship between several domains of neighborhood deprivation and PCa while accounting for potential mediators, such as PSA screening and healthcare utilization. As the first study to examine domains of nSES in PCa, this research provides an opportunity to investigate non-genetic variation in PCa risk. Additionally, we reaffirmed the association between germline genetics and PCa while attempting to identify novel gene-by-nSES interactions. While such interactions were not identified, more research focused on other exposures, specifically those that modify disease development and not only disease detection, is necessary to improve our knowledge of PCa etiology. Data Availability Statement: The research was conducted with approved access to UK Biobank data under application number 14105 (PI: Witte). UK Biobank data are publicly available by request from https://www.ukbiobank.ac.uk. Informed consent was obtained from all study participants. UK Biobank received ethics approval from the Research Ethics Committee (REC reference: 11/NW/0382). PRS 269 (PRS ID: PGS000662) and PRS 451 (PRS ID: PGS003765) are publicly available on the PGS catalog from https://www.pgscatalog.org/. Funding: This work was supported by the National Human Genome Research Institute of the National Institutes of Health under award No. 5T32HG000044 (to J.J.).

The funder had no role in the design of the study; the collection, analysis, or interpretation of the data; or the writing of the manuscript and decision to submit it for publication. 13 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint Conflicts of Interest: No authors have conflicts to disclose.

1. Centers for Disease Control and Prevention. Prostate Cancer Incidence by Age and Stage at Diagnosis, United States - 2001-2019. Atlanta, GA: Centers for Disease Control and Prevention, US Department of Health and Human Services; 2023. (USCS Data Brief). Report No.: 34. 2. Bergengren O, Pekala KR, Matsoukas K, Fainberg J, Mungovan SF, Bratt O, et al. 2022 Update on Prostate Cancer Epidemiology and Risk Factors—A Systematic Review. Eur Urol [Internet]. 2023 May 16 [cited 2023 Jul 13]; Available from: https://www.sciencedirect.com/science/article/pii/S0302283823027860 3. Wang A, Shen J, Rodriguez AA, Saunders EJ, Chen F, Janivara R, et al. Characterizing prostate cancer risk through multi-ancestry genome-wide discovery of 187 novel risk variants. Nat Genet. 2023 Dec;55(12):2065–74. 4. Coughlin SS. A review of social determinants of prostate cancer risk, stage, and survival. Prostate Int. 2020 Jun 1;8(2):49–54. 5. Riveros C, Al-Toubat M, Chalfant V, Elshafei A, Feibus A, Forero A, et al. The impact of socioeconomic status on the survival of men with early-onset prostate cancer. Am J Clin Exp Urol. 2023 Apr 15;11(2):146–54. 6. DeRouen MC, Schupp CW, Koo J, Yang J, Hertz A, Shariff-Marco S, et al. Impact of individual and neighborhood factors on disparities in prostate cancer survival. Cancer Epidemiol. 2018 Apr 1;53:1–11. 7. Perez-Cornago A, Key TJ, Allen NE, Fensom GK, Bradbury KE, Martin RM, et al. Prospective investigation of risk factors for prostate cancer in the UK Biobank cohort study. Br J Cancer. 2017 Nov;117(10):1562–71. 8. Salmon C, Quesnel-Vallée A, Barnett TA, Benedetti A, Cloutier MS, Datta GD, et al. Neighbourhood social deprivation and risk of prostate cancer. Br J Cancer. 2023 May 15;1–11. 9. Wu CC, Lin CH, Chiang HS, Tang MJ. A population-based study of the influence of socioeconomic status on prostate cancer diagnosis in Taiwan. Int J Equity Health. 14 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint 2018 Jun 15;17(1):79. 10. Pichardo MS, Minas TZ, Pichardo CM, Bailey-Whyte M, Tang W, Dorsey TH, et al. Association of Neighborhood Deprivation With Prostate Cancer and Immune Markers in African American and European American Men. JAMA Netw Open. 2023 Jan 20;6(1):e2251745. 11. Herrera-Luis E, Benke K, Volk H, Ladd-Acosta C, Wojcik GL. Gene–environment interactions in human health. Nat Rev Genet. 2024 May 28;1–17. 12. Hou K, Xu Z, Ding Y, Mandla R, Shi Z, Boulier K, et al. Calibrated prediction intervals for polygenic scores across diverse contexts. Nat Genet. 2024 Jul;56(7):1386–96. 13. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, et al. UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age. PLOS Med. 2015 Mar 31;12(3):e1001779. 14. Kachuri L, Graff RE, Smith-Byrne K, Meyers TJ, Rashkin SR, Ziv E, et al. Pan-cancer analysis demonstrates that integrating polygenic risk scores with modifiable risk factors improves risk prediction. Nat Commun. 2020 Nov 27;11(1):6084. 15. Rashkin SR, Graff RE, Kachuri L, Thai KK, Alexeeff SE, Blatchins MA, et al. Pan-cancer study detects genetic risk variants and shared genetic basis in two large cohorts. Nat Commun. 2020 Sep 4;11(1):4423. 16. GOV.UK [Internet]. [cited 2022 May 5]. English indices of deprivation 2019. Available from: https://www.gov.uk/government/statistics/english-indices-of-deprivation-2019 17. GOV.UK [Internet]. [cited 2024 May 21]. English indices of deprivation 2010. Available from: https://www.gov.uk/government/statistics/english-indices-of-deprivation-2010 18. Townsend P, Phillimore P, Beattie, Alastair. Health and Deprivation: Inequality and the North. 1988. 19. Conti DV, Darst BF, Moss LC, Saunders EJ, Sheng X, Chou A, et al. Trans-ancestry genome-wide association meta-analysis of prostate cancer identifies new susceptibility loci and informs genetic risk prediction. Nat Genet. 2021 15 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint Jan;53(1):65–75. 20. Newcombe PJ, Conti DV, Richardson S. JAM: A Scalable Bayesian Framework for Joint Analysis of Marginal SNP Effects. Genet Epidemiol. 2016;40(3):188–201. 21. Ruan Y, Lin YF, Feng YCA, Chen CY, Lam M, Guo Z, et al. Improving polygenic prediction in ancestrally diverse populations. Nat Genet. 2022 May;54(5):573–80. 22. Moses KA, Zhao Z, Bi Y, Acquaye J, Holmes A, Blot WJ, et al. The impact of sociodemographic factors and PSA screening among low-income Black and White men: data from the Southern Community Cohort Study. Prostate Cancer Prostatic Dis. 2017 Dec;20(4):424–9. 23. Wang J, Buto P, Ackley SF, Kobayashi LC, Graff RE, Zimmerman SC, et al. Association between cancer and dementia risk in the UK Biobank: evidence of diagnostic bias. Eur J Epidemiol. 2023 Oct 1;38(10):1069–79. 24. Shafique K, Oliphant R, Morrison DS. The impact of socio-economic circumstances on overall and grade-specific prostate cancer incidence: a population-based study. Br J Cancer. 2012 Jul;107(3):575–82. 25. Major JM, Norman Oliver M, Doubeni CA, Hollenbeck AR, Graubard BI, Sinha R. Socioeconomic status, healthcare density, and risk of prostate cancer among African American and Caucasian men in a large prospective study. Cancer Causes Control. 2012 Jul 1;23(7):1185–91. 26. Fu J, Fu C, Wang RS, Geynisman DM, Ghatalia P, Lynch SM, et al. Current Status and Future Direction to Address Disparities in Diversity, Equity, and Inclusion in Prostate Cancer Care. Curr Oncol Rep. 2023 Jul 1;25(7):699–708. 27. Patel AP, Wang M, Ruan Y, Koyama S, Clarke SL, Yang X, et al. A multi-ancestry polygenic risk score improves risk prediction for coronary artery disease. Nat Med. 2023 Jul;29(7):1793–803. 28. Shams H, Shao X, Santaniello A, Kirkish G, Harroud A, Ma Q, et al. Polygenic risk score association with multiple sclerosis susceptibility and phenotype in Europeans. Brain. 2023 Feb 1;146(2):645–56. 29. Mars N, Kerminen S, Feng YCA, Kanai M, Läll K, Thomas LF, et al. Genome-wide risk prediction of common diseases across ancestries in one million people. Cell Genomics. 2022 Apr 13;2(4):100118. 16 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint 30. Kachuri L, Chatterjee N, Hirbo J, Schaid DJ, Martin I, Kullo IJ, et al. Principles and methods for transferring polygenic risk scores across global populations. Nat Rev Genet. 2024 Jan;25(1):8–25. 31. Westerman KE, Sofer T. Many roads to a gene-environment interaction. Am J Hum Genet. 2024 Apr 4;111(4):626–35. 32. Zhu X, Yang Y, Lorincz-Comi N, Li G, Bentley AR, de Vries PS, et al. An approach to identify gene-environment interactions and reveal new biological insight in complex traits. Nat Commun. 2024 Apr 22;15(1):3385. 33. Martin AR, Gignoux CR, Walters RK, Wojcik GL, Neale BM, Gravel S, et al. Human Demographic History Impacts Genetic Risk Prediction across Diverse Populations. Am J Hum Genet. 2017 Apr 6;100(4):635–49. 34. Oh SS, Galanter J, Thakur N, Pino-Yanes M, Barcelo NE, White MJ, et al. Diversity in Clinical and Biomedical Research: A Promise Yet to Be Fulfilled. PLOS Med. 2015 Dec 15;12(12):e1001918. 35. Sangaramoorthy M, Shariff-Marco S, Conroy SM, Yang J, Inamdar PP, Wu AH, et al. Joint Associations of Race, Ethnicity, and Socioeconomic Status With Mortality in the Multiethnic Cohort Study. JAMA Netw Open. 2022 Apr 11;5(4):e226370. 17 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint Main Figure/Tables Table 1: Baseline characteristics of men included in the analysis Demographic and distribution characteristics for cases and controls for the European and African ancestries analyzed in the study. All values were obtained at time of assessment. Anthropometrics - Age, body mass index (BMI), and height (cm) - are non-standardized. Neighborhood socioeconomic status (nSES) deprivation variables include 2010 English Indices of Deprivation and Townsend Index. Index means and standard deviation (SD) are standardized for the total sample after stratifying on 18 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint population, but before case/control status. Number of clinical visits is defined as the number of primary care and hospital inpatient visits per individual in the year prior to assessment. Median and interquartile range (IQR) for number of clinical visits is defined per population before stratifying on case/control status. PSA screening status per individual is determined by indicator for self-reported PSA screening at baseline (ever/never). Total number and percentages were calculated within both population and case/control status. 19 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint Table 2: Odds ratios for prostate cancer by polygenic risk score (PRS) percentile stratified by genetic population Odds ratios (OR) and 95%confidence intervals for the association between a specified Polygenic Risk Score 269 (PRS 269) decile and Prostate Cancer (PCa) compared to individuals in the 40-60% percentile of scores (reference) for European and African ancestry samples. The number of cases and controls for each population and percentile group is given. PRS 269 values were standardized per population before analysis. 20 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint Figure 1: Odds ratios for prostate cancer in relation to nSES indices of deprivation in (A) European group individuals and (B) African group individuals Odds ratios (ORs) and 95%confidence interval estimates between nSES deprivation indices and for the (A) European, and (B) African ancestries. All indices of deprivation, including index of multiple deprivation (IMD), were standardized per group prior to modeling. ORs<1 signify increased nSES deprivation and less neighborhood resources lower odds of prostate cancer 21 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint Figure 2: Odds ratios for prostate cancer in relation to nSES indices of deprivation stratified by PSA screening status Odds ratios (ORs) and 95%confidence interval estimates between nSES deprivation indices and PCa for the European ancestry stratified by indicator for self-reported PSA screening at baseline (ever/never). All indices of deprivation, including index of multiple deprivation (IMD), were standardized prior to stratifying on screening status. P-values 22 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint calculated as a difference of means z-test for each nSES variable stratified on PSA screening group. ORs<1 signify increased nSES deprivation and less neighborhood resources lower odds of prostate cancer 23 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint Table 3: Odds ratios for prostate cancer in relation to polygenic risk score (PRS) by neighborhood deprivation quartiles. Odds ratios (OR) and 95% confidence intervals for the association between continuous PRS269 values and PCa stratified by individuals in each neighborhood socioeconomic status (nSES) deprivation quartile for men of European ancestry. PRS 269 values were standardized in the European ancestry sample prior to separating on deprivation quartile. Heterogeneity P-value calculated from a Cochran’s Q test comparing PRS 269 associations within each stratified nSES deprivation index. 24 . CC-BY-NC 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.24311312doi: medRxiv preprint