Risk Stratification for Endometrial Cancer Reveals Independent Contributions of Polygenic Risk and Body Mass Index

Background Obesity is a major risk factor for endometrial cancer, but it is unknown whether it impacts the association between genetic risk and endometrial cancer. We incorporated polygenic risk score and epidemiological risk factors in the prediction of and investigated associations of BMI and polygenic risk score with endometrial cancer risk

We generated polygenic risk score for endometrial cancer in 129,829 unrelated female participants of European ancestry (including 956 incident cases with endometrial cancer) in the UK Biobank and predicted endometrial cancer using endometrial cancer polygenic risk score and established epidemiological risk factors, including BMI. We evaluated the performance of endometrial cancer prediction models by odds ratios and area under the receiver operating characteristic curves (AUCs) to using logistic regression. Individual and joint associations of BMI and polygenic risk score with endometrial cancer were assessed using Cox proportional hazards models.

An integrated model incorporating both polygenic risk score and epidemiological risk factors achieved a modest, but statistically significant, improvement in predicting endometrial cancer status compared with the model that included epidemiologic risk factors alone (AUC = 0.74 versus 0.73; P = 3.98 × 10 -5). Obese participants (BMI ≥ . CC-BY-NC-ND 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 (which was not certified by peer review) The copyright holder for this preprint this version posted February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 4 30 kg/m2) in the top polygenic risk tertile had the highest endometrial cancer risk. We observed independent effects of genetic risk and BMI on endometrial cancer risk.

Integrating polygenic risk score with epidemiological risk factors may offer insights into population stratification for endometrial cancer susceptibility. Higher endometrial cancer polygenic risk is associated with endometrial cancer, irrespective of BMI.

Endometrial cancer, polygenic risk score, body mass index, risk stratification . CC-BY-NC-ND 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 (which was not certified by peer review) The copyright holder for this preprint this version posted February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 5

Endometrial cancer is the most common gynaecological cancer in developed countries, with 420,242 new cases and 97,704 new deaths estimated globally in 2022. 1 Notably, over the last three decades, the incidence and mortality of endometrial cancer has been increasing worldwide.2 Previous studies have used epidemiological risk factors including age, BMI, parity, duration of oral contraceptive use, age at menarche, age at menopause, to predict endometrial cancer status at moderate accuracy, with AUC values varying from 0.61 to 0.77 depending on the risk factors included and specific study populations. 3-6 Polygenic risk score (PRS) approaches, the cumulative dosage effects of multiple genetic variants identified by genome-wide association studies (GWAS), hold promise for disease stratification. 7 Incorporating polygenic risk scores into epidemiological models have achieved marginal improvement in the prediction of endometrial cancer development. 3,5,6 However, those studies have only included genome-wide significant or sub-genome-wide significant variants into the construction of the endometrial cancer PRS. Obesity, typically measured by BMI, is the strongest known modifiable risk factor for endometrial cancer.8 With the rising global prevalence of obesity, 9 the incidence of endometrial cancer is further expected to increase. In addition to obesity and genetic variation, factors such as age, ages at menopause and menarche, parity and endogenous sex hormone levels also affect endometrial cancer risk. 8,10-19 Despite the impact of all these factors, the aetiology of endometrial cancer is not fully understood and there is a significant gap in research examining the integration of risk factors for endometrial cancer prediction. Indeed, the identification of individuals . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 6 at an elevated risk of endometrial cancer, irrespective of strong risk factors such as obesity, may be crucial for early detection and facilitate the implementation of targeted screening and preventive strategies. This study uniquely evaluates the combined use of PRS and established risk factors in the prediction of endometrial cancer and explores the individual and joint effects of BMI and genetic susceptibility on endometrial cancer risk in the UK Biobank.

Study Population A flowchart outlining the study design and population is presented in Figure 1. This cohort study was based on data from the UK Biobank, which is a prospective cohort with extensive phenotypic and genotypic data for over 500,000 UK participants aged 40-70 at enrolment. Details of the UK Biobank can be found in Bycroft et al. (2018). 20 Briefly, participants were genotyped using either UK BiLEVE Axiom Array (807,411 genetic variants) or UK Biobank Axiom Array (825,927 genetic variants). Genetic variants were imputed using the 1000 Genomes phase 3, the UK10K, and the Haplotype Reference Consortium datasets as the imputation reference panels, which resulted in 93,095,623 autosomal single nucleotide polymorphisms (SNPs), short indels and large structural variants and 3,963,705 variants on the X chromosome. Selection of Endometrial Cancer Cases and Cancer-free Participants Following the exclusion of withdrawn participants, a set of 181,201 unrelated females of European ancestry was defined as the intersection of the “White British ancestry” group (UK Biobank Data Field 22006) and the “used in genetic principal components” group (UK Biobank Data Field 22020) created by Bycroft et al. . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 7 (2018),20 the latter comprising high quality samples filtered to avoid closely related samples (Figure 1A). A total of 1,867 endometrial cancer cases were identified among the 181,201 unrelated females of White British ancestry using the International Classification of Disease 10 (ICD10) subcategory C54.1 (Malignant neoplasm of corpus uteri, Endometrium) in UK Biobank Data Fields 40001 (Underlying (primary) cause of death), 40002 (Contributory (secondary) causes of death), 40006 (Type of cancer), 41202 (Diagnoses – main), 41204 (Diagnoses – secondary), and 41270 (Diagnoses). Diagnosis date was defined as the time of the first endometrial cancer record. From the unrelated females of European ancestry, 133,322 females who had an intact uterus (no hysterectomy) and no prior cancer diagnosis (except non- melanoma skin) were included as controls. Prior cancers were identified based on ICD9 (40013, 41203, and 41271), ICD10 (40001, 40002, 40006, 41202, 41204, and 41270), and self-reported cancers (20001). We restricted analyses to 982 incident endometrial cancer cases. This included the removal of 566 prevalent cases who were diagnosed before recruitment, 240 cases without known date of diagnosis, and 79 cases who were diagnosed within 12 months to mitigate issues related to delayed diagnosis or delayed linkage to cancer registry. To evaluate endometrial cancer risk prediction models ( Figure 1B & C), we additionally excluded 4,475 females (26 cases and 4,449 non-cases) with missing values of BMI, age at menarche, number of live births, or ever taken oral contraceptive (OC) pill, resulting in 956 cases and 128,873 cancer-free cohort participants. Generation of endometrial cancer PRS . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 8 GWAS summary statistics on endometrial cancer risk were sourced from the latest Endometrial Cancer Association Consortium (ECAC) GWAS analysis (12,906 cases and 108,979 controls), which included 636 endometrial cancer cases and 62,853 cancer-free female controls from the UK Biobank. 13 To avoid potential bias due to sample overlap between the GWAS dataset and the PRS validation dataset, the ECAC GWAS summary statistics were derived again by excluding UK Biobank samples, resulting in 12,270 endometrial cancer cases and 46,126 controls (Figure 1B). 21 All GWAS variants were directly genotyped or well imputed (imputation score > 0.4) and had a minor allele frequency (MAF) > 1%. Posterior effect sizes for genome- wide variants were derived using the SBayesR 22 method implemented in genome- wide complex trait Bayesian (GCTB) software. 23 The linkage disequilibrium (LD) matrix was computed based on 1.1 million HapMap 3 variants using a banded matrix with a window size of 3 cM per SNP in a random sample of 50,000 unrelated UK Biobank samples (https://cnsgenomics.com/software/gctb/#LDmatrices). For UK Biobank individual-level genotypic data, standard GWAS quality controls were conducted to select genetic variants and samples for endometrial cancer PRS generation following the guidelines outlined by Choi et al. (2020) 24 (see https://choishingwan.github.io/PRS-Tutorial/). Briefly, genetic variants with a MAF < 0.01, a missing genotype rate exceeding 1%, or departing from Hardy-Weinberg Equilibrium (P < 1 × 10 -10) were excluded, leaving 8,947,018 variants available for potential construction of the endometrial cancer risk PRS. All samples had no more than 10% missing genotypes. Per-individual PRS was calculated as the genome- wide sum of the per-variant posterior effect size multiplied by allele dosage using . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 9 PLINK (https://www.cog-genomics.org/plink/). 25 PRS values for individual females were centered to zero. Selection of established endometrial cancer risk factors Drawing on insights from epidemiological and Mendelian randomization studies of endometrial cancer,11,18,26,27 we selected and incorporated seven key risk factors into our prediction models for endometrial cancer status (case or non-case). The selected risk factors were: measured BMI (UK Biobank data field 21001), reported age at menarche (2714), age at natural menopause (3581), number of live births (2734), ever taken OC pill (2784), and levels of SHBG (30830) and testosterone (30850) during the initial assessment visit. The proportion of participants with missing information for variables was very low for BMI, age at menarche, number of live births and ever taken OC pill (missingness range 0.3% - 2.9%). The proportion of missing was greater for SHBG (14.3%) and testosterone levels (19.5%) likely due to issues associated with biospecimen processing, including sample quantity and quality failures. Age at menopause was missing for 3.4% of participants and not available for 42.6% of participants because they had not reached menopause. In order to generate risk models that would be useful pre-menopausal, we generated a polygenic score (PGS) for age at menopause using the GWAS summary statistics provided by Ruth et al. (2021) 17 and the same procedure as for the derivation for endometrial cancer PRS described above. We used age at menopause PGS in endometrial cancer risk prediction and in the analysis of associations of endometrial cancer with BMI and PRS. We similarly leveraged PGS methods for SHBG and testosterone levels. The use of PGS for these factors in risk prediction models will allow for replication in other cohorts that typically do not have these measured. PGS . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 10 for levels of SHBG and testosterone were generated using the female-stratified GWAS summary statistics provided by Ruth et al. (2020) 27 and the same procedure as for the derivation for endometrial cancer PRS. The resulting PGS values for SHBG and testosterone instead of their measured values were included as covariates to assess the performance of our prediction models and in the analysis of associations of endometrial cancer with BMI and PRS. Age at initial assessment (UK Biobank data field 21003), and the top 10 genetic principal components (PCs) were included in all prediction models as covariates. To estimate genetic PCs, a genetic relationship matrix among individuals was created using HapMap3 variants and the GCTA-GREML method. 28 The genetic relationship matrix was used to derive genetic PCs using the GCTA software (version 1.94.1).29 Evaluation of PRS model performance and established risk factors We assessed the predictive performance of the endometrial cancer PRS model in the UK Biobank using logistic regression to calculate the Nagelkerke’s R 2 and variance on the liability scale explained by PRS as described previously. 30 Participants were divided into percentiles based on their endometrial cancer PRS distribution, and their estimated odds ratios (ORs) for endometrial cancer risk were calculated using the middle two deciles (40%-60%) as the reference group. AUCs were reported for the endometrial cancer PRS and the following: 1) each of the seven established endometrial cancer risk factors; 2) the epidemiologic model comprising all seven established endometrial cancer risk factors; and 3) an integrated model combining the epidemiological model with the endometrial cancer PRS. All analyses were adjusted for age at initial visit and the top 10 genetic PCs. We used stratified bootstrap with 2000 replicates to compute their corresponding . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 11 95% confidence intervals (CIs). We calculated the net reclassification index (NRI) to quantify the improvement in the reclassification of endometrial cancer cases and non-cases by the integrated model as compared to the traditional epidemiological model (i.e. without PRS). BMI and PRS associations with endometrial cancer UK Biobank incident endometrial cancer cases and non-cases were categorised by BMI (BMI < 25kg/m², 25kg/m² ≤ BMI < 30kg/m², and BMI ≥ 30kg/m²) or endometrial cancer PRS tertiles from non-cases. We evaluated associations of BMI and endometrial cancer PRS with endometrial cancer status using Cox proportional hazard models. We tested the proportional hazards assumption for covariates included in a model fit by testing for independence between the scaled Schoenfeld residuals and time. P-values for trend were estimated using endometrial cancer PRS and BMI as continuous variables. The multivariable models were adjusted with each other for BMI groups and endometrial cancer PRS tertiles, and additionally accounted for age at menarche, number of live births (as a categorical factor), ever taken oral contraceptive pill, PGS for age at menopause, PGS for SHBG levels, PGS for testosterone levels, age at initial assessment, and the top 10 genetic PCs. We additionally performed a sensitivity analysis where multivariable models were adjusted with each other for BMI and endometrial cancer PRS as continuous variables. Follow-up time was calculated from the baseline date (date when attended assessment centre during their initial visit) to the date of endometrial cancer diagnosis or death (whichever occurred first). Cumulative incidence rate of endometrial cancer during the follow up period were generated for the different BMI and endometrial cancer PRS groups. We tested the interactions between BMI and endometrial cancer PRS by adding an interaction term in the multivariable model. . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 12 We further explored the joint associations of BMI groups and PRS tertiles with endometrial cancer using the Cox proportional model accounting for all covariates mentioned above. A nine-group comprehensive variable was thus formed, and the first PRS tertile and a normal weight (BMI < 25kg/m 2) group was considered as the

group. All analyses were performed using R software version 4.2.2 (https://www.R- project.org/). Distribution of baseline characteristics was assessed by descriptive statistics and compared between case and non-case groups using χ 2 tests for categorical variables and t-tests for continuous variables. Cox proportional hazards ratio models were conducted using the survival (version 3.5-5) and cumulative incidence rate plots were generated using the survminer (version 0.4.9) R packages. NRI calculations were performed using the PredictABEL R package (version 1.2-4). 31 AUC were assessed using the pROC (version 1.18.0) R package. 32 Statistical tests were two-sided with a P < 0.05 considered statistically significant. Data analysis was conducted from February 22 to Sept 21, 2023.

Baseline Characteristics of Study Population This analysis comprised 134,304 unrelated female participants of the European ancestry, including 982 cases of endometrial cancer and 133,322 cancer-free cohort participants (Figure 1A). Consistent with established knowledge, endometrial cancer cases, in comparison to non-cases, exhibited older age at initial assessment, earlier age at menarche, later age at menopause, lower SHBG levels, higher testosterone levels, and a lower frequency of oral contraceptive pill use (Table 1). While the mean . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 13 number of live births did not significantly differ between cases and controls, a higher proportion of cases were nulliparous compared to non-cases (22.6% vs. 18.9%). As anticipated, we observed a higher prevalence of obesity (BMI ≥ 30 kg/m 2) among participants with endometrial cancer compared to non-cases ( P < 2.2 × 10 -16; Table 1). Integration of PRS and established risk factors for the prediction of endometrial cancer The endometrial cancer PRS was strongly associated with endometrial cancer risk amongst White-British unrelated participants (P value < 2.2 × 10-16) (Figure 2A). The Nagelkerke’s R2, a pseudo-R 2 statistic measuring proportion of variance explained by endometrial cancer PRS, was 0.9% and variance on the liability-scale explained by endometrial cancer PRS was 2.2%. Compared to the middle quintile of participants (40-60%), participants in the top 10% and top 1% of the PRS distribution had a 1.98-fold ( P = 3.03 × 10 -9) and 3.06-fold ( P = 7.10 × 10 -07) increased risk of developing endometrial cancer respectively ( eTable 1). The PRS model predicted endometrial cancer status at an AUC of 0.67 (95% CI, 0.65-0.69). While this was slightly higher than for six of the established endometrial cancer risk factors (AUC range 0.65-0.66), it was lower than the prediction accuracy for BMI (AUC = 0.71; 95% CI 0.70-0.73) (Figure 2B; eTable 2). The epidemiological model that included all seven established risk factors could predict endometrial cancer status with an AUC of 0.73 (95% CI 0.71-0.74); this accuracy was improved by 1% on the addition of the endometrial cancer PRS (i.e. the integrated model, AUC = 0.74; 95% CI 0.72- 0.75; P = 3.98 × 10 -5). The continuous net reclassification improvement for endometrial cancer prediction was 0.25 (95% CI 0.19 - 0.31; P < 1 × 10 -4) for the integrated model compared with the epidemiological model. . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 14 Associations of BMI and PRS with endometrial cancer risk The proportional hazards regression model assumes that the ratio of hazards between two groups is constant over time. There was no evidence to support violation of the proportional hazard assumption in the current analysis (global test P = 0.08). Cumulative incidence curves indicated that only participants in the top tertile of the PRS distribution displayed increased risk compared to the bottom and middle tertiles; whereas higher cumulative incidence rate was associated with both the overweight (25 Kg/m 2 ≤ BMI < 30 Kg/m 2) and obese (BMI ≥ 30 Kg/m 2) groups (Figure 3A). BMI and PRS were independently associated with endometrial cancer risk (Figure 3B ). In the model mutually adjusted for BMI groups and PRS tertiles, compared with the bottom PRS tertile, the top PRS tertile displayed an increased risk (1.71-fold) for endometrial cancer (95% CI, 1.45-2.00; P = 3.7 × 10 -11). For BMI, overweight and obese groups presented a 1.56-fold (95% CI, 1.31-1.86; P = 7.2 × 10-7) and a 3.03-fold (95% CI, 2.55-3.59; P = 4.7 × 10 -37), increased risk of endometrial cancer, respectively, compared with normal BMI group (Figure 3B). When endometrial cancer PRS and BMI were treated as continuous variables, an increase of one standard deviation of endometrial cancer PRS was associated with 2.13-fold risk of endometrial cancer (95% CI, 1.79-2.54; P = 2.52 × 10 -17), while an increase of BMI by 1 kg/m 2 was associated with a 1.09-fold risk (95% CI, 1.08-1.11; P = 2.52 × 10 -80). There was no evidence of interaction between BMI and endometrial cancer PRS (P for interaction = 0.39). Upon stratification by BMI group, we observed that only those with the greatest polygenic load (i.e. the top PRS tertile) in each group had an increased risk of endometrial cancer risk. Participants in the top PRS tertile and the highest BMI group exhibited the greatest risk (HR = 4.94; 95% CI, 3.65-6.68; P = 7.67 × 10 -25; Figure 4). Even for participants with a normal . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 15 BMI, those in the top PRS tertile had a 2.01-fold increased risk (95% CI, 1.45-2.78; P = 2.50 × 10-5), compared with the bottom PRS tertile. Increased risks for the top PRS tertile persisted after additionally adjusting for continuous BMI (eFigure1).

In this cohort study of the UK Biobank, our initial focus was on assessing the predictive performance of endometrial cancer PRS and established risk factors. Firstly, we generated a risk distribution for PRS, revealing that individuals in the top 1-10% of the PRS had a risk comparable to that associated with a first-degree family history of endometrial cancer 33,34. The incorporation of PRS with the epidemiological risk factors in a risk prediction model led to a modest improvement in performance compared to the model that included the non-genetic risk factors alone. Subsequently, we evaluated associations of BMI and PRS with endometrial cancer, finding a joint association with endometrial cancer risk, with obese participants who also had the highest PRS exhibiting the greatest risk. Participants in the top PRS tertile experienced a lower risk if they were not obese. Conversely, individuals in the top PRS tertile had increased endometrial cancer risk compared to other tertiles, even when they had a normal weight. These findings highlight the potential use of endometrial cancer PRS to identify high-risk individuals in the UK Biobank, which is particularly significant given the lack of clinical guidelines for endometrial cancer screening in the general population 35. Indeed, current screening recommendations are primarily targeted at women with or at risk of Lynch syndrome who account for only ~3% of endometrial cancer cases. 36,37 . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 16 Previous studies have shown that PRS alone provides limited benefit in population screening, individual risk prediction, and population risk stratification. 38,39 However, disease risk stratification can be improved through the integration of PRS and other risk factors,40 as we have also demonstrated. One more appropriate application of PRS is to facilitate personalized cancer screening and management practices. For instance, inclusion of PRS into breast cancer risk estimation may reduce screening and enable more personalised risk management strategies for CHEK2 and ATM pathogenic variant carriers. 41 PRS has also been shown to add value in distinguishing type 1 from type 2 and monogenic diabetes in adults with a pre- existing diagnosis of diabetes and in selecting the best treatment for different types of diabetes. 42 As we observed marginal improvement by incorporating endometrial cancer PRS with established risk factors, endometrial cancer PRS may aid in stratifying individuals with Lynch syndrome pathogenic variants. To the best of our knowledge, this is the first study to quantify the association of BMI with endometrial cancer risk across different genetic risk levels defined by PRS. Our

suggest that, irrespective of PRS, endometrial cancer risk is positively associated with BMI and can be substantially mitigated by weight reduction. An analysis of the Women’s Health Initiative observational study found that women experiencing weight loss of ≥ 5% was linked to a 29% decrease in endometrial cancer risk, and intentional weight loss of ≥ 5% in obese females was associated with 56% lower endometrial cancer risk. 43 A separate meta-analysis of 13 published studies confirmed a similar association between intentional weight loss and reduced endometrial cancer risk; additionally, the analysis revealed that bariatric surgery was associated with a remarkable 59% reduction in endometrial cancer risk.44 . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 17 The findings that endometrial cancer risk associates with a higher PRS after accounting for established risk factors suggest that endometrial cancer can progress via other mechanisms. Furthermore, although obesity is often defined by BMI, it does not fully represent the distribution of body fat within the human body. In a recent study, it was reported that BMI formed a distinct cluster of fat distribution traits, such as visceral-to-subcutaneous adipose tissue ratio, waist-to-hip ratio without adjustment for BMI, and waist adjusted for BMI. The study further revealed that one unit increase in waist-to-hip ratio or visceral-to-subcutaneous adipose tissue ratio was associated with elevated hazard ratio for endometrial cancer. 45 Therefore, including other aspects of obesity such as fat distribution and other risk factors not assessed in this study such as duration of hormonal therapy usage, LDL-cholesterol levels, circulating estrogen levels, sedentary behaviours, and smoking (summarized in 10,11) may further improve our capability to predict endometrial cancer in the general population. Our study has several limitations. Firstly, as only a very small proportion of participants had repeated measures or reports for BMI, we could not assess the effects of longitudinal changes of BMI on endometrial cancer risk. Secondly, we only included some known risk factors in our models. Other risk factors not included in this study such as fat distribution and physical activities may modify the joint association of BMI and PRS with endometrial cancer risk. Thirdly, due to the lack of endometrial cancer GWAS in non-European populations, we were unable to assess the ability of risk models to distinguish endometrial cancer in non-European populations. . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 18

The findings of this study demonstrate that an endometrial cancer prediction model incorporating both epidemiological risk factors and PRS has the best performance. The improvement of prediction of endometrial cancer by PRS is anticipated to increase as additional genetic risk variants are identified by larger GWAS. Importantly, BMI and endometrial cancer PRS exert independent effects on the risk of developing endometrial cancer, revealing a substantial increase in risk for obese individuals with high PRS. While underscoring the potential protective impact of weight loss, our findings also indicate that elevated PRS poses an increased risk, even in individuals with a normal weight. These insights emphasise the complex interplay between genetic susceptibility, lifestyle factors, and obesity in endometrial cancer risk, reinforcing the need for personalized and nuanced approaches to screening and preventive interventions. List of abbreviations PRS, polygenic risk score; PGS, polygenic score; BMI, body mass index; GWAS, genome-wide association study . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 19 Figure legends Figure 1. Flowchart for the selection of study participants in the UK Biobank and directed acyclic graph of the study design. Figure 2. Prediction performance of endometrial cancer polygenic risk score (PRS) in the UK Biobank. Figure 3. Cumulative incidence curves and multivariable-adjusted effects of endometrial cancer polygenic risk score (PRS) and BMI on endometrial cancer risk. Figure 4. The joint association of genetic risk and BMI with endometrial cancer. Supplementary materials eTable 1: Endometrial cancer risk by PRS percentiles eTable 2: Comparison of different predictors to discriminate endometrial cancer cases in the UK Biobank eFigure1: The joint association of genetic risk and BMI with endometrial cancer with additional adjustment for continuous BMI . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 20 Table 1 Baseline characteristics of participants in the UK Biobank. Baseline characteristics Overall Cases Controls P-value Number of female participants 134304 982 (0.7%) 133322 (99.3%) Age at initial assessment, years (SD) 55.7 (8.0) 59.5 (6.6) 55.7 (8.0) < 2.2 × 10 -16 BMI (SD), continuous 26.9 (5.1) 30.5 (6.9) 26.9 (5.1) < 2.2 × 10 -16 BMI categoriesa BMI < 25kg/m2 55090 (41.0%) 210 (21.4%) 54880 (41.2%) < 2.2 × 10 -16 25kg/m2 <= BMI = 30kg/m2 30128 (22.4%) 427 (43.5%) 29701 (22.3%) Missing (%) 383 (0.3%) 5 (0.5%) 378 (0.3%) Age at menarche, years (SD) 13.0 (1.6) 12.7 (1.6) 13.0 (1.6) < 9.5 × 10 -10 Number of females who reported their age at menarche (%) 130449 (97.1%) 961 (97.9%) 129488 (97.1%) Missing (%) 3855 (2.9%) 21 (2.1%) 3834 (2.9%) Age at menopause, years (SD) 50.4 (4.4) 51.7 (4.3) 50.4 (4.4) < 2.2 × 10 -16 Number of post-menopausal females (%) 72459 (54.0%) 754 (76.8%) 71705 (53.8%) Number of pre-menopausal females (%) 57218 (42.6%) 184 (18.7%) 57034 (42.8%) Missing (%) 4627 (3.4%) 44 (4.5%) 4583 (3.4%) Age at menopause PGSb (SD) 0 (1.43) 0.17 (1.44) 0 (1.43) 2.1 × 10 -4 SHBG levels (nmol/L; SD) 62.5 (30.7) 50.6 (26.0) 62.6 (30.7) < 2.2 × 10 -16 Number of females that had detectable SHBG levels 115035 (85.7%) 854 (87.0%) 114181 (85.6%) Missing (%) 19269 (14.3%) 128 (13.0%) 19141 (14.4%) SHBG PGSb (SD) 0 (0.22) -0.05 (0.23) 0 (0.22) 8.2 × 10 -11 Testosterone levels (nmol/L; SD) 1.1 (0.6) 1.2 (0.6) 1.1 (0.6) 5.4 × 10 -7 Number of females that had detectable testosterone levels 108095 (80.5%) 820 (83.5%) 107275 (80.5%) . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 21 Baseline characteristics Overall Cases Controls P-value Missing (%) 26209 (19.5%) 162 (16.5%) 26047 (19.5%) Testosterone PGSb (SD) 0 (0.34) 0.05 (0.36) 0 (0.34) 2.8 × 10 -6 Number of live births (SD), count 1.8 (1.2) 1.7 (1.2) 1.8 (1.2) 0.2 Number of live births categoriesa 0 (%) 25428 (18.9%) 222 (22.6%) 25206 (18.9%) 1.1 × 10 -2 1 (%) 17798 (13.3%) 118 (12.0%) 17680 (13.3%) > 1 (%) 90999 (67.8%) 642 (65.4%) 90357 (67.8%) Missing (%) 79 (0.1%) 0 79 (0.1%) Ever taken oral contraceptive pilla Yes (%) 112428 (83.7%) 712 (72.5%) 111716 (83.8%) < 2.2 × 10 -16 No (%) 21645 (16.1%) 270 (27.5%) 21375 (16.0%) Missing (%) 231 (0.2%) 0 231 (0.2%) Abbreviations - SD: standard deviation; BMI: body mass index; PGS: polygenic score; SHBG: sex hormone binding globulin aPercentages may not add up to 100% due to rounding; bPGS (polygenic scores) of age at menopause, SHBG, and testosterone were shifted to mean zero . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 22 . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 23 Figure 1. Flowchart for the selection of study participants in the UK Biobank and directed acyclic graph of the study design. (A) Flowchart of inclusion and exclusion criteria. (B) Flowchart of the generation of endometrial cancer polygenic risk scores (PRSs) and evaluation of its prediction of endometrial cancer status. (C) The directed acyclic graph for assessment of endometrial cancer incidence in different BMI and endometrial cancer PRS groups. QCs included filtered out SNPs with MAF 10% of samples, departing from Hardy-Weinberg Equilibrium (P < × 10 -6), and low imputation quality (< 0.4). *4,475 females (26 cases and 4,449 non-cases) within missing values of BMI, age at menarche, number of live births, or ever taken OC pill were excluded. . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 24 A. B. Figure 2. Prediction performance of endometrial cancer polygenic risk score (PRS) in the UK Biobank. (A) The distribution of endometrial cancer PRS by White British unrelated female cases and controls in the UK Biobank. (B) forest plots of AUC of prediction models included endometrial cancer PRS (EC PRS; in green), each individual risk factors (in blue), all risk factors (epidemiological model; in red), and both EC PRS and risk factors (integrated model; in red). Each model was adjusted for age a t initial assessment and the top 10 genetic principal components. . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 25 A B Figure 3. Cumulative incidence curves and multivariable-adjusted effects of endometrial cancer polygenic risk score (PRS) and BMI on endometrial cancer risk. (A) Cumulative incidence curves were drawn for PRS tertiles and BMI groups in the UK Biobank. The unit of follow-up time was months, and the start point was defined as the 12 th month (1 year after recruitment). (B) Multivariable models were adjusted for either PRS or BMI group and additionally adjusted for age at menarche, number of live births, ever taken oral contraceptive pill, and PGS values of age at . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 26 menopause and levels of SHBG and testosterone, as well as age at initial assessment and the top 10 genetic principal components. . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 27 Figure 4. The joint association of genetic risk and BMI with endometrial cancer. Multivariable models were adjusted for age at initial assessment, age at menarche, number of live births, ever taken oral contraceptive pill, PRS values of age at menopause and SHBG and testosterone and the top 10 genetic principal components. The dashed vertical line indicates a Hazard ratio of 1. . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 28 Author Contributions: Drs Wang and O’Mara had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Concept and design: Wang, O’Mara. Acquisition, analysis, or interpretation of data: All authors. Drafting of the manuscript: Wang. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Wang, O’Mara. Administrative, technical, or material support: O’Mara. Supervision: O’Mara. Conflict of Interest Disclosures: None reported. Disclaimer: Where authors are identified as personnel of the International Agency for Research on Cancer/World Health Organization, the authors alone are responsible for the views expressed in this article and they do not necessarily represent the decisions, policy or views of the International Agency for Research on Cancer /World Health Organization. Funding/Support: This work was supported by a co-funded Worldwide Cancer Research and Cancer Australia project grant awarded to T.A.O’M, E.J.C. and M.J.G (grant number 22-0253). T.A.O’M. is supported by a National Health and Medical Research Council (NHMRC) of Australia Investigator Fellowship (APP1173170). E.J.C is supported by a National Institute for Health and Care Research (NIHR) Advanced Fellowship (NIHR300650) and the NIHR Manchester Biomedical . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 29 Research Centre (NIHR203308).The endometrial cancer genome-wide association analyses were supported by the NHMRC (APP552402, APP1031333, APP1109286, APP1111246 and APP1061779), the U.S. National Institutes of Health (R01- CA134958), European Research Council (EU FP7 Grant), Wellcome Trust Centre for Human Genetics (090532/Z/09Z) and Cancer Research UK. OncoArray genotyping of ECAC cases was performed with the generous assistance of the Ovarian Cancer Association Consortium (OCAC), which was funded through grants from the U.S. National Institutes of Health (CA1X01HG007491-01 (C.I. Amos), U19-CA148112 (T.A. Sellers), R01-CA149429 (C.M. Phelan) and R01-CA058598 (M.T. Goodman); Canadian Institutes of Health Research (MOP-86727 (L.E. Kelemen)) and the Ovarian Cancer Research Fund (A. Berchuck). OncoArray genotyping of the BCAC controls was funded by Genome Canada Grant GPH-129344, NIH Grant U19 CA148065, and Cancer UK Grant C1287/A16563. All studies and funders are listed in O’Mara et al (2018). Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Data Sharing Statement This research project (Project Application Number 25331) was approved by the UK Biobank in accordance with their established access procedure. UK Biobank data is available to bona fide researchers for health-related research in the public interest. Endometrial cancer GWAS summary statistics used to generate endometrial cancer polygenic risk scores can be downloaded from the GWAS Catalog (Study accession: . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 30 GCST006464). GWAS summary statistics used to generate polygenic scores for age at natural menopause can be downloaded from the GWAS Catalog (Study accession: GCST90320256). Female-stratified GWAS summary statistics used to generate polygenic scores for SHBG and testosterone can be downloaded from the GWAS Catalog (Study accession: GCST90012107 for SHBG and GCST90012112 for testosterone). Polygenic scores generate in this study will be available at the PGS Catalog upon publication. . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 31

1. Bray F, L aver sanne M , S ung H, e t a l. Gl obal canc er st a t i stic s 202 2: GLO BOCA N e st i mate s o f inc idenc e and mortal i ty worldw ide for 3 6 canc ers in 185 coun trie s . CA Canc er J Cl in . May -Jun 2024 ;74(3) :229-263 . d oi:10. 3322 /caa c.2 18 34 2. Gu B, S hang X , Yan M, e t al. Va r i a t i on s in inc idence and mo rtali t y rate s o f end ome trial ca ncer a t t he globa l , regiona l , and na tion al level s, 1 990-2019. Gy n e c o l O n c o l . May 20 21;161(2):5 73 - 580. doi:10.1 016 /j.ygy no.2021. 01.036 3. Bafli gil C , Thomps on D J, Lopha ta nano n A , et al. Dev el opment a nd eva lua tion o f p o lyge nic risk score s for p r e dicti on o f endome trial c anc er r i s k in Europe a n wome n. Ge n et M e d . Se p 2022 ;24(9) :1847-18 56. d oi:1 0.1016 /j.gi m.2022.0 5.014 4. Hüs i ng A, Do s s u s L, Ferr ari P , et a l . An ep i demiolog ica l model f o r p redi ction of e n dometri al ca ncer ri sk in Europe . Eu r J Epi dem iol . Ja n 2 016;31(1):5 1 - 6 0. d oi:10 .1007 / s 1 0654- 015-00 30-9 5. Kac huri L, G ra ff RE , Smi th- Byrne K, e t al. Pan -ca ncer a nalysi s d emon stra te s tha t in teg r a t i ng poly genic ris k score s with modi fiabl e ri sk fa cto rs im p r ov e s r i sk pr e dict ion. Nat Co m mun . Nov 2 7 2020 ;11(1) :6084. doi:10 .1 038/s414 67 -02 0-1960 0-4 6. Shi J, Kra ft P , Ro sne r B, et a l. Ri sk predic ti on model s fo r end ome tr i a l ca ncer: d e vel opme nt and vali dation in an in te rnati ona l con sor tium. J N atl Ca nc er In st . Jan 23 2023 ;doi:10.1 093 / j nci/djad0 14 7. Torka mani A, W in e inger NE, T opol EJ. T h e per son al an d cli nica l utili ty of polyge nic risk scor e s. Na t Rev Gen e t . S ep 2 018;19(9 ):5 81 -590. doi :10 .1038/ s41576 -018 -0018 -x 8. Se t i awan VW , Yang HP, Pike MC, e t al. T y pe I and I I endome tri al c anc ers: h a ve the y diffe ren t risk f ac t o rs ? J C l in O n c o l. J u l 10 20 13;31( 20):26 07-18 . doi :10 .1200 /JC O.2012.48 .2 596 9. Bluhe r M . O be si ty: g lobal epid emio logy a nd pathog en e si s . Nat Re v Endo crin ol . M a y 2019 ;15(5) :288-298 . d oi:10. 1038 / s 41 574 -019-017 6 -8 10 . Wang X, Glubb D M, O'Mar a TA. 1 0 Year s of G W AS di scove r y in e ndom etr i al c ance r : Aetiol og y, func ti on and tran sla tion. EBi o Medic ine . F eb 2 3 2022;7 7:103895 . doi:1 0.1016 /j.ebi om .2022.1 03895 11 . Rag lan O, Kal lial a I, Ma rkozann e s G, e t al . Risk fac t o rs for en domet rial canc e r: An umbrell a revie w of th e lit era tur e . In t J Ca nc er . Oct 1 201 9;145(7):171 9 -1730. do i :10.100 2 / ijc .3196 1 12 . Cheng THT, Th ompson DJ, O'Mar a T A, e t al. Five endome tri al canc er ri s k loc i iden t ifie d through g enom e-wi de as soc iatio n a naly s is. Nat G e n et. Jun 20 16;48(6 ):667 -67 4. d oi:10 .1038/ ng.35 62 13 . O'Mara TA, Glubb DM, A ma n t F, e t a l. Id enti fic atio n of nin e new s us c eptibili ty loci for endom e t ria l ca nce r. N at Co mmu n . Aug 9 2018;9(1):31 66 . doi:1 0 .1038/ s 4 1467-01 8-05 427-7 14 . Spu r dle A B, Thomps on D J, Ahme d S, et a l . Gen ome-w id e a s s o ciati on study iden tifi es a co mmon va r ia nt a s s oc ia ted w ith ri s k o f e ndome t ri al can cer . Na t Ge net . May 2011;43 (5) :451-455 . doi:1 0.1038 /ng.81 2 15 . Pai nte r JN, O'Ma ra TA, M a r q ua rt L , et a l. Gene tic Ri s k Sc or e M endelia n Randomi zation Sho ws t h a t O be sity M e a s ured as B ody Mas s Ind ex, b ut no t Wai s t : H ip Ra tio, I s Ca usal for Endom etr i a l Canc er. C an ce r E p i de m B iom ar . N ov 2016;25 ( 1 1) :1503-151 0. d oi:10 . 1158/1055- 9965 .Epi-16 -014 7 16 . Day FR, Tho mp s on DJ, Helg a so n H, et al . Geno mic a naly se s id enti fy hundre ds o f v aria nts as s o cia ted with ag e at men a r c he a n d s u ppor t a r ole f or puberty timi ng in c anc er r i s k . Na t G e n e t . 2017 ;49(6) :834-841 . d oi:10. 1038 /ng.384 1 17 . Ruth KS , Day F R, Hu ssai n J, e t al. Gen etic insight s in to biol ogica l mec ha nism s gove rnin g h u m an o v a r i an a g ei n g. Na ture . Aug 19 2021 ;596(7872)do i :10.1038 /s41586-02 1 -0 3779-7 18 . D'Ur s o S, Ar um ugam P, W e id er T, e t a l. Mende li an rando miza t io n anal y s i s of fa ct ors rela t e d to ovu latio n an d rep r o ductive functi on a nd endome t ria l ca nce r ri sk. BMC Me d . No v 1 2022 ;20(1) :419. doi:10 .11 86/s1291 6 -022 -02585 - w 19 . Coll abo rative Grou p on Epidemi ologic al S t udi e s on Endome t ri al C. Endo m etr i a l ca ncer a nd oral c ont racep tiv e s : an indi vidua l pa rtici pa nt met a- ana ly si s o f 27 276 women wit h endom etria l . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 32 ca ncer f rom 36 epidemi ologic al s tudie s. La ncet O nc ol . Se p 2015;1 6 (9) :1061-107 0 . doi:1 0.1016 /S147 0-2045 (15)00 212 -0 20 . Byc roft C , Fr eema n C, P etko va D , et a l . The U K Biobank r e s o urce with de ep phe n oty ping a nd g e no m i c da t a . Na ture . Oct 2018 ;562 (7726):2 03-209 . doi :10 .1038 /s41586 -01 8-057 9-z 21 . Kho P F, Mor tlock S, Rog er s PAW , et a l . G enetic an aly se s of g ynec ologic al d i s e a s e i denti fy gen etic r ela t i on s hip s be t w e en ut erin e fib r o id s a nd endo m etr i a l ca ncer , a nd a nov el en dome tr i al ca ncer g ene tic r i sk r eg ion a t the WNT4 1 p36.12 locus . Hum Gene t . S ep 20 21;140( 9):13 53-1365 . doi:1 0.1007 /s00439 -021 -02 312-0 22 . Ll oyd-Jon e s LR, Zeng J , Sidor enk o J, e t al . Improved polyge nic p r e d iction by Bay e s i a n multipl e reg re ssion on s u mmary st a t i s tic s . Nat Co mm u n . Nov 8 2019;10(1) :50 86. doi:1 0.1038 /s41467 -019 -12 653-0 23 . Zen g J, de Vl aming R, Wu Y, e t a l . Si gnatu re s of n ega t i ve sel ec tion in the gen etic ar chitec tu r e of huma n comp lex tr ait s. Nat G e n et . Ma y 2018;50(5):7 46 -753. do i:10 .1038 /s415 88-01 8-0101 -4 24 . Choi SW, M ak T S, O'Reil ly PF . Tu torial : a guide to pe rfo r m ing pol ygeni c risk scor e a nalys e s. Nat Proto c . S ep 202 0;15 (9):2759 -27 72. d oi:1 0.1038/ s41596 -020 - 0 353 -1 25 . Purce ll S, N ea l e B, T odd -Bro wn K, et al. P L INK: a tool s et f or who le -genom e a s s oc i ation a nd popul ati on-ba sed link age an aly se s. A m J Hum Ge ne t . Sep 2007;81 (3) :55 9-75 . doi: 1 0.1086/5197 95 26 . Wang X, Kho PF , Rama chan dran D , e t al. Mul t i -tr ait G WAS id enti fie s a novel endo metrial ca ncer ri sk lo cus tha t a ssoc iat e s with te s to s te rone l evel s. iScie nce . 2023;26 :1065 90. doi:D OI: ht t p s:/ / doi .or g / 10 .1016 /j.i sci.2 0 23.1065 90 27 . Ruth KS , Day F R, Tyrr ell J , e t al. U sing hu m an gen etic s to un der s t a nd th e di sea s e imp acts o f te st o s te rone in men an d women. Nat ure Me dicin e . 2020 ;26 (2):252 -258 . d oi:10. 1 038/s 41 591 -020 - 0751 -5 28 . Yan g J, Beny amin B, McEv oy BP, e t al. Co mmo n SNP s expla in a larg e propo rtion o f the herit abil ity f or human heigh t. Na t G e n et . Jul 2010;42(7 ):5 65-9 . doi:1 0 .1038/ng.6 0 8 29 . Yan g J, L ee SH, G odda r d M E , Vi ssc her P M. G CTA: a t ool for g enome -wi de co mpl ex t rai t ana lysi s. Am J H um G ene t . J an 7 2011;8 8(1):76 -82. d o i:10.101 6 / j.a jhg.20 10.11 .01 1 30 . Le e SH, Godd ard ME , W r a y NR, Vi ssc her PM. A bet ter co e ffic ien t of d e te r mi na tio n for gen etic pro fil e analy s i s. Ge ne t E pide m iol . Apr 2012 ;36(3 ):214 -24. doi:10 .1 002/gep i.21 614 31 . Kundu S, Aulc henko YS, v an Duijn C M, Ja nssen s A C. Pr edic tA BEL: a n R pac kag e f or the as s e ssme nt o f ri sk predic tion model s. Eur J E p i d e mio l . Ap r 2011 ;26(4 ):261 -4. doi : 10.100 7/s10654 - 011-95 67-4 32 . Robin X, Turck N, Hainard A, e t a l. pR OC : an open - s ou r c e p ac kag e fo r R and S+ to an alyze and compa r e R O C c urve s . BMC Bioi nf orma t i cs . Mar 17 20 11;12 :77. doi:10 .1 186/1 471-2105 - 1 2 -77 33 . Win AK, Re ec e JC, Ryan S. Fa mily hi s to ry a nd ri s k of end om etria l ca ncer : a s y stem atic revie w a n d m e t a - a n a l y si s. Ob st e t G yn e c ol . J an 2 0 15;125(1) :89 - 9 8 . doi:10 .1097 /A O G .000 0 00000000 0563 34 . Johna tty S E, Ta n YY, Buchan an DD, et al . F amily histo r y of ca nce r predi c t s e ndom etrial ca ncer ri sk in dep enden tly of Ly nch Syndrome: Implic a tion s fo r g ene tic coun selli ng . Gyne col Onc ol . Nov 20 17;147(2) :38 1-387 . doi :10.1016 /j. y gyno.2 017.08.0 11 35 . Cro s bi e EJ, Ki tson S J , McAlpin e JN, Muk h opadhy ay A, P owel l ME, Sin gh N. Endom e t ria l ca ncer. Lan ce t . Ap r 9 2022 ;399(10333 ): 1 412-1428 . doi:10 .1 016/S0140 -67 36(22 )0 0323 -3 36 . Rya n NA J , Gla i r e MA , Blak e D , Ca bre ra- D andy M, Eva ns D G, Cro sbie E J. T he pr opo rtion o f endom e t ria l ca nce r s a ssoc iat ed w ith Ly nch syndrome : a sy ste matic r ev iew o f t h e litera tur e and meta -an aly si s. Gene t Med . O c t 2019;21 ( 10 ) :2 167-2180 . doi:10 .10 38/s414 36-019 - 0536 - 8 37 . Rya n NA J , Mc Mahon R , Tobi S , et al . The propor tion o f e ndom etr i al tumour s a sso c iated w ith Ly nch s yn dr om e ( PETAL S): A p ro spec tive cros s- secti ona l st ud y . PL oS Med . Sep 2 02 0;17(9) :e1 003263. doi:1 0.1371 /journal .p med.1003 263 38 . Hingorani A D , Grat ton J , F inan C , et al . P e r f ormanc e o f pol ygeni c ri sk sco re s in scr eeni ng, predic tio n, and r i sk str ati fica t i o n: s e cond ary analy s i s o f da t a in the P olyge nic S cor e Ca t a log. BM J Med . 2023;2 (1) :e0 00554. d oi :10.113 6 / bmjm ed-2023-000 554 . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint 33 39 . Huntley C, To rr B, Sud A , et al . U tili ty of p olyg enic r i s k s c ore s i n UK can cer scr eeni ng : a model ling a naly sis . La ncet Onc ol . May 10 202 3;24(6):658 -66 8. doi :10 .1016 /S147 0 -2045(23 )001 56-0 40 . Le e A, Mav add at N , Wilc ox AN, e t al . B OADI CEA: a compreh e n s iv e brea s t ca ncer risk predic tio n model incorp ora ting genetic a nd nongene tic risk fac tor s. Ge n et M e d . Au g 2019 ;21(8) :1708-17 18. d oi:1 0.1038 / s414 36-018 - 0 406 -9 41 . Gao C , Poll ey EC, Har t SN, et al . Risk o f Br ea st C anc er Among Ca rrier s o f P athog e ni c Va r i an t s in Brea st Can c er Pr edi spo si t i on G ene s Va rie s by Pol ygen ic Ri sk S core. J C lin Onco l . Aug 10 2021 ;39(23) :2564 - 2 573. d oi :10.1200 /J C O.20.0199 2 42 . Lu cke t t AM, Weed o n MN, Haw kes G, Le sli e RD, O ram RA , Gr a n t SFA . Utili ty of gen etic ri s k s c o res i n t y p e 1 d i abet es . Di a b e t o l o g i a . Sep 2 023;66(9 ):15 89-1600 . doi:10 .10 07/s 00125 -023-05955 -y 43 . Lu o J, Chle bo wski RT, H e n dr y x M, e t al . I nten tiona l W eight L o ss and Endome trial Cance r R is k . J Clin Onc ol . Ap r 10 2017;35 (11) :1189-11 9 3. doi:10 .1200 /J CO.20 16.70.5 822 44 . Zha ng X, Rh oade s J, Caa n B J, e t al. In ten ti onal w eigh t l o ss, we ight c ycli ng, an d end ometria l ca ncer ri sk: a s y stema tic revie w and me t a-an al ys i s. I n t J G y neco l Ca nc er . Nov 201 9;29(9) :1361 - 1 371 . doi:1 0.1136 /ijgc -2019-00 0728 45 . Rask -And er sen M, Iv an sson E , H o glund J , E k WE, Ka r l sson T, Joh a n sson A. Adip o s it y and sex- s p e c if ic ca n c er r is k. C ance r C ell . Ju n 1 2 2023; 41(6) :1186-119 7 e4 . d oi:10.101 6 / j. c ce ll.2023 .05.010 . CC-BY-NC-ND 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 February 21, 2025. ; https://doi.org/10.1101/2025.02.19.25322538doi: medRxiv preprint