Cancer Risk Prediction Research Articles

The efficacy of machine learning models in lung cancer risk prediction with explainability.

Among many types of cancers, to date, lung cancer remains one of the deadliest cancers around the world. Many researchers, scientists, doctors, and people from other fields continuously contribute to this subject regarding early prediction and diagnosis. One of the significant problems in prediction is the black-box nature of machine learning models. Though the detection rate is comparatively satisfactory, people have yet to learn how a model came to that decision, causing trust issues among patients and healthcare workers. This work uses multiple machine learning models on a numerical dataset of lung cancer-relevant parameters and compares performance and accuracy. After comparison, each model has been explained using different methods. The main contribution of this research is to give logical explanations of why the model reached a particular decision to achieve trust. This research has also been compared with a previous study that worked with a similar dataset and took expert opinions regarding their proposed model. We also showed that our research achieved better results than their proposed model and specialist opinion using hyperparameter tuning, having an improved accuracy of almost 100% in all four models.

Mammographic Texture versus Conventional Cumulus Measure of Density in Breast Cancer Risk Prediction: A Literature Review.

Mammographic textures show promise as breast cancer risk predictors, distinct from mammographic density. Yet, there is a lack of comprehensive evidence to determine the relative strengths as risk predictor of textures and density and the reliability of texture-based measures. We searched the PubMed database for research published up to November 2023, which assessed breast cancer risk associations [odds ratios (OR)] with texture-based measures and percent mammographic density (PMD), and their discrimination [area under the receiver operating characteristics curve (AUC)], using same datasets. Of 11 publications, for textures, six found stronger associations (P < 0.05) with 11% to 508% increases on the log scale by study, and four found weaker associations (P < 0.05) with 14% to 100% decreases, compared with PMD. Risk associations remained significant when fitting textures and PMD together. Eleven of 17 publications found greater AUCs for textures than PMD (P < 0.05); increases were 0.04 to 0.25 by study. Discrimination from PMD and these textures jointly was significantly higher than from PMD alone (P < 0.05). Therefore, different textures could capture distinct breast cancer risk information, partially independent of mammographic density, suggesting their joint role in breast cancer risk prediction. Some textures could outperform mammographic density for predicting breast cancer risk. However, obtaining reliable texture-based measures necessitates addressing various issues. Collaboration of researchers from diverse fields could be beneficial for advancing this complex field.

Computational single-cell methods for predicting cancer risk.

Despite recent biotechnological breakthroughs, cancer risk prediction remains a formidable computational and experimental challenge. Addressing it is critical in order to improve prevention, early detection and survival rates. Here, I briefly summarize some key emerging theoretical and computational challenges as well as recent computational advances that promise to help realize the goals of cancer-risk prediction. The focus is on computational strategies based on single-cell data, in particular on bottom-up network modeling approaches that aim to estimate cancer stemness and dedifferentiation at single-cell resolution from a systems-biological perspective. I will describe two promising methods, a tissue and cell-lineage independent one based on the concept of diffusion network entropy, and a tissue and cell-lineage specific one that uses transcription factor regulons. Application of these tools to single-cell and single-nucleus RNA-seq data from stages prior to invasive cancer reveal that they can successfully delineate the heterogeneous inter-cellular cancer-risk landscape, identifying those cells that are more likely to turn cancerous. Bottom-up systems biological modeling of single-cell omic data is a novel computational analysis paradigm that promises to facilitate the development of preventive, early detection and cancer-risk prediction strategies.

Development and external validation of a head and neck cancer risk prediction model.

Head and neck cancer (HNC) incidence is on the rise, often diagnosed at late stage and associated with poor prognoses. Risk prediction tools have a potential role in prevention and early detection. The IARC-ARCAGE European case-control study was used as the model development dataset. A clinical HNC risk prediction model using behavioral and demographic predictors was developed via multivariable logistic regression analyses. The model was then externally validated in the UK Biobank cohort. Model performance was tested using discrimination and calibration metrics. 1926 HNC cases and 2043 controls were used for the development of the model. The development dataset model including sociodemographic, smoking, and alcohol variables had moderate discrimination, with an area under curve (AUC) value of 0.75 (95% CI, 0.74-0.77); the calibration slope (0.75) and tests were suggestive of good calibration. 384 616 UK Biobank participants (with 1177 HNC cases) were available for external validation of the model. Upon external validation, the model had an AUC of 0.62 (95% CI, 0.61-0.64). We developed and externally validated a HNC risk prediction model using the ARCAGE and UK Biobank studies, respectively. This model had moderate performance in the development population and acceptable performance in the validation dataset. Demographics and risk behaviors are strong predictors of HNC, and this model may be a helpful tool in primary dental care settings to promote prevention and determine recall intervals for dental examination. Future addition of HPV serology or genetic factors could further enhance individual risk prediction.

Artificial Intelligence-Powered Imaging Biomarker Based on Mammography for Breast Cancer Risk Prediction.

The purposes of this study were to develop an artificial intelligence (AI) model for future breast cancer risk prediction based on mammographic images, investigate the feasibility of the AI model, and compare the AI model, clinical statistical risk models, and Mirai, a state of-the art deep learning algorithm based on screening mammograms for 1-5-year breast cancer risk prediction. We trained and developed a deep learning model using a total of 36,995 serial mammographic examinations from 21,438 women (cancer-enriched mammograms, 17.5%). To determine the feasibility of the AI prediction model, mammograms and detailed clinical information were collected. C-indices and area under the receiver operating characteristic curves (AUCs) for 1-5-year outcomes were obtained. We compared the AUCs of our AI prediction model, Mirai, and clinical statistical risk models, including the Tyrer-Cuzick (TC) model and Gail model, using DeLong's test. A total of 16,894 mammograms were independently collected for external validation, of which 4002 were followed by a cancer diagnosis within 5 years. Our AI prediction model obtained a C-index of 0.76, with AUCs of 0.90, 0.84, 0.81, 0.78, and 0.81, to predict the 1-5-year risks. Our AI prediction model showed significantly higher AUCs than those of the TC model (AUC: 0.57; p < 0.001) and Gail model (AUC: 0.52; p < 0.001), and achieved similar performance to Mirai. The deep learning AI model using mammograms and AI-powered imaging biomarkers has substantial potential to advance accurate breast cancer risk prediction.

Validation of the BOADICEA model in a prospective cohort of BRCA1/2 pathogenic variant carriers

BackgroundNo validation has been conducted for the BOADICEA multifactorial breast cancer risk prediction model specifically in BRCA1/2 pathogenic variant (PV) carriers to date. Here, we evaluated the performance of BOADICEA...

A set of molecular biomarkers as an indicator of cancer risk.

e12563 Background: Biomarkers are molecular indicators of a biological status and are often used to determine the presence or absence of cancer Molecular changes in both in of cancer cells and their surrounding environment including, blood other body fluids, and tissues that react to the cancer provide biomarker materials that may be effective predictor of disease risk or disease outcome. Effective biomarkers for the disease associated risk and early detection of cancer improves potential patient outcome. An emerging theme is to develop multi-omics panels of biomarker with sufficient sensitivity and specificity for the pre-symptomatic detection of cancer. We applied approximately 300 genomics and proteomics-based genomics (PbG) biomarkers, to identify a panel of biomarkers associated with different types of cancers using patient and donor RNA samples collected from platelets Platelet RNA mirrors the RNA of nucleated cells and can be an important indicator of tissue states particularly in under pro-inflammatory conditions [PMC9555784]. Methods: Publicly available RNA-seq data of N = 2351 blood platelet samples, (including n= 390 asymptomatic controls) from 18 different tumor types was analyzed to identify the panel of biomarkers associated with highest risk cancers. Approximately 300 genomic and PbG signatures (developed from gene expression-based signatures on patients stratified according to protein expression (IHC) and survival. The predictive value of each gene signature (including commercial once ROR, OncotypeDx, etc.) was evaluated by logistic regression modeling in univariate and multivariate settings. Results: Out of 300 biomarkers, we found a set of PbG biomarkers including ratio of BRG1 expression between nuclear and cytoplasm, cytoplasmic C-terminal-binding protein 2 (CtBP2) in express in cytoplasm and ratio of nucleus and cytoplasm, Nucleus Tripartite motif-containing 28 (TRIM28), Vascular endothelial growth factor (VEGF), Mitogen-activated protein kinases (MAPK), androgen receptor (AR) express in Nuclear vs cytoplasm, Glycoprotein 78 (Gp78) in cytoplasm, E2F3, and a member of the epidermal growth factor (EGF) receptor family (ERBB2) are highly predictive (0.77 (0.76 ~ 0.79)) of cancer associated risk on a multivariate training/validation setting. The BRG1 biomarker has the highest odds ratio (11.33) within model, known for coordinating and maintaining key signaling pathways that promote oncogenesis in different cancer types including leukemia, breast, and prostate cancer. Cancer Risk model = 3.4 - 1.01 ERBB2 + 0.3 GP78_Cyto - 3.2 AR_Nucl + 0.7 CtBP2_NucVsCyto - 0.3 E2F3 - 0.99 AR_NucVsCyto + 0.87 VEGF - 1.9 MAPK + 2.43 BRG1_NucVsCyto + 1.35 TRIM28_Nucl + 1.23 CtBP2_Cyto. Conclusions: Utilizing novel genomic, and proteomic technologies, risk prediction and early detection of cancer could identify patients at a time when outcomes are superior and treatment is manageable. Further evaluation of this model is recommended.

Improving lung cancer health equity by applying deep learning to low dose CT screening of minority and disadvantaged patients.

1584 Background: In the US, disparities in lung cancer mortality exist for African American, Hispanic, and other minorities. Standard of care low dose CT screening (LDCT) detects early-stage disease and improves mortality, yet disparities are perpetuated in screening by eligibility criteria derived from cohorts underrepresenting these minorities. One such cohort is the National Lung Screening Trial (NLST) cohort which is 92% White. Consequently, guidelines for lung cancer screening may be insufficient to address the unique needs of diverse populations. We hypothesize that Artificial Intelligence prediction of future lung cancer risk from an individual’s LDCT can partially mitigate racial and ethnic disparities and improve health system practice guidelines by individualizing screening risk as compared to current general guidelines. Here, we benchmark a Resnet18 3D neural network trained on NLST LDCT images, Sybil, on the diverse patient population of the University of Illinois Health system (UIH) which is 20% White and 60% African American. Methods: A real-world cohort from UIH consisting of 1,450 CT studies was evaluated alongside 60,378 CT studies from the NLST cohort. All CT studies evaluated by the model were not used in model training. Using Youden’s J index as a probability cutoff, sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were evaluated. Receiver operating characteristic (ROC) and precision-recall (PR) curves were generated to assess model performance between cohorts. NLST data were truncated to achieve equivalent incidence of lung cancer with UIH when generated PR curves. Results: Multi-year prediction performance (ROC-AUC and PR-AUC) between cohorts are summarized (Table). For prediction of lung cancer within 1-year of LDCT in the UIH cohort, the model respectively demonstrated sensitivity, specificity, positive predictive value, and negative predictive value among White (0.80, 0.77, 0.54, 0.92), African American (0.84, 0.78, 0.39, 0.97) races and Hispanic (0.75, 0.73, 0.60, 0.84) and non-Hispanic (0.87, 0.77, 0.42, 0.97) ethnicities. Conclusions: Model performance was similar between the NLST (92% White) and a diverse, real-world cohort at UIH (20% White) though decreases in ROC-AUC performance in year 1 predictions and may be due to insufficient representation of minority populations during model training. Prospective studies involving larger and more representative patient populations should be conducted to further optimize the model and evaluate its clinical utility to improve lung cancer health equity in minority populations.[Table: see text]

Evaluating the potential of micronuclei DNA from erythrocytes for early detection or recurrence risk prediction of colorectal cancer.

Evaluating the potential of micronuclei DNA from erythrocytes for early detection or recurrence risk prediction of colorectal cancer.

Functional partial least squares with censored outcomes: Prediction of breast cancer risk with mammogram images

Functional partial least squares with censored outcomes: Prediction of breast cancer risk with mammogram images

The Associations and Causal Relationships of Ovarian Cancer - Construction of a Prediction Model.

To explore the risk and protective factors for developing ovarian cancer and construct a risk prediction model. Information related to patients diagnosed with ovarian cancer on the electronic medical record data platform of three tertiary hospitals in Guangdong Province from May 2018 to September 2023 was collected as the case group. Patients with non-ovarian cancer who attended the clinic during the same period were included in the control group. Logistic regression analysis was used to screen the independent variables and explore the factors associated with the development of ovarian cancer. An ovarian cancer risk prediction model was constructed using a decision tree C4.5 algorithm. The ROC and calibration curves were plotted, and the model was validated. Logistic regression analysis identified independent risk and protective factors for ovarian cancer. The sample size was divided into training and test sets in a ratio of 7:3 for model construction and validation. The AUC of the training and test sets of the decision tree model were 0.961 (95% CI:0.944-0.978) and 0.902 (95% CI:0.840-0.964), respectively, and the optimal cut-off values and their coordinates were 0.532 (0.091, 0.957), and 0.474 (0.159, 0.842) respectively. The accuracies of the training and test sets were 93.3% and 84.2%, respectively, and their sensitivities were 95.7% and 84.2%, respectively. The constructed ovarian cancer risk prediction model has good predictive ability, which is conducive to improving the efficiency of early warning of ovarian cancer in high-risk groups.

Association of polygenic-based breast cancer risk prediction with patient management.

10527 Background: Guidelines recommend individuals with ≥20% lifetime risk of breast cancer (BC) undergo enhanced management, including annual screening mammography (SM) as early as age 30, annual breast MRI, and genetic counseling (GC). Lifetime BC risk can be estimated using a validated risk model, such as Tyrer-Cuzick (TC). A risk predictor that combines TC with a polygenic risk score (“combined risk score” or CRS) significantly improves risk prediction over TC alone. Little is known about how clinicians manage patients based on CRS. Here, we describe management following receipt of CRS results. Methods: De-identified administrative claims data from the Optum Labs Data Warehouse were linked with de-identified TC and CRS results originally provided to ordering clinicians. Patients were divided into four subgroups based on their lifetime risk predicted by CRS and by TC alone (high risk: ≥20%, average risk: &lt;20%): A (CRS-high/TC-high), B (CRS-high/TC-avg), C (CRS-avg/TC-high), and D (CRS-avg/TC-avg). Patient management claims were compared 360 days before and after the CRS test date and differences evaluated with McNemar’s test. Differences in post-test management were analyzed using multivariable logistic regression with adjustment for clinical factors. Results: Before receiving risk results, 11.9%, 7.7%, 8.8%, and 5.0% of patients in A, B, C, and D received a SM before age 40, respectively; 3.4%, 2.6%, 2.3%, and 0.9% received an MRI, respectively; and 4.3%, 4.7%, 6.0%, and 3.6% received GC, respectively. After receiving risk results, SM in those under age 40 significantly increased 1.6-2.2-fold in A, B, and C; MRI significantly increased 4.7-5.6-fold in A, B, and C; and GC significantly increased 1.9-2.3-fold in A and B. SM, MRI, and GC did not increase in D. Prophylactic mastectomy occurred in only 0.17% of all patients after receiving results. After adjustment for confounders, those in A, B, and C were 3.8-5.2 times more likely to undergo SM; 11.6-23.1 times more likely to undergo MRI; and 2.0-2.9 times more likely to undergo GC, compared to those in Group D (all p&lt;0.001). Conclusions: Patients with a ≥20% lifetime risk for BC were more likely to undergo enhanced management compared to those with a lifetime risk &lt;20%, regardless of whether their risk was based on the CRS or on TC. These results suggest that clinicians recommended management aligned with guidelines for those with ≥20% lifetime risk, even when such risk was predicted by the CRS. [Table: see text]

Effect of novel polygenetic variant of HUNT lung cancer model on lung cancer risk assessment over clinical model among light smokers (average &lt;9 pack-years).

e20066 Background: As lung cancer incidence and mortality increase, methods of early diagnosis are needed to increase survival. The validated HUNT Lung Cancer Model (HUNT LCM) predicts individual 6-year risk of lung cancer in ever smokers with an AUC of 0.87 based on eight clinical variables: sex, age, BMI, pack-years, smoking intensity number of cigarettes per day, quit time, daily cough, and daily indoors smoke exposure. We aimed to explore whether the performance of the HUNT model can be improved by adding genetic information, and in what group of ever smokers these differences in performance emerges. Methods: Based on the eight clinical variables of the HUNT LCM and 22 single nucleotide polymorphisms (SNPs) previously identified in the literature as highly associated with lung cancer risk, the novel lung cancer risk prediction model HUNT Lung-SNP was developed. Clinical and genetical data of ever-smokers from the HUNT2 population study (n = 30746, median follow-up 15.26 years) where 160 individuals were diagnosed with lung cancer within six years were used for the model fitting. External validation was performed in another population-based cohort, the Tromsø Study (n = 2663) where 39 were diagnosed with lung cancer within 6 years. Results: The integrated discrimination improvement index (IDI) between the HUNT Lung-SNP and the HUNT LCM shows that the HUNT Lung-SNP significantly improve the LC risk stratification within six years with an IDI of 0.019 (95% CI 0.015-0.025), p &lt; 0·001 and of 0.013 (95% CI 0.008-0.018), p &lt; 0·001 in the HUNT2 and Tromsø cohorts, respectively. Of the cases, 21/160 (13%) and 10/39 (25%) in the HUNT and Tromsø cohort respectively, were low risk by the HUNT LCM but high risk by the HUNT Lung-SNP model. The main difference in characteristics among cases predicted by the HUNT Lung-SNP model, but missed by the HUNT LCM, was the average number of pack-years: 8.22 vs. 29.8 (p &lt; 0.001) in HUNT2 and 7.7 vs. 33.7 (p &lt; 0.001) in Tromsø cohort. Conclusions: A genetic approach combined with clinical variables of the HUNT LCM can improve the lung cancer risk prediction significantly by increasing the sensitivity among light smokers. Thus, we identified a high-risk population in two independent cohorts that would not be eligible for screening by clinical cutoffs nor with a clinical risk prediction model. This novel HUNT Lung-SNP needs further validation in other populations to evaluate its clinical utility.

Harnessing population-wide health data to predict cancer risk

Harnessing population-wide health data to predict cancer risk

Novel models for the functional characterization of SDHA germline variants to predict cancer risk.

Novel models for the functional characterization of SDHA germline variants to predict cancer risk.

Association between Metabolic-Dysfunction-Associated Steatotic Liver Disease and Hepatic Cancer: Current Concepts and Future Challenges.

Background: This study systematically reviewed the association between metabolic-dysfunction-associated steatotic liver disease (MASLD) and the development of hepatic cancer. Previous research has highlighted MASLD as a predisposing condition. Aim: To collect recent global data on the relationship between MASLD and hepatic cancer. Methods: A systematic review was conducted, which included an analysis of studies on the relationship between MASLD and the incidence of hepatic cancers, focusing on the role of fibrosis and MASLD severity as predictors of cancer risk. Following standard methodological frameworks for the assessment of longitudinal studies, the review gathered information on fibrosis scores, hepatocellular carcinoma (HCC) incidence, and other types of hepatic neoplasms. Results: A total of 522 studies were initially identified, of which 6 studies were appropriate for the review. They collectively revealed that the stage of fibrosis in MASLD is a significant independent predictor of mortality and liver-related events, with higher fibrosis stages correlating with greater risk. Longitudinal data showed that increases in FIB-4 scores were linked to a higher risk of developing HCC and cirrhosis. MASLD was also associated with an increased risk of non-hepatic cancers such as colorectal cancer in males and breast cancer in females. The severity of MASLD was found to be a modifiable risk factor for biliary tract cancer (BTC), with the risk further amplified by diabetes. Moreover, lifestyle factors and comorbidities, such as smoking and diabetes, were identified as modifiers of cancer risk in MASLD patients. Conclusions: The systematic review identified the association between MASLD and an elevated risk of hepatic cancer, establishing a clear link between the severity of liver fibrosis and the incidence of HCC and other hepatic neoplasms. This supports the need for screening for hepatic cancer in patients with MASLD, particularly in the presence of advanced fibrosis or other risk-modifying factors.

Interpreting disease genome-wide association studies and polygenetic risk scores given eligibility and study design considerations.

Genome-wide association studies (GWAS) have been helpful in identifying genetic variants predicting cancer risk and providing new insights into cancer biology. Increasing use of genetically informed care, as well as genetically informed prevention and treatment strategies, have also drawn attention to some of the inherent limitations of cancer genetic data. Specifically, genetic endowment is lifelong. However, those recruited into cancer studies tend to be middle-aged or older people, meaning the exposure most likely starts before recruitment, as opposed to exposure and recruitment aligning, as in a trial or a target trial. Studies in survivors can be biased as a result of depletion of the susceptibles, here specifically due to genetic vulnerability and the cancer of interest or a competing risk. In addition, including prevalent cases in a case-control study will make the genetics of survival with cancer look harmful (Neyman bias). Here, we describe ways of designing GWAS to maximize explanatory power and predictive utility, by reducing selection bias due to only recruiting survivors and reducing Neyman bias due to including prevalent cases alongside using other techniques, such as selection diagrams, age-stratification, and Mendelian randomization, to facilitate GWAS interpretability and utility.

An improved epigenetic counter to track mitotic age in normal and precancerous tissues

The cumulative number of stem cell divisions in a tissue, known as mitotic age, is thought to be a major determinant of cancer-risk. Somatic mutational and DNA methylation (DNAm) clocks are promising tools to molecularly track mitotic age, yet their relationship is underexplored and their potential for cancer risk prediction in normal tissues remains to be demonstrated. Here we build and validate an improved pan-tissue DNAm counter of total mitotic age called stemTOC. We demonstrate that stemTOC’s mitotic age proxy increases with the tumor cell-of-origin fraction in each of 15 cancer-types, in precancerous lesions, and in normal tissues exposed to major cancer risk factors. Extensive benchmarking against 6 other mitotic counters shows that stemTOC compares favorably, specially in the preinvasive and normal-tissue contexts. By cross-correlating stemTOC to two clock-like somatic mutational signatures, we confirm the mitotic-like nature of only one of these. Our data points towards DNAm as a promising molecular substrate for detecting mitotic-age increases in normal tissues and precancerous lesions, and hence for developing cancer-risk prediction strategies.

A domain knowledge-based interpretable deep learning system for improving clinical breast ultrasound diagnosis.

Though deep learning has consistently demonstrated advantages in the automatic interpretation of breast ultrasound images, its black-box nature hinders potential interactions with radiologists, posing obstacles for clinical deployment. We proposed a domain knowledge-based interpretable deep learning system for improving breast cancer risk prediction via paired multimodal ultrasound images. The deep learning system was developed on 4320 multimodal breast ultrasound images of 1440 biopsy-confirmed lesions from 1348 prospectively enrolled patients across two hospitals between August 2019 and December 2022. The lesions were allocated to 70% training cohort, 10% validation cohort, and 20% test cohort based on case recruitment date. Here, we show that the interpretable deep learning system can predict breast cancer risk as accurately as experienced radiologists, with an area under the receiver operating characteristic curve of 0.902 (95% confidence interval = 0.882 - 0.921), sensitivity of 75.2%, and specificity of 91.8% on the test cohort. With the aid of the deep learning system, particularly its inherent explainable features, junior radiologists tend to achieve better clinical outcomes, while senior radiologists experience increased confidence levels. Multimodal ultrasound images augmented with domain knowledge-based reasoning cues enable an effective human-machine collaboration at a high level of prediction performance. Such a clinically applicable deep learning system may be incorporated into future breast cancer screening and support assisted or second-read workflows.

Prediction of esophageal cancer risk based on genetic variants and environmental risk factors in Chinese population

BackgroundResults regarding whether it is essential to incorporate genetic variants into risk prediction models for esophageal cancer (EC) are inconsistent due to the different genetic backgrounds of the populations studied. We aimed to identify single-nucleotide polymorphisms (SNPs) associated with EC among the Chinese population and to evaluate the performance of genetic and non-genetic factors in a risk model for developing EC.MethodsA meta-analysis was performed to systematically identify potential SNPs, which were further verified by a case-control study. Three risk models were developed: a genetic model with weighted genetic risk score (wGRS) based on promising SNPs, a non-genetic model with environmental risk factors, and a combined model including both genetic and non-genetic factors. The discrimination ability of the models was compared using the area under the receiver operating characteristic curve (AUC) and the net reclassification index (NRI). The Akaike information criterion (AIC) and Bayesian information criterion (BIC) were used to assess the goodness-of-fit of the models.ResultsFive promising SNPs were ultimately utilized to calculate the wGRS. Individuals in the highest quartile of the wGRS had a 4.93-fold (95% confidence interval [CI]: 2.59 to 9.38) increased risk of EC compared with those in the lowest quartile. The genetic or non-genetic model identified EC patients with AUCs ranging from 0.618 to 0.650. The combined model had an AUC of 0.707 (95% CI: 0.669 to 0.743) and was the best-fitting model (AIC = 750.55, BIC = 759.34). The NRI improved when the wGRS was added to the risk model with non-genetic factors only (NRI = 0.082, P = 0.037).ConclusionsAmong the three risk models for EC, the combined model showed optimal predictive performance and can help to identify individuals at risk of EC for tailored preventive measures.