Abstract
Background
Germline BRCA1/2 testing is recommended for all ovarian cancer patients, as identifying pathogenic variants (PVs) informs treatment and enables family cascade testing. However, in resource-limited settings, high testing costs often limit feasibility. An alternative approach is to use predictive models to prioritize patients at risk, optimizing resource allocation. Existing models are largely Western-derived; in Asian populations, models exist for breast cancer—not ovarian cancer.
Methods
Using data from a multi-center study of 1,126 Asian patients with ovarian cancer (including 147 BRCA PV carriers), we developed models incorporating cancer history, clinicopathological features, and reproductive factors to estimate likelihood of carrying PVs. We assessed discrimination, calibration, and accuracy, and compared genetic testing costs to a universal testing strategy.
Results
Our final model demonstrated good calibration and strong discriminatory power, with an area under the curve of 0.80 (95% confidence interval:0.74–0.87). Factors included in the model were age at diagnosis, ethnicity, personal and family cancer history, and clinicopathological features. The model had 77% accuracy at the optimal threshold for testing, compared to 13% accuracy with universal testing, reducing genetic testing costs from ~USD4,371 to ~USD1,974 per identified carrier. Notably, while maintaining 100% sensitivity, the model reduced testing by 15%, yielding potential savings of ~USD68,182 per 800 patients each year.
Conclusion
Targeted testing using a prediction model potentially offers a more efficient and scalable alternative when universal testing is not feasible, optimizing impact in resource-limited settings.
Impact
This work supports more affordable and equitable BRCA testing pathways for Asian ovarian cancer patients.
Keywords: BRCA testing, Ovarian cancer, Asian, Predictive model, Targeted testing
Introduction
Genetic testing for germline BRCA1 and BRCA2 pathogenic variants (PVs) in patients with breast or ovarian cancer improves survival via targeted therapies such as PARP inhibitors (1). It also facilitates family cascade testing—often more cost-effective than treatment, especially in low-resource settings. Although ovarian cancer is less common than breast cancer, BRCA mutations are more prevalent in ovarian cancer, making systematic genetic testing essential to identify carriers and offer appropriate interventions (2). Up to 60% of carriers report no family history of cancer, so family history alone misses many BRCA PV carriers (2–4). Accordingly, guidelines such as the Mainstreaming Cancer Genetic Testing (MCG) (5) and the National Comprehensive Cancer Network (NCCN) (6) recommend universal testing for all ovarian cancer patients, irrespective of tumor subtype or grade (7–9). However, universal access to genetic testing remains challenging. In high-income countries, resources are available but underutilized by ethnic minorities (10) partly due to barriers such as lack of referrals, whereas the challenge in low- and middle-income countries (LMICs) is the lack of resources, including funding and manpower. Most services are funded by charitable or research funding and are largely inaccessible to the wider population (2,11). In such contexts, identifying individuals who have a higher probability of carrying a germline alteration, may be needed to balance cost and access.
BRCA carrier prediction models for breast cancer have been developed globally, including Malaysia (12–15). These models incorporate clinicopathological, demographic, and cancer history features to generate personalized risk scores, identifying individuals likely to carry BRCA PVs and prioritize them for genetic testing. We previously developed the Asian Genetic Risk Calculator (ARiCa) for breast cancer patients (15), calibrated to a BRCA2-predominant spectrum and breast-specific predictors (such as hormone receptor status and bilateral disease). Ovarian cancer, by contrast, exhibits a BRCA1-predominant spectrum and distinct epidemiology (such as histologic subtype and staging system), making ARiCa less suitable (2). Moreover, existing ovarian models are largely Western-derived—developed and validated primarily in high-risk groups—tend to underestimate carrier probabilities in Asian populations (12,13,16,17), underscoring the need for a cancer-specific model tailored to Asian women in the general population.
Hence, this study aims to develop and validate population-specific BRCA carrier prediction model for ovarian cancer patients in the general population, and to compare its performance to the current standard of universal testing. The findings can assist in guiding genetic testing decisions in low-resource settings, enabling a more efficient and equitable allocation of healthcare resources.
Materials and Methods
Study population
The study participants included women clinically diagnosed with ovarian, fallopian, or primary peritoneal cancer, recruited from two cohorts: [1] The Malaysian Ovarian Cancer Genetic (MyOvCa) study, a single-centre hospital-based case cohort where consecutive cases from a private hospital were recruited between October 2008 and February 2015 (3), and [2] The Mainstreaming Genetic Counselling for Ovarian Cancer Patients in Malaysia (MaGiC) study, a multi-center hospital-based case cohort, sourcing cases from 23 different sites, both private and public hospitals across Malaysia (2). Further details on patients and recruitment sites are provided in previous studies (2,3).
Participants donated a blood or saliva sample that was processed and stored, completed a questionnaire that included information on lifestyle and reproductive history for ovarian cancer, and provided written informed consent. Germline DNA was sequenced using an amplicon-based Hi-Plex-NGS method on an Illumina platform (California, USA), as described previously (3). All identified pathogenic and variants of uncertain significance were confirmed by Sanger sequencing. The recruitment and genetic studies were approved by the Ethics Committees of the Ministry of Health Malaysia (NMRR-16-1322-31114), University Malaya Medical Centre (UMMC 20163-2255), Hospital Universiti Sains Malaysia (USM/JEPeM/17060286), Universiti Kebangsaan Malaysia (JEP 2017 814), and Subang Jaya Medical Centre (RSDH 201612.2).
Statistical analyses
Model development
Variables that were considered for the model development were age at diagnosis for ovarian cancer, ethnicity (Chinese, Malay, Indian, or Other), laterality (unilateral or bilateral), tumor grade, cancer stage, subtype, personal cancer history (breast, uterine, cervical, or colorectal cancers), and presence of first- and second-degree family history of breast or ovarian cancer. The grade index of carcinomas was determined based on the morphology of cancer cells observed from excised tumors under a microscope, as documented in histopathological reports. Grade 1 was assigned to well-differentiated carcinomas, grade 2 to moderately differentiated carcinomas, and grade 3 to poorly differentiated carcinomas. Cancer stage was assigned based on the International Federation of Gynecology and Obstetrics (FIGO) system, which is determined by the extent of spread and metastasis. Cancer subtype analyzed in present study were serous and non-serous subtypes encompassing endometroid, clear cell, mucinous, mixed, adenocarcinoma, and rare/unclassified. Both grade and subtype were further grouped together according to criteria established in previous publications: [1] Subtype-Grade v1 - high-grade serous, low-grade serous, endometrioid, clear cell, mucinous, and other (mixed, adenocarcinoma, and rare/unclassified) (18), [2] Subtype-Grade v2 - high-grade serous, high-grade endometrioid, high-grade clear cell, and other (low-grade serous, low- grade endometrioid, low-grade clear cell, mucinous, mixed, adenocarcinoma, and rare/unclassified) (19). Additionally, hormonal use and reproductive factors considered were oral contraceptive use (ever or never), age at menarche, menopausal status (pre-menopause or post-menopause), parity status (parous or nulliparous), and history of tubal ligation (ever or never).
The study sample was split into training (70%) and validation (30%) sets using stratified random sampling by carrier status to preserve mutation prevalence (13%) in both datasets; all other variables were split randomly. Equality tests indicated broadly similar distributions of key variables and missingness between the two datasets (Supplementary Table S1). Laterality, tumor grade, and cancer stage had missing rates exceeding 10%: 27% for laterality, 33% for grade, and 17% for stage (Table 1). Missing data in both the training and validation sets were imputed using multiple imputation by chained equations, assuming missing at random (MAR). Personal history of other cancers and BRCA PV carrier status were included in the imputation model, as they have demonstrated importance as predictors of the variables in this study. We assessed the plausibility of the MAR assumption by fitting logistic regression models for the missingness indicator of each variable against observed covariates. Missingness was associated with at least one observed variable included in the imputation model, suggesting that missingness was related to observed information and supporting multiple imputation under MAR assumption. We generated 100 imputed datasets for the analysis, with each imputed dataset analyzed separately and then combined according to Rubin’s rules (20).
Table 1. Characteristics of study population.
| Variable | Total n (%) (n=1,126) | Missing* n (%) | Chinese n (%) (n=489) | Malay n (%) (n=450) | Indian n (%) (n=115) | Other n (%) (n=68) | P-value |
|---|---|---|---|---|---|---|---|
| Demographic | |||||||
| Age at diagnosis, mean (sd) | 51.84 (11.3) | 9 (0.8) | 52.94 (11.14) | 50.31 (11.2) | 54.4 (11.8) | 49.6 (11.3) | <0.001 |
| Ethnicity, n (%) | 4 (0.4) | ||||||
| Chinese | 489 (43.6) | - | - | - | - | ||
| Malay | 450 (40.1) | - | - | - | - | ||
| Indian | 115 (10.2) | - | - | - | - | ||
| Other | 68 (6.1) | - | - | - | - | ||
| Hormonal use and reproductive history | |||||||
| Oral contraceptive, n (%) | 14 (1.2) | 0.017 | |||||
| Never | 872 (78.4) | 393 (81.4) | 332 (74.4) | 97 (84.3) | 49 (73.1) | ||
| Ever | 240 (21.6) | 90 (18.6) | 114 (25.6) | 18 (15.7) | 18 (26.9) | ||
| Age at menarche, mean (sd) | 13.02 (1.52) | 18 (1.6) | 13.12 (1.6) | 12.92 (1.5) | 12.99 (1.5) | 13 (1.6) | 0.240 |
| Menopausal status, n (%) | 19 (1.7) | 0.009 | |||||
| Pre-menopause | 213 (19.2) | 88 (18.5) | 89 (19.9) | 14 (12.2) | 22 (32.4) | ||
| Post-menopause | 894 (80.8) | 388 (81.5) | 358 (80.1) | 101 (87.8) | 46 (67.6) | ||
| Parity status, n (%) | 8 (0.7) | 0.303 | |||||
| Nulliparous | 367 (32.8) | 160 (33.0) | 155 (34.5) | 29 (25.2) | 23 (33.8) | ||
| Parous | 751 (67.2) | 325 (67.0) | 294 (65.5) | 86 (74.8) | 45 (66.2) | ||
| Tubal Ligation, n (%) | 53 (4.7) | 0.277 | |||||
| Never | 963 (89.7) | 417 (88.7) | 389 (91.3) | 98 (86.7) | 59 (93.7) | ||
| Ever | 110 (10.3) | 53 (11.3) | 37 (8.7) | 15 (13.3) | 4 (6.3) | ||
| Family history | |||||||
| FFHBC, n (%) | 13 (1.2) | 0.881 | |||||
| No | 1009 (90.7) | 433 (89.8) | 409 (91.3) | 105 (91.3) | 61 (91.0) | ||
| Yes | 104 (9.3) | 49 (10.2) | 39 (8.7) | 10 (8.7) | 6 (9.0) | ||
| FFHOC, n (%) | 15 (1.3) | 0.989 | |||||
| No | 1058 (95.2) | 459 (95.4) | 425 (95.1) | 109 (94.8) | 64 (95.5) | ||
| Yes | 53 (4.8) | 22 (4.6) | 22 (4.9) | 6 (5.2) | 3 (4.5) | ||
| SFHBC, n (%) | 13 (1.2) | 0.632 | |||||
| No | 1027 (92.3) | 449 (93.2) | 408 (91.1) | 106 (92.2) | 63 (94.0) | ||
| Yes | 86 (7.7) | 33 (6.8) | 40 (8.9) | 9 (7.8) | 4 (6.0) | ||
| SFHOC, n (%) | 14 (1.2) | 0.418 | |||||
| No | 1094 (98.4) | 471 (97.9) | 444 (99.1) | 112 (97.4) | 66 (98.5) | ||
| Yes | 18 (1.6) | 10 (2.1) | 4 (0.9) | 3 (2.6) | 1 (1.5) | ||
| Personal history | |||||||
| Type of cancer, n (%) | 7 (0.6) | 0.728 | |||||
| Ovarian | 1066 (95.3) | 465 (95.1) | 425 (94.4) | 108 (93.9) | 67 (98.5) | ||
| Fallopian tube | 20 (1.8) | 7 (1.4) | 11 (2.4) | 3 (2.6) | 0 (0.0) | ||
| Peritoneal | 32 (2.9) | 14 (2.9) | 13 (2.9) | 4 (3.5) | 1 (1.5) | ||
| Other cancer, n (%) | 0 (0.0) | ||||||
| Breast cancer | 48 (66.8) | 27 (71.1) | 13 (59.1) | 6 (66.7) | 2 (50.0) | 0.211 | |
| Uterine cancer | 14 (19.4) | 5 (13.1) | 6 (27.3) | 2 (22.2) | 2 (50.0) | 0.557 | |
| Cervical cancer | 5 (6.9) | 3 (7.9) | 1 (4.5) | 1 (11.1) | 0 (0.0) | 0.662 | |
| Colorectal cancer | 5 (6.9) | 3 (7.9) | 2 (9.1) | 0 (0.0) | 0 (0.0) | 0.770 | |
| Tumor characteristics | |||||||
| Laterality, n (%) | 308 (27.4) | 0.825 | |||||
| Unilateral | 524 (64.1) | 208 (62.5) | 228 (64.6) | 50 (64.9) | 37 (68.5) | ||
| Bilateral | 294 (35.9) | 125 (37.5) | 125 (35.4) | 27 (35.1) | 17 (31.5) | ||
| Grade, n (%) | 372 (33.0) | 0.064 | |||||
| Grade 1 | 78 (10.3) | 19 (6.7) | 47 (13.6) | 4 (6.2) | 8 (13.8) | ||
| Grade 2 | 4 (0.6) | 2 (0.7) | 1 (0.3) | 1 (1.5) | 0 (0.0) | ||
| Grade 3 | 672 (89.1) | 263 (92.6) | 298 (86.1) | 60 (92.3) | 50 (86.2) | ||
| Stage, n (%) | 186 (16.5) | 0.004 | |||||
| Stage 1 | 274 (29.2) | 106 (27.3) | 125 (31.4) | 28 (28.0) | 15 (27.8) | ||
| Stage 2 | 128 (13.6) | 62 (16.0) | 40 (10.1) | 11 (11.0) | 14 (25.9) | ||
| Stage 3 | 438 (46.6) | 187 (48.2) | 177 (44.6) | 55 (55.0) | 19 (35.2) | ||
| Stage 4 | 100 (10.6) | 33 (8.5) | 55 (13.9) | 6 (6.0) | 6 (11.1) | ||
| Subtype, n (%) | 33 (2.9) | 0.193 | |||||
| Serous | 565 (51.6) | 241 (51.3) | 218 (49.3) | 67 (59.3) | 39 (58.2) | ||
| Endometrioid | 225 (20.6) | 85 (18.1) | 99 (22.4) | 28 (24.8) | 13 (19.4) | ||
| Clear cell | 184 (16.8) | 90 (19.1) | 74 (16.7) | 9 (8.0) | 10 (14.9) | ||
| Mucinous | 40 (3.7) | 17 (3.6) | 19 (4.3) | 4 (3.5) | 0 (0.0) | ||
| Mixed | 25 (2.3) | 10 (2.1) | 14 (3.2) | 0 (0.0) | 1 (1.5) | ||
| Adenocarcinoma | 14 (1.3) | 5 (1.1) | 6 (1.4) | 2 (1.8) | 1 (1.5) | ||
| Rare/Unclassified | 40 (3.7) | 22 (4.7) | 12 (2.7) | 3 (2.6) | 3 (4.5) | ||
| Outcome | |||||||
| BRCA PVs carrier status, n (%) | 0 (0.0) | 0.043 | |||||
| Non-carrier | 979 (86.9) | 443 (90.6) | 375 (83.3) | 96 (83.5) | 61 (85.9) | ||
| BRCA1 | 97 (8.6) | 32 (6.5) | 49 (10.9) | 12 (10.4) | 7 (9.9) | ||
| BRCA2 | 50 (4.5) | 14 (2.9) | 26 (5.8) | 7 (6.1) | 3 (4.2) | ||
Sample:1,126 ovarian cancer patients from the Malaysian Ovarian Cancer Genetic (OVC) study and the Mainstreaming Genetic Counselling for Ovarian Cancer Patients in Malaysia (MaGiC) study before imputation.
Abbreviations: FFHBC, First Degree Family History for Breast Cancer; FFHOC, First Degree Family History for Ovarian Cancer; SFHBC, Second Degree Family History for Breast Cancer; SFHOC, Second Degree Family History for Ovarian Cancer; PV, pathogenic variant.
BRCA carrier prediction models were developed using logistic regression on the training set. Candidate predictors entered the multivariable model if they met a liberal univariable threshold (p-value<0.20) to avoid excluding potentially informative variables; retention in the final model used a stricter criterion (p-value<0.05). A correlation test (r) was performed to assess multicollinearity among variables. For highly correlated variables, the one with the higher odds ratio (OR) was prioritized for inclusion. Therefore, menopausal status, parity status, and presence of family cancer history were retained, while age at menopause, number of children, age at diagnosis and number of affected relatives were excluded due to their strong correlation with the retained variables (r=0.696, r=0.835, and r>0.999, respectively).
In principle, variables with p-value≥0.05 were excluded from the final models; however, we retained key predictors such as age at diagnosis, personal and family cancer history, and tumor characteristics irrespective of their p-values in the gene-specific and overall BRCA models, given their well-established importance in the literature (3,4,21). Oral contraceptive use and menopausal status were also retained because they improved discrimination between carriers and non-carriers. Hormonal use has been reported to be less common among carriers diagnosed with ovarian cancer, while menopausal status captures gene-specific differences in age at diagnosis, particularly the later onset observed in BRCA2 PV carriers (22). Recognizing that certain variables may not be relevant for specific mutation types, we explored eight different model combinations of candidate predictors in this study: [1] a baseline model encompassing demographics, personal, and family cancer history components (Model 1), [2] a reproductive model that adds oral contraceptive use, menopausal status, and parity to the baseline (Model 2), [3] three subtype models that add tumor grade and subtype to the baseline, specified as Model 3 (grade and subtype entered separately), Model 4 (Subtype–Grade v1), and Model 5 (Subtype–Grade v2); and [4] three full models combining the reproductive model with tumor grade and subtype, specified as Model 6 (grade and subtype entered separately), Model 7 (Subtype–Grade v1), and Model 8 (Subtype–Grade v2). Accordingly, we developed an overall BRCA model as well as BRCA-specific models tailored specifically for BRCA1 and BRCA2. In total, we have developed 24 models: 8 models for overall BRCA prediction and 16 BRCA-specific models (8 for BRCA1 and 8 for BRCA2).
Validation of performance of risk assessment methods
Model discrimination and calibration were assessed in the validation set. Discrimination was evaluated using the area under the receiver operating curve (AUC) (23). Calibration was assessed using the Hosmer–Lemeshow (HL) test (HL, 15 or less) and calibration plots to visualize the distribution of observed versus expected numbers of BRCA PV carriers across deciles of predicted carrier probabilities (24). The optimal carrier probability threshold for genetic testing was chosen based on the intersection of sensitivity and specificity curves (25). Additional performance measures included sensitivity (percentage of true carriers detected), specificity (percentage of true non-carriers detected), accuracy (percentage of true carriers and non-carriers detected), and screening rate (percentage of eligible patients for BRCA testing).
We conducted a simple cost analysis using a cost-threshold scenario (base-case ~USD570 per test) and sensitivity analyses at reduced unit test cost (~USD115 per test) and varying subsidy levels. Efficiency was assessed using detection and cost ratios. The detection ratio is the number of tests needed to find one BRCA PV carrier; the cost ratio is the testing cost per carrier detected. We estimated total testing costs and cost per carrier using the annual incidence (~800 cases) and mutation prevalence (13%), assuming 100% compliance in low-risk groups. All data were analyzed using R version 4.0.3 (RRID:SCR_001905), and a p-value<0.05 was considered statistically significant unless otherwise stated.
Results
In this cross-sectional multi-center case study of 1,126 patients with ovarian cancer, 147 (13.0%) had a germline BRCA PV, with 9% BRCA1 and 4% BRCA2 PV carriers (Table 1). The majority were Chinese (44%) and Malay (40%), with a mean age at diagnosis of 51.8 years (standard deviation=11.3). Indian, Chinese, and Malay women had higher proportions of advanced-stage (classified as stage 3 and 4) disease (61.0%, 58.5%, and 56.7%, respectively), compared to Other ethnic group (46.3%; p-value=0.004). Additionally, Indian and Malay women had higher proportions of carriers compared to Chinese women (16.5% and 16.7% versus 9.4%, respectively; p-value=0.043). Compared to non-carriers, both BRCA1 and BRCA2 PV carriers are more likely to have serous epithelial ovarian cancers, present at later stages with higher grades (classified as grade 2 and 3), and report a personal history of breast cancer (p-value<0.05; Supplementary Table S2). A higher proportion of BRCA PV carriers reported a first- or second-degree relative with breast and/or ovarian cancer (p-value<0.05), but more than half had no family history. While these trends hold true for both BRCA1 and BRCA2 PV carriers, BRCA2 PV carriers were more likely to be post-menopausal or parous (p-value<0.05). Conversely, BRCA1 PV carriers were more likely to have a family history of ovarian cancer compared to BRCA2 PV carriers (p-value<0.05). Due to these differences, we decided to explore separate models for BRCA1, BRCA2, and overall BRCA in subsequent analyses.
Prediction models were developed using 788 ovarian cancer cases (103 BRCA PV carriers) and validated using 338 cases (44 BRCA PV carriers; Supplementary Fig. S1). In univariable analysis, family history of breast cancer, type of cancer, personal history of breast cancer, tumor grade, cancer stage, and subtype were associated with overall BRCA, BRCA1, and BRCA2 PV carrier status (p-value<0.2; Supplementary Table S3). Age at diagnosis, ethnicity, laterality, personal history of colorectal cancer, menopausal status, and parity were associated only with BRCA2 PV carrier status, whereas family history of ovarian cancer was associated exclusively with BRCA1 PV carrier status (p-value<0.2).
The best-performing model was selected based on the highest AUC and the lowest non-significant HL score in the validation set (Supplementary Table S4). Among the eight models, subtype models (Model 3, 4, and 5) were the best-performing models for overall BRCA and BRCA1, whereas reproductive model (Model 2) for BRCA2. Even though the full models (Model 6, 7, and 8) have comparatively high AUCs, addition of hormonal use and reproductive components to the subtype models did not improve AUCs of the model performance for overall BRCA, BRCA1, and BRCA2. Subtype models with grade and subtype as individual variables (Model 3) performed better in terms of calibration compared to those incorporating the combined versions of grade and subtype components (Model 4 and 5). For BRCA2, addition of tumor characteristics (Model 3-8) reduces the discriminatory power, addition of hormonal use and reproductive components to the baseline model (Model 2), on the other hand improves discrimination while maintaining good calibration. Therefore, Model 3 was the best-performing model for overall BRCA (AUC=0.80, HL=13.8) and BRCA1 (AUC=0.80, HL=8.7), whereas Model 2 performed best for BRCA2 (AUC=0.72, HL=15.0; Figure 1). Overall, the AUCs were comparable across models.
Figure 1. Receiver operating curve and calibration plot of best-performing models by BRCA PV carrier status.
Sample: 338 ovarian cancer patients from the Malaysian Ovarian Cancer Genetic (MyOvCa) study and the Mainstreaming Genetic Counselling for Ovarian Cancer Patients in Malaysia (MaGiC) study in imputed validation set.
Abbreviations: AUC, Area Under Curve; 95% CI, 95% Confidence Interval; HL, Hosmer-Lemeshow; ROC, Receiver Operating Curve.
Note: Variables included in BRCA-specific models (Model 2 and 3): Age of diagnosis, ethnicity, oral contraceptive use, menopausal status, parity status, family history of breast or ovarian cancer (first and second degree), personal history of cancer (ovarian, breast, or colorectal cancer), tumor grade, cancer stage and subtype. Variables included in Overall BRCA model (Model 3): Age of diagnosis, ethnicity, family history of breast or ovarian cancer (first and second degree), personal history of cancer (ovarian, breast, or colorectal cancer), laterality, tumor grade, cancer stage and subtype.
Table 2 presents the final set of variables included in each model and summarizes the corresponding adjusted estimates. Malay ethnicity, personal history of breast cancer, advanced cancer stage, presence of first- and second-degree family history of breast cancer or ovarian cancer, and clear cell subtype were associated with overall BRCA PV carrier status (p-value<0.05). These variables were also associated with BRCA1 PV carrier status, except for second-degree family history of breast cancer, whereas BRCA2 was only associated with Malay ethnicity, personal history of breast cancer, first-degree family history of breast cancer, and parity status (p-value<0.05).
Table 2. Multivariable regression of best-performing models by BRCA PV carrier status.
| Model | BRCA-specific models | Overall BRCA model | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Model 3 | Model 2 | Model 3 | ||||||||||
| Variable | BRCA1 versus Non-carrier (n=753) | BRCA2 versus Non-carrier (n=720) | BRCA versus Non-carrier (n=788) | |||||||||
| OR | 95% CI | P-value | OR | 95% CI | P-value | OR | 95% CI | P-value | ||||
| Demographic | ||||||||||||
| Age at diagnosis | 0.97 | 0.94 | 1.00 | 0.089 | 0.99 | 0.96 | 1.04 | 0.917 | 0.99 | 0.96 | 1.01 | 0.261 |
| Ethnicity | ||||||||||||
| Chinese | Reference | Reference | Reference | |||||||||
| Malay | 2.52 | 1.23 | 5.17 | 0.012 | 4.20 | 1.64 | 10.77 | 0.003 | 3.09 | 1.70 | 5.61 | <0.001 |
| Indian | 1.61 | 0.56 | 4.63 | 0.375 | 2.85 | 0.80 | 10.20 | 0.108 | 2.06 | 0.87 | 4.86 | 0.100 |
| Other | 0.49 | 0.10 | 2.54 | 0.397 | 1.26 | 0.14 | 11.17 | 0.836 | 0.62 | 0.16 | 2.36 | 0.482 |
| Hormonal use and reproductive history | ||||||||||||
| Oral contraceptive | ||||||||||||
| Never | Reference | |||||||||||
| Ever | - | - | - | - | 1.06 | 0.47 | 2.43 | 0.881 | - | - | - | - |
| Menopausal status | ||||||||||||
| Pre-menopause | Reference | |||||||||||
| Post-menopause | - | - | - | - | 8.17 | 0.86 | 77.64 | 0.068 | - | - | - | - |
| Parity status | ||||||||||||
| Nulliparous | Reference | |||||||||||
| Parous | - | - | - | - | 5.27 | 1.49 | 18.63 | 0.010 | - | - | - | - |
| Family history | ||||||||||||
| Family history of cancer | Reference (No) | Reference (No) | Reference (No) | |||||||||
| FFHBC (yes) | 6.46 | 2.94 | 14.21 | <0.001 | 5.52 | 2.29 | 13.33 | <0.001 | 5.84 | 3.05 | 11.16 | <0.001 |
| FFHOC (yes) | 7.64 | 3.06 | 19.05 | <0.001 | 2.39 | 0.56 | 10.16 | 0.239 | 5.17 | 2.26 | 11.80 | <0.001 |
| SFHBC (yes) | 1.94 | 0.71 | 5.28 | 0.197 | 1.95 | 0.62 | 6.18 | 0.254 | 2.35 | 1.03 | 5.37 | 0.042 |
| SFHOC (yes) | 17.75 | 3.28 | 95.98 | 0.001 | - | - | - | - | 7.88 | 1.69 | 36.80 | 0.009 |
| Personal history | ||||||||||||
| Type of cancer | ||||||||||||
| Ovarian | Reference | Reference | ||||||||||
| Fallopian tube | 2.74 | 0.59 | 12.68 | 0.196 | - | - | - | - | 1.68 | 0.47 | 6.00 | 0.427 |
| Peritoneal | 0.15 | 0.02 | 1.42 | 0.098 | - | - | - | - | 0.19 | 0.04 | 0.90 | 0.036 |
| Other cancer | Reference (No) | Reference (No) | Reference (No) | |||||||||
| Breast cancer (yes) | 7.13 | 2.31 | 21.98 | 0.001 | 4.78 | 1.31 | 17.42 | 0.018 | 4.96 | 1.94 | 12.70 | <0.001 |
| Colorectal cancer (yes) | - | - | - | - | 5.94 | 0.40 | 88.76 | 0.196 | 2.86 | 0.17 | 49.05 | 0.469 |
| Tumor characteristics | ||||||||||||
| Laterality | ||||||||||||
| Unilateral | - | - | - | - | - | - | - | - | Reference | |||
| Bilateral | - | - | - | - | - | - | - | - | 1.72 | 0.91 | 3.26 | 0.093 |
| Grade | ||||||||||||
| Grade 1-2 | Reference | - | - | - | - | Reference | ||||||
| Grade 3 | 4.90 | 0.68 | 35.13 | 0.113 | - | - | - | - | 3.33 | 0.84 | 13.17 | 0.087 |
| Stage | ||||||||||||
| Stage 1 | Reference | Reference | ||||||||||
| Stage 2 | 6.31 | 1.34 | 29.65 | 0.020 | - | - | - | - | 2.92 | 0.85 | 10.06 | 0.090 |
| Stage 3 | 4.27 | 0.36 | 10.61 | 0.041 | - | - | - | - | 4.78 | 1.65 | 13.89 | 0.004 |
| Stage 4 | 9.43 | 0.69 | 14.68 | 0.004 | - | - | - | - | 5.77 | 1.77 | 18.77 | 0.004 |
| Subtype | ||||||||||||
| Other a | Reference | Reference | ||||||||||
| Mucinous | 2.56 | 0.12 | 53.88 | 0.545 | - | - | - | - | 0.81 | 0.06 | 10.66 | 0.871 |
| Clear cell | - | - | - | - | - | - | - | - | 0.14 | 0.02 | 0.90 | 0.038 |
| Endometrioid | 1.94 | 0.40 | 11.83 | 0.443 | - | - | - | - | 1.12 | 0.33 | 3.83 | 0.861 |
| Serous | 3.19 | 0.74 | 15.07 | 0.136 | - | - | - | - | 2.06 | 0.70 | 6.08 | 0.188 |
Sample: 788 ovarian cancer patients from the Malaysian Ovarian Cancer Genetic (MyOvCa) study and the Mainstreaming Genetic Counselling for Ovarian Cancer Patients in Malaysia (MaGiC) study in imputed training set.
Abbreviations: OR, Odds ratio; 95% CI, 95% Confidence Interval; FFHBC, First Degree Family History for Breast Cancer; FFHOC, First Degree Family History for Ovarian Cancer; SFHBC, Second Degree Family History for Breast Cancer; SFHOC, Second Degree Family History of Ovarian Cancer.
Includes mixed, adenocarcinoma, rare, and unclassified.
Note: All variables listed in this table were included in the corresponding final models. Variables included in BRCA-specific models (Model 2 and 3): Age of diagnosis, ethnicity, oral contraceptive use, menopausal status, parity status, family history of breast or ovarian cancer (first and second degree), personal history of cancer (ovarian, breast, or colorectal cancer), tumor grade, cancer stage and subtype. Variables included in Overall BRCA model (Model 3): Age of diagnosis, ethnicity, family history of breast or ovarian cancer (first and second degree), personal history of cancer (ovarian, breast, or colorectal cancer), laterality, tumor grade, cancer stage and subtype.
We identified an optimal testing threshold for each model, with the overall BRCA model requiring a higher threshold (Supplementary Fig. S2). At these thresholds, we estimated the proportion of ovarian cancer patients who would be offered BRCA testing and compared performance (Table 3). Overall BRCA model would screen 20% fewer patients (33% v 53%) to detect 5% more carriers compared to BRCA-specific models (73% v 68%). We then compared the overall BRCA model with universal testing. Table 4 showed that, at the optimal threshold, the model achieved higher accuracy while screening 67% fewer patients and costing three times less than universal testing, successfully capturing over 70% of carriers. Notably, achieving 100% sensitivity with the model would require screening 85% of ovarian cancer patients, which is still more efficient than the current guideline recommending testing of all affected individuals.
Table 3. Performance of best performing models at optimal thresholds.
| Model | BRCA-specific models | Overall BRCA model | |
|---|---|---|---|
| BRCA1 (n=323) | BRCA2 (n=309) | BRCA (n=338) | |
| Threshold, % | 6.6 | 4 | 12.2 |
| Sensitivity, % | 69 (66-73) | 67 (61-73) | 73 (71-75) |
| Specificity, % | 68 (67-68) | 67 (66-67) | 73 (72-73) |
| Eligible for genetic testing, % | 35 (30-40) | 35 (30-40) | 33 (28-38) |
| Detection ratio | 6:1 | 3:1 | |
| Total number of patients screened, n (%) | 178 (53) | 111 (33) | |
| Total number of carriers identified, n (%) | 30 (68) | 32 (73) | |
| Total number of carriers missed, n (%) | 14 (32) | 12 (27) | |
Sample: 338 ovarian cancer patients from the Malaysian Ovarian Cancer Genetic (OVC) study and the Mainstreaming Genetic Counselling for Ovarian Cancer Patients in Malaysia (MaGiC) study in imputed validation set.
Note: Variables included in BRCA-specific models (Model 2 and 3): Age of diagnosis, ethnicity, oral contraceptive use, menopausal status, parity status, family history of breast or ovarian cancer (first and second degree), personal history of cancer (ovarian, breast, or colorectal cancer), tumor grade, cancer stage and subtype. Variables included in Overall BRCA model (Model 3): Age of diagnosis, ethnicity, family history of breast or ovarian cancer (first and second degree), personal history of cancer (ovarian, breast, or colorectal cancer), laterality, tumor grade, cancer stage and subtype.
Table 4. Comparison of performance across different BRCA carrier status assessment methods.
| Risk assessment method | Overall BRCA model | Universal testing | |
|---|---|---|---|
| Threshold | 12.2a | 1b | - |
| Sensitivity (%) | 73 | 100 | 100 |
| Accuracy (%) | 77 | 26 | 13 |
| Eligible (%) | 33 | 85 | 100 |
| Detection ratio | 3:1 | 6:1 | 8:1 |
| Total cost of genetic testing (USD) | 150,000 | 386,364 | 454,545 |
| Genetic testing cost ratio (USD per detected carrier) | 1,974:1 | 3,715:1 | 4,371:1 |
Sample: 338 ovarian cancer patients from the Malaysian Ovarian Cancer Genetic (MyOvCa) study and the Mainstreaming Genetic Counselling for Ovarian Cancer Patients in Malaysia (MaGiC) study in imputed validation set.
Optimal threshold.
Threshold at 100% consistent with the standard clinical practice.
Note: Total cost of genetic testing (~USD570 per test) and corresponding cost per carrier detected were calculated based on country-specific annual incidence (800 cases) and mutation prevalence (13%), assuming 100% uptake and compliance in high-risk and low-risk groups, respectively. Variables included in Overall BRCA model (Model 3): Age of diagnosis, ethnicity, family history of breast or ovarian cancer (first and second degree), personal history of cancer (ovarian, breast, or colorectal cancer), laterality, tumor grade, cancer stage and subtype.
Additionally, we presented a cost-threshold scenario comparing the total cost of genetic testing at different sensitivity levels, assuming an ideal scenario of 100% uptake among high-risk individuals and 100% compliance among low-risk individuals. Figure 2 illustrates that the model was more cost-effective, screening 15% fewer patients and saving up to ~USD68,182 (~MYR 300,000) while achieving 100% sensitivity—comparable to universal testing. When the unit cost was reduced from ~USD570 to ~USD115 (Supplementary Fig. S3), our conclusions were unchanged: the estimated 15% reduction in testing volume persisted, while affordability improved despite smaller absolute savings. To examine cost variation, we plotted total costs across different subsidy levels and corresponding screening rates, adjusted for observed uptake among ovarian cancer patients from a recent study (26). Total costs remained consistently lower for the overall BRCA model at every subsidy level (Supplementary Fig. S4).
Figure 2. Screening rate and the corresponding total cost of genetic testing based on annual ovarian cancer incidence at varying sensitivity under base-case unit test cost.
Sample: 338 ovarian cancer patients from the Malaysian Ovarian Cancer Genetic (MyOvCa) study and the Mainstreaming Genetic Counselling for Ovarian Cancer Patients in Malaysia (MaGiC) study in imputed validation set.
Note: Total cost of genetic testing (~USD570 per test) was calculated based on country-specific annual incidence (800 cases), assuming 100% uptake and compliance in high-risk and low-risk groups, respectively.
Discussion
Summary of main results
Despite the importance of germline BRCA1 or BRCA2 PVs testing for risk management and treatment selection, genetic testing uptake remains low in Asia due to resource limitations. An efficient mutation prediction model is therefore crucial for improving detection rates, identifying individuals who would benefit most from testing, and optimizing resource allocation. We demonstrated that our proposed model for ovarian cancer patients achieves 100% sensitivity—matching universal testing—while reducing the number of patients screened by 15%, offering a more resource-efficient approach to capture all carriers.
Results in the context of published literature
The clinical features and risk factors seen in women with BRCA mutations in this study are similar to those reported in other Asian studies and partly overlap with patterns seen in breast cancer (3,15,21,27–30). While tumor grade is a key predictor for BRCA mutation in breast cancer (15), ovarian cancer’s aggressive nature and the predominance of high-grade tumors (90% classified as grade 3 in present study) limit the discriminatory power of grade (31). Breast cancer typically progresses more slowly and has a higher 5-year survival rate, with tumor grades varying widely across early and late stages (32). In contrast, ovarian cancer is inherently aggressive, often presenting as high-grade from the outset and associated with lower survival rates (32). As a result, cancer stage—which better reflects disease progression and prognosis—emerged as a more informative predictor for BRCA mutation in ovarian cancer, underscoring the importance of a cancer-specific model that accounts for key factors influencing BRCA mutation risk.
Comparison with current practice in Western setting
Current BRCA testing approaches in the West typically follow either universal testing (based on NCCN guideline) for epithelial ovarian cancer or criteria-driven referral based on mutation prediction models (e.g., BRCAPRO and CanRisk) (33–36). However, these models rely heavily on detailed pedigree information, including age at diagnosis across multiple relatives, to estimate carrier probability and apply a fixed referral threshold (≥5% based on NCCN guideline) (36). In contrast, our models differ in both structure and intended use. Rather than relying primarily on extended family history, it incorporates routinely available clinical and tumor characteristics, allowing implementation in settings where pedigree information is often incomplete or unavailable. Notably, in our cohort, the threshold that maximized sensitivity and specificity was 12.2%, which is higher than the recommended 5% referral threshold in Western settings; applying a 5% threshold in our cohort would result in substantially broader testing with limited additional gain in sensitivity, highlighting the need for context-specific calibration of referral thresholds.
Implications for practice and future research
Despite international recommendations to test all ovarian cancer patients, implementation gaps persist in many Asian settings: in Singapore, only 18.1% of eligible ovarian cancer patients were referred to a cancer genetics clinic, and in Malaysia only ~17% proceeded with testing when self-funded (26,37). These gaps likely reflect persistent barriers in routine care, including high testing costs and limited clinical genetics services (2,7,10,11,37). Consistent with this, a local study showed that genetic testing uptake among ovarian cancer patients is influenced by the subsidy levels, with higher uptake observed when testing is fully subsidized (26). In LMICs, constraints extend beyond laboratory price alone to include counselling resources, referral logistics, and overall service delivery capacity. A strategy that preserves sensitivity while reducing testing volume may therefore improve feasibility and scalability. Previously, we showed that targeted BRCA testing using a BRCA carrier prediction model spares low-risk individuals from unnecessary testing—an approach relevant to low-resource settings (15). Building on our prior work, our current study demonstrated that the proposed BRCA carrier prediction model for ovarian cancer supports a scalable referral-threshold strategy that can be adapted to available resources. In our data, a 1% threshold achieved 100% sensitivity while recommending testing for 85% of patients, whereas a 12.2% threshold recommended testing for 33% with 73% sensitivity, illustrating how programs can choose a cutoff that balances missed carriers against feasible testing volume and cost. In practice, as capacity and budgets increase, the threshold can be lowered to expand testing; when resources are constrained, it can be raised to focus testing on those at highest risk.
Although BRCA testing is often associated with targeted therapies and improved survival (38), its greatest impact in low-resource settings is enabling risk management and cancer prevention. By identifying BRCA PV carriers, we can implement preventive measures not only for affected individuals but also for their at-risk relatives. A systematic review found that BRCA cascade screening—involving the testing of both affected individuals and their unaffected relatives—is a cost-effective approach (39). A study conducted in a middle-income Asian country further demonstrated that familial cascade testing can be cost-saving (40). Our proposed mutation prediction model supports this preventive approach by prioritizing cancer patients most likely to carry BRCA PVs, thereby facilitating more targeted and efficient cascade testing in resource-constrained environments.
Study limitations and strengths
This study is limited by a modest sample size and few carriers, increasing the risk of overfitting—especially in the gene-specific models. This reflects the rarity of ovarian cancer and the challenges of referring patients for genetic counselling and testing across geographically dispersed hospitals (2). While BRCA1/2 prevalence appears broadly similar across Asian populations, underlying genetic architecture and baseline risk profiles may still vary across East Asian, South Asian, and other Asian populations, which could affect model performance and the optimal referral threshold. Our findings should therefore be viewed as exploratory, and external validation in larger, more diverse Asian cohorts—potentially with re-calibration—is needed before broader generalization or use in different settings. We did not include Ashkenazi Jewish ancestry in the model, unlike most existing models, as such variants are extremely rare in Malaysia. Moreover, self-reported ethnicity has been shown in prior studies to be highly concordant with genetic ancestry, making it a reliable measure for our population (41). Our model focused exclusively on BRCA genes because the number of carriers of other susceptibility genes was too small for meaningful analysis. Future work should expand the gene panel and consider incorporating homologous recombination deficiency (HRD) testing for broader risk assessment. We acknowledge that univariable selection of candidate predictors does not capture interactions; however, the included predictors were not highly correlated. Comparison with existing models (e.g., BRCAPRO and CanRisk) was not undertaken because they typically require detailed multigenerational pedigree/family-history inputs that were not consistently available in our cohort, whereas our model uses routinely collected clinical and tumor predictors. Lastly, this base-case cost analysis reflects a cost-threshold scenario rather than a full health economic evaluation, excludes cascade testing and downstream costs that warrant further decision modelling, and the relative advantage of targeted testing will depend on future cost trajectories and health system capacity. Nonetheless, this is the first Asian multi-center study with comprehensive data on clinicopathological and demographic features, marking an important step toward developing a cancer-specific BRCA carrier prediction model tailored for an unselected population and assessing its efficiency in a low-resource setting.
Conclusions
Our study raises important questions about the effectiveness and feasibility of a cancer-specific BRCA carrier prediction model compared to universal testing. Given the substantial need for genetic testing, a tailored approach that takes into account available subsidies and healthcare capacity is essential. In resource-limited settings, our model improves efficiency by minimizing unnecessary testing and prioritizing high-risk individuals, while providing a scalable pathway to broader testing by adjusting referral thresholds to match local resource capacity and genetic testing costs.
Supplementary Material
Acknowledgements
We thank all the participants and their families for taking part in the research studies and all the researchers, clinicians, technicians, and administrative staff, for their outstanding support with recruitment, data collection, and sample management. MyOvCa and MaGiC thank all research staff at Cancer Research Malaysia, University Malaya, participating Ministry of Health Malaysia hospitals, Subang Jaya Medical Centre, Beacon Hospital, Gleneagles Penang, Hospital Universiti Sains Malaysia, KPJ Ampang Puteri Specialist Hospital, KPJ Johor Specialist Hospital, KPJ Sabah Specialist Hospital, Loh Guan Lye Specialist Centre, Mount Miriam Cancer Hospital, Pantai Hospital Kuala Lumpur, Penang Adventist Hospital, Universiti Kebangsaan Malaysia Medical Centre and Sunway Medical Centre who assisted in recruitment and interviews for their contributions and commitment to this study.
We want to thank Siti Norhidayu Hasan, Lau Shao Yan, and Habibatul Saadiah Isa for assistance with DNA preparation for MaGiC; Lee Sheau Yee, Daphne SC Lee, Wong Siu Wan and Lee Yong Quan for their assistance in curating family history data for MyOvCa and MaGiC. We thank Mohamad Faiz Mohamed Jamli and Boon Kiong Lim for their contributions in providing resources for this study.
This work was supported by charitable funding from Yayasan Sime Darby, Yayasan PETRONAS, the Khind Starfish Foundation, and the AstraZeneca External Investigator Grant awarded to S. Y. Yoon. S. H. Teo is supported by the Wellcome Trust (Grant No.: v203477/Z/16/Z). W. K. Ho is supported by the Wellcome Trust Career Development Award (Grant No.: 227824/Z/23/Z).
Footnotes
CONFLICT OF INTEREST
SYY received speaker’s honoraria from Astra Zeneca and she is the vice president of Genetic Counselling Society Malaysia.
Data availability
The data generated in this study are not publicly available because they contain sensitive patient information and are subject to ethical data protection requirements, but are available from the corresponding author on reasonable request and with appropriate approvals.
References
- 1.Ledermann JA, Harter P, Gourley C, Friedlander M, Vergote I, Rustin G, et al. Overall survival in patients with platinum-sensitive recurrent serous ovarian cancer receiving olaparib maintenance monotherapy: an updated analysis from a randomised, placebo-controlled, double-blind, phase 2 trial. The lancet oncology. 2016;17(11):1579–89. doi: 10.1016/S1470-2045(16)30376-X. [DOI] [PubMed] [Google Scholar]
- 2.Yoon S-Y, Wong SW, Lim J, Ahmad S, Mariapun S, Padmanabhan H, et al. Oncologist-led BRCA counselling improves access to cancer genetic testing in middle-income Asian country, with no significant impact on psychosocial outcomes. Journal of medical genetics. 2022;59(3):220–9. doi: 10.1136/jmedgenet-2020-107416. [DOI] [PubMed] [Google Scholar]
- 3.Hasmad HN, Lai KN, Wen WX, Park DJ, Nguyen-Dumont T, Kang PCE, et al. Evaluation of germline BRCA1 and BRCA2 mutations in a multi-ethnic Asian cohort of ovarian cancer patients. Gynecologic oncology. 2016;141(2):318–22. doi: 10.1016/j.ygyno.2015.11.001. [DOI] [PubMed] [Google Scholar]
- 4.Kwon BS, Byun JM, Lee HJ, Jeong DH, Lee TH, Shin K-H, et al. Clinical and genetic characteristics of BRCA1/2 mutation in Korean ovarian cancer patients: a multicenter study and literature review. Cancer Research and Treatment: Official Journal of Korean Cancer Association. 2019;51(3):941–50. doi: 10.4143/crt.2018.312. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.George A, Riddell D, Seal S, Talukdar S, Mahamdallie S, Ruark E, et al. Implementing rapid, robust, cost-effective, patient-centred, routine genetic testing in ovarian cancer patients. Scientific reports. 2016;6(1):29506. doi: 10.1038/srep29506. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Palomaki GE, McClain MR, Steinort K, Sifri R, LoPresti L, Haddow JE. Screen-positive rates and agreement among six family history screening protocols for breast/ovarian cancer in a population-based cohort of 21-to 55-year-old women. Genetics in Medicine. 2006;8(3):161–8. doi: 10.1097/01.gim.0000204458.84988.f2. [DOI] [PubMed] [Google Scholar]
- 7.Kurian AW, Abrahamse P, Furgal A, Ward KC, Hamilton AS, Hodan R, et al. Germline Genetic Testing After Cancer Diagnosis. JAMA. 2023;330(1):43–51. doi: 10.1001/jama.2023.9526. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Daly MB, Pilarski R, Yurgelun MB, Berry MP, Buys SS, Dickson P, et al. NCCN guidelines insights: genetic/familial high-risk assessment: breast, ovarian, and pancreatic, version 1.2020: featured updates to the NCCN guidelines. Journal of the National Comprehensive Cancer Network. 2020;18(4):380–91. doi: 10.6004/jnccn.2020.0017. [DOI] [PubMed] [Google Scholar]
- 9.Esplin ED, Nielsen SM, Bristow SL, Garber JE, Hampel H, Rana HQ, et al. Universal germline genetic testing for hereditary cancer syndromes in patients with solid tumor cancer. JCO Precision Oncology. 2022;6:e2100516. doi: 10.1200/PO.21.00516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lin J, Sharaf RN, Saganty R, Ahsan D, Feit J, Khoury A, et al. Achieving universal genetic assessment for women with ovarian cancer: are we there yet? A systematic review and meta-analysis. Gynecologic oncology. 2021;162(2):506–16. doi: 10.1016/j.ygyno.2021.05.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Nakamura S, Kwong A, Kim S-W, Iau P, Patmasiriwat P, Dofitas R, et al. Current status of the management of hereditary breast and ovarian cancer in Asia: first report by the Asian BRCA consortium. Public health genomics. 2016;19(1):53–60. doi: 10.1159/000441714. [DOI] [PubMed] [Google Scholar]
- 12.Lee A, Mavaddat N, Wilcox AN, Cunningham AP, Carver T, Hartley S, et al. BOADICEA: a comprehensive breast cancer risk prediction model incorporating genetic and nongenetic risk factors. Genetics in Medicine. 2019;21(8):1708–18. doi: 10.1038/s41436-018-0406-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Lindor NM, Johnson KJ, Harvey H, Shane Pankratz V, Domchek SM, Hunt K, et al. Predicting BRCA1 and BRCA2 gene mutation carriers: comparison of PENN II model to previous study. Familial cancer. 2010;9:495–502. doi: 10.1007/s10689-010-9348-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Kang E, Park SK, Lee JW, Kim Z, Noh W-C, Jung Y, et al. KOHBRA BRCA risk calculator (KOHCal): a model for predicting BRCA1 and BRCA2 mutations in Korean breast cancer patients. Journal of human genetics. 2016;61(5):365–71. doi: 10.1038/jhg.2015.164. [DOI] [PubMed] [Google Scholar]
- 15.Ang BH, Ho WK, Wijaya E, Kwan PY, Ng PS, Yoon SY, et al. Predicting the likelihood of carrying a BRCA1 or BRCA2 mutation in Asian breast cancer patients. Journal of clinical oncology: official journal of the American Society of Clinical Oncology. 2022;40(14):1542. doi: 10.1200/JCO.21.01647. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kwong A, Wong CH, Suen DT, Co M, Kurian AW, West DW, et al. Accuracy of BRCA1/2 mutation prediction models for different ethnicities and genders: experience in a southern Chinese cohort. World journal of surgery. 2012;36:702–13. doi: 10.1007/s00268-011-1406-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hung F-H, Wang YA, Jian J-W, Peng H-P, Hsieh L-L, Hung C-F, et al. Evaluating BRCA mutation risk predictive models in a Chinese cohort in Taiwan. Scientific Reports. 2019;9(1):10229. doi: 10.1038/s41598-019-46707-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Kommoss S, Gilks CB, du Bois A, Kommoss F. Ovarian carcinoma diagnosis: the clinical impact of 15 years of change. British journal of cancer. 2016;115(8):993–9. doi: 10.1038/bjc.2016.273. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Jones PM, Drapkin R. Modeling high-grade serous carcinoma: how converging insights into pathogenesis and genetics are driving better experimental platforms. Frontiers in oncology. 2013;3:217. doi: 10.3389/fonc.2013.00217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Sterne JA, White IR, Carlin JB, Spratt M, Royston P, Kenward MG, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. Bmj. 2009;338 doi: 10.1136/bmj.b2393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Hirasawa A, Imoto I, Naruto T, Akahane T, Yamagami W, Nomura H. Prevalence of pathogenic germline variants detected by multigene sequencing in unselected Japanese patients with ovarian cancer. Oncotarget. 2017;8(68):112258–67. doi: 10.18632/oncotarget.22733. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Schrijver LH, Antoniou AC, Olsson H, Mooij TM, Roos-Blom M-J, Azarang L, et al. Oral contraceptive use and ovarian cancer risk for BRCA1/2 mutation carriers: an international cohort study. American Journal of Obstetrics and Gynecology. 2021;225(1):51.e1–e17. doi: 10.1016/j.ajog.2021.01.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982;143(1):29–36. doi: 10.1148/radiology.143.1.7063747. [DOI] [PubMed] [Google Scholar]
- 24.Archer KJ, Lemeshow S. Goodness-of-fit test for a logistic regression model fitted using survey sample data. The Stata Journal. 2006;6(1):97–105. [Google Scholar]
- 25.Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clinical chemistry. 1993;39(4):561–77. [PubMed] [Google Scholar]
- 26.Padmanabhan H, Lim KK, Hassan NT, Ahmad Bashah NS, Lee Y-Q, Lim J, Teo IH, Gunasagran Y, Kalimuthu RK, Omar J, Mohd Abas MN, et al. Exploring willingness to pay out-of-pocket for genetic testing among ovarian cancer patients in Malaysia, a middle-income country in Asia. Journal of Gynecologic Oncology. 2025;(37):e32. doi: 10.3802/jgo.2026.37.e32. In press. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Bu H, Chen J, Li Q, Hou J, Wei Y, Yang X, et al. BRCA mutation frequency and clinical features of ovarian cancer patients: A report from a Chinese study group. Journal of Obstetrics and Gynaecology Research. 2019;45(11):2267–74. doi: 10.1111/jog.14090. [DOI] [PubMed] [Google Scholar]
- 28.Enomoto T, Aoki D, Hattori K, Jinushi M, Kigawa J, Takeshima N, et al. The first Japanese nationwide multicenter study of BRCA mutation testing in ovarian cancer: CHARacterizing the cross-sectionaL approach to Ovarian cancer geneTic TEsting of BRCA (CHARLOTTE) International Journal of Gynecologic Cancer. 2019;29(6) doi: 10.1136/ijgc-2019-000384. [DOI] [PubMed] [Google Scholar]
- 29.Gupta S, Rajappa S, Advani S, Agarwal A, Aggarwal S, Goswami C, et al. Prevalence of BRCA1 and BRCA2 mutations among patients with ovarian, primary peritoneal, and fallopian tube cancer in India: a multicenter cross-sectional study. JCO Global Oncology. 2021;7:849–61. doi: 10.1200/GO.21.00051. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Paik ES, Heo EJ, Choi CH, Kim JH, Kim JW, Kim YM, et al. Prevalence and clinical characterization of BRCA1 and BRCA2 mutations in Korean patients with epithelial ovarian cancer. Cancer Science. 2021;112(12):5055–67. doi: 10.1111/cas.15166. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Kroeger PT, Jr, Drapkin R. Pathogenesis and heterogeneity of ovarian cancer. Current Opinion in Obstetrics and Gynecology. 2017;29(1):26–34. doi: 10.1097/GCO.0000000000000340. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Botesteanu DA, Lipkowitz S, Lee JM, Levy D. Mathematical models of breast and ovarian cancers. Wiley Interdisciplinary Reviews: Systems Biology and Medicine. 2016;8(4):337–62. doi: 10.1002/wsbm.1343. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Antoniou AC, Pharoah P, Smith P, Easton DF. The BOADICEA model of genetic susceptibility to breast and ovarian cancer. British journal of cancer. 2004;91(8):1580–90. doi: 10.1038/sj.bjc.6602175. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Parmigiani G, Chen S, Iversen ES, Jr, Friebel TM, Finkelstein DM, Anton-Culver H, et al. Validity of models for predicting BRCA1 and BRCA2 mutations. Annals of internal medicine. 2007;147(7):441–50. doi: 10.7326/0003-4819-147-7-200710020-00002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Carver T, Hartley S, Lee A, Cunningham AP, Archer S, Babb de Villiers C, et al. CanRisk Tool—A web interface for the prediction of breast and ovarian cancer risk and the likelihood of carrying genetic pathogenic variants. Cancer epidemiology, biomarkers & prevention. 2021;30(3):469–73. doi: 10.1158/1055-9965.EPI-20-1319. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Network NCC. 2025 Genetic/Familial High-Risk Assessment: Breast, Ovarian, Pancreatic, and Prostate. 2025.
- 37.Soon JJH, Wu J, Ishak NDB, See WQ, Dorosan M, Yuen J, et al. Leveraging Electronic Health Records to Examine the Real-World Rates of Cancer Genetics Referrals in a Singapore Health Care Cluster. JCO Oncology Practice. 2026;22(1):91–9. doi: 10.1200/OP-24-00858. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.DiSilvestro P, Banerjee S, Colombo N, Scambia G, Kim B-G, Oaknin A, et al. Overall survival with maintenance olaparib at a 7-year follow-up in patients with newly diagnosed advanced ovarian cancer and a BRCA mutation: the SOLO1/GOG 3004 trial. Journal of Clinical Oncology. 2023;41(3):609–17. doi: 10.1200/JCO.22.01549. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Koldehoff A, Danner M, Civello D, Rhiem K, Stock S, Müller D. Cost-Effectiveness of targeted genetic testing for breast and ovarian cancer: a systematic review. Value in Health. 2021;24(2):303–12. doi: 10.1016/j.jval.2020.09.016. [DOI] [PubMed] [Google Scholar]
- 40.Lertwilaiwittaya P, Tantai N, Maneeon S, Kongbunrak S, Nonpanya N, Hurst AC, et al. A cost-utility analysis of BRCA1 and BRCA2 testing in high-risk breast cancer patients and family members in Thailand: a cost-effective policy in resource-limited settings. Frontiers in Public Health. 2023;11:1257668. doi: 10.3389/fpubh.2023.1257668. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Ho W-K, Tai M-C, Dennis J, Shu X, Li J, Ho PJ, et al. Polygenic risk scores for prediction of breast cancer risk in Asian populations. Genetics in Medicine. 2022;24(3):586–600. doi: 10.1016/j.gim.2021.11.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data generated in this study are not publicly available because they contain sensitive patient information and are subject to ethical data protection requirements, but are available from the corresponding author on reasonable request and with appropriate approvals.