Development of a risk score based on clinical–pathological features to predict the presence of germline BRCA1/2 pathogenic variants in ovarian cancer patients

Abstract

Background

Identification of germline BRCA1/2 pathogenic variants is crucial for tailoring ovarian cancer treatment and prevention. The purpose of the study was to develop a model to predict BRCA1/2 status in ovarian cancer patients.

Patients and methods

The association between clinical–pathological features and BRCA1/2 status was analysed in a series of 1009 ovarian cancer patients, using Fisher’s exact test. Logistic regression models and a decision tree classification algorithm were used to develop a risk score.

Results

Compared with noncarriers, BRCA1/2 carriers (n = 216; 21.4%) presented more frequently with serous histotype non-low-grade (92.3% versus 71.6%, P < 0.001), family history of ovarian cancer (31.6% versus 5.7%, P < 0.001), family history of breast cancer (53.7% versus 31.6%, P < 0.001), previous breast cancer (20.9% versus 8.5%, P < 0.001), advanced stage (78.8% versus 69.5%, P = 0.019) and younger age (56.9 years versus 60.8 years, P < 0.001). Multivariable logistic regression on 648 patients with complete data confirmed histotype, family history of breast/ovarian cancer, previous breast cancer and age as independently and significantly associated with BRCA1/2 status. After refining the categorization of variables according to decision tree classification algorithm results, odds ratios derived from multivariable logistic regression were used to assign weights from 1 to 3 to each feature (non-low-grade serous histotype = 3, low-grade serous/high-grade endometrioid histotype/family history of ovarian cancer = 2, age at diagnosis <50 years/family and personal history of breast cancer = 1) and to develop a score ranging from 0 to 10, associated with increasing risk of BRCA1/2 variants (from 0.6% for score 0 to 88% for score ≥7). The area under the curve of the score was 0.78 (95% confidence interval 0.74-0.82) and the optimal cut-off was ≥4 points with a sensitivity of 81% and a specificity of 62.3%.

Conclusions

The proposed risk score may improve assessment and counselling of ovarian cancer patients.

Highlights

• •BRCA testing uptake in ovarian cancer is still suboptimal; mainstreaming may hamper genetic counselling personalization.
• •Predicting the likelihood of BRCA1/2 variants would help tailoring pretest counselling and prioritizing testing.
• •A score predicting the probability of BRCA1/2 variants was developed based on a few, widely available, clinical features.

Introduction

Although relatively uncommon, epithelial ovarian cancer (OC) remains an international public health challenge due to its high mortality rate.^1,2 Among several factors involved in its aetiology, germline pathogenic variants in the BRCA1/2 genes are known to account for up to 20% of cases.3, 4, 5 Since carriers of these variants benefit from targeted therapies such as poly (ADP-ribose) polymerase (PARP) inhibitors,^6,7 as well as from specific prevention of additional tumours including breast cancer (BC), BRCA1/2 gene testing has become crucial in the clinical management of OC patients,^8,9 This led scientific societies and health organizations to recommend ‘universal’ BRCA testing in women with epithelial OC (with the exception of mucinous and borderline types). As the limited availability of genetic counselling services would have hampered testing access and/or timing, it was also recommended that BRCA test is directly ordered by oncologists and gynaecologists after a focused pretest counselling session through a ‘mainstream’ testing approach.^10,11

Nevertheless, implementation of universal OC testing in the real world is far to be achieved, with BRCA analysis still underutilized worldwide. If in underdeveloped/developing countries availability of cancer genetics services is limited,¹² and testing all OC patients may not be economically affordable, even in high-income countries uptake rates are yet suboptimal, ranging from <60% to 70% (increasing up to 79% after an improvement initiative).13, 14, 15

Moreover, from the patient’s perspective, mainstreaming may represent a suboptimal approach if compared with traditional path. Indeed, the paradigm shift from comprehensive genetic counselling delivered by genetic health care professionals to short pretest sessions provided by nongenetic professionals has reduced opportunities to receive extensive and tailored information and to make informed choices. Women undergoing mainstream BRCA testing showed poorer knowledge and reported not having the opportunity of making an informed decision about genetic testing; in addition, nongenetic health professionals involved in the mainstreaming model often expressed limited familiarity with genetics and little confidence with genetic counselling.16, 17, 18

The knowledge of the individual likelihood of carrying a BRCA defect may help select, prioritize or reinforce recommendation about testing whenever needed and would enable oncologists and gynaecologists to provide tailored information in pretest counselling.

Prior probability is influenced by several clinical and pathological features: risks of carrying a germline BRCA1/2 pathogenic variant associated with each feature have been recently estimated through a systematic review with meta-analysis¹⁹ and may be combined to assess the probability of any OC patient to carry a germline BRCA1/2 pathogenic variant before testing.

The aim of this study was to develop a risk score predicting the presence of germline BRCA1/2 pathogenic variants on the basis of clinical features of OC patients, taking advantage of the evidence emerged from the systematic review and of the availability of a large cohort of patients.

Materials and methods

Study sample

The study included all epithelial OC patients tested for BRCA1/2 from 2012 to 2022 at the Bologna HUB of the Emilia-Romagna Network for genetic risk of BC and OC, according to the regional protocol.^20,21 For each patient, information on clinical–pathological features was collected, including those identified as predictive in a previous meta-analysis (personal history of BC, stage and age at diagnosis).¹⁹ Unlike in the meta-analysis, grade and histotype were not considered independent variables but were integrated into the same category in accordance with the updated classification criteria (https://www.pathologyoutlines.com/topic/ovarytumorwhoclassif.html)²² and the presence of BC and OC in first/second-degree relatives, which had not been assessed in the meta-analysis because it could not be extrapolated from most papers, was systematically collected in the present study.

BRCA test results were interpreted according to the ENIGMA Variant Curation Expert Panel classification criteria (https://enigmaconsortium.org/)^23,24: the detection of germline pathogenic or likely pathogenic variants identified ‘BRCA1/2 carriers’.

The study was approved by the Ethics Committee of ‘Area Vasta Emilia Centro’ of Emilia-Romagna region (CB-AVEC), Italy (490/2022 /Oss/AOUBo) on 23 August 2022. All patients provided a written informed consent to the use of their data for research purposes; the dataset was pseudonymised before statistical analysis.

Statistical analyses

The frequencies of dichotomous variables were compared between carriers and noncarriers of BRCA1/2 pathogenic variants, using Fisher’s exact test.

A multivariable logistic regression model was used to predict the probability of being carriers of a BRCA1/2 germline pathogenic variant as a function of personal BC history, histotype, grade, Federation of Gynaecology and Obstetrics (FIGO) stage, age at diagnosis, OC family history and BC family history. Five dummy variables were created for histotype, representing the contrasts of histotypes 2 (low-grade serous), 3 (high-grade endometrioid), 4 (low-grade endometrioid), 5 (clear cell), and 7 (mesonephric-like adenocarcinoma) versus histotype 1 (non-low-grade serous), which was used as the reference category. The latter category included patients with pathology reports of: ‘high-grade serous carcinoma’, ‘serous carcinoma’ without the specification of grade, ‘undifferentiated carcinoma’ and ‘mixed histotypes with a high-grade serous component’, which were shown to be comparable BRCA predictors in the meta-analysis.¹⁹

Then, a decision tree classification algorithm was used to refine the categorization of histotype and to identify subgroups with a different probability of being carriers of a BRCA1/2 germline pathogenic variant as a function of combinations of the same variables listed in the preceding text. This supervised learning algorithm uses a decision tree to classify patients into mutually exclusive subgroups based on the values of predictor variables. The tree is grown using the Chi-square Automatic Interaction Detection (CHAID) algorithm. At each step, CHAID selects the predictor that has the strongest association with the dependent variable. Categories of each predictor are merged if they are not significantly different with respect to the dependent variable.

A cross-validation was carried out dividing the sample into 10 subsamples, or folds. The cross-validated risk estimate for the final tree was calculated as the average of the risks for all of the trees. This analysis was carried out using IBM SPSS, version 28 (IBM, Armonk, NY).

The results of the decision tree classification algorithm were used to improve the categorization of variables to be included in the logistic regression model, and to avoid sparse data. Specifically, age at diagnosis was dichotomised as <50 years and ≥50 years, and histotypes were grouped into three categories (1 = non-low-grade serous; 2 = low-grade serous and high-grade endometrioid, 3 = other histotypes), with the third used as the reference category. A weight was then assigned to each included variable, and these values were summed to compute a risk score. Specifically, a weight of 1 was assigned to the variable with the lowest odds ratio (OR) and the weights of other variables were reparametrized proportionally.

Receiver operating characteristic (ROC) analysis was then used to analyse the performance of each variable and of the predictive score in terms of area under the curve (AUC). The optimal cut-off balancing sensitivity and specificity was the one maximizing Youden’s J index, which is equal to sensitivity + specificity – 1.

Results

Patient features

From 2012 to 2022, 1009 OC patients were tested for BRCA1/2 at the Medical Genetics Laboratory of IRCCS Policlinico di Sant’Orsola of Bologna; among those, 216 (21.4%) carried germline BRCA1/2 pathogenic variants. In 497 patients (49.5%), germline-only DNA testing for both sequence and copy number variants (CNVs) in BRCA1 and BRCA2 genes was carried out, whereas 512 patients (50.5%) underwent BRCA1/2 sequence analysis in tumour DNA, followed by blood testing aimed at checking sequence variants found in tissue, if any, through Sanger sequencing, and at detecting CNVs using multiplex ligation-dependent probe amplification (MLPA). In the latter group, 27 patients (5.3%) were found to harbour exclusively somatic pathogenic variants, thus were considered as noncarriers for statistical analyses.

The major features of patients included in the dataset are summarized in Table 1 and reported in more detail in Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2025.105300.

Table 1.

Clinical–pathological features of patients included in the dataset

Feature		BRCA1/2 carriers (N = 216)	Noncarriers (N = 793)	P value
Histotype, n (%)	Missing data	8	43	—
	Non-low-grade serousa Other histotypes •Low-grade serous •High-grade endometrioidb •Low-grade endometrioidc •Clear cell •Mesonephric-like adenocarcinoma	192 (92.3) 16 (7.7) •1 (0.5) •11 (5.3) •3 (1.4) •1 (0.5) •0 (0.0)	537 (71.6) 213 (28.4) •29 (3.9) •45 (6.0) •81 (10.8) •52 (6.9) •6 (0.8)	<0.001d
FIGO stage, n (%)	Missing data	51	209	—
	I-II	35 (21.2)	178 (30.5)	0.019d
	III-IV	130 (78.8)	406 (69.5)
Age at diagnosis, years	Missing data	1	17	—
	Mean (standard deviation)	56.9 (11.1)	60.8 (12.0)	<0.001e
	≤50, n (%)	63 (29.3)	165 (21.3)	0.017d
	>50, n (%)	152 (70.7)	611 (78.7)
Personal breast cancer history, n (%)	Missing data	5	59	—
	Negative	167 (79.1)	671 (91.4)	<0.001d
	Positive	44 (20.9)	63 (8.5)
Family history of ovarian cancer, n (%)	Missing data	7	97	—
	Negative	143 (68.4)	656 (94.3)	<0.001d
	Positive	66 (31.6)	40 (5.7)
Family history of breast cancer, n (%)	Missing data	2	96	—
	Negative	99 (46.3)	477 (68.4)	<0.001d
	Positive	115 (53.7)	220 (31.6)

FIGO, International Federation of Gynecology and Obstetrics.

^aHigh-grade serous: 172/630 (27.3%); unspecified serous: 5/24 (20.8%); mixed tumour with high-grade serous component: 4/25 (16.0%); carcinosarcoma with high-grade serous component: 1/7 (14.3%); undifferentiated carcinoma: 10/43 (23.3%).

^bHigh-grade endometrioid: 10/51 (19.6%); mixed tumour with high-grade endometrioid component: 0/3 (0.0%); carcinosarcoma with high-grade endometrioid component: 1/2 (50.0%).

^cLow-grade endometrioid: 3/81 (3.7%); mixed tumour with low-grade endometrioid component: 0/3 (0.0%).

^dChi-square test.

^et-test.

Among the 1009 patients in the dataset, 648 had complete data for all the features of interest and were selected for subsequent statistical analyses. No significant differences were found between included and nonincluded patients for any of the analysed variables (Supplementary Table S2, available at https://doi.org/10.1016/j.esmoop.2025.105300).

Predictors of germline BRCA1/2 pathogenic variants

In the logistic regression model, all variables analysed except FIGO stage were significantly associated with the presence of germline BRCA1/2 pathogenic variants, as shown in Table 2. Older patients and those with low-grade endometrioid (compared with non-low-grade serous) tumours were at lower risk of carrying a BRCA1/2 germline pathogenic variant. The presence of a family history of OC was associated with a 9.5-fold risk, while family and a personal history of BC were associated with a 1.8-fold and 4-fold risk, respectively.

Table 2.

Logistic regression results (significant variables are in boldface)

Coefficients	Estimate	Standard error	OR	P value	95% confidence interval (OR scale)
(Intercept)	1.015	0.639	2.759
Age <50 years	0.346	0.171	2.021	0.043	1.010	1.977
Stage (reference: III-IV)	0.135	0.286	1.145	0.637	0.653	2.006
Family history of OC	2.253	0.298	9.515	<0.001	5.311	17.049
Family history of BC	0.582	0.218	1.789	0.008	1.167	2.743
Personal history of BC	1.389	0.324	4.010	<0.001	2.126	7.562
Histotype 2 (versus 1)	−1.743	1.083	0.175	0.108	0.021	1.462
Histotype 3 (versus 1)	−0.927	0.502	0.396	0.065	0.148	1.058
Histotype 4 (versus 1)	−3.493	1.062	0.030	0.001	0.004	0.244
Histotype 5 (versus 1)	−16.714	563.406	5.512 × 10−8	0.976	0.000	∞
Histotype 7 (versus 1)	−16.597	2709.241	6.196 × 10−8	0.995	0.000	∞

Histotype 1, non-low-grade serous; histotype 2, low-grade serous; histotype 3, high-grade endometrioid; histotype 4, low-grade endometrioid; histotype 5, clear cell; histotype 7, mesonephric-like adenocarcinoma.

BC, breast cancer; OC, ovarian cancer; OR, odds ratio.

The Nagelkerke R² goodness of fit index was satisfactory and indicated that 35% of the variability of the dependent variable was explained by the predictors. Performance diagnostics showed high accuracy (81.2%) and AUC (79.1%), and a better performance in detecting noncases (specificity: 95.2%) than cases (sensitivity: 35.5%), as summarized in Supplementary Table S3, available at https://doi.org/10.1016/j.esmoop.2025.105300.

Decision tree classification

The cross-validated tree partitioned the sample into six homogeneous subgroups of patients (nodes 3, 4, 6, 7, 8 and 9 in Figure 1) with different probabilities (from 0% to 75%) of carrying a germline BRCA1/2 pathogenic variant. The first split divided patients into those with (node 1) and those without (node 2) a family history of OC. Patients in node 1 were subsequently split into two nodes according to the stage. The risk of being a carrier of a germline BRCA1/2 pathogenic variant was 75% for stage III/IV (node 3) and 42.3% for stage I/II (node 4).

Figure 1.

Decision tree classification model. In each node the number and the percentage with and without BRCA1/2 pathogenic variants is provided in tabular form and as a bar diagram. Carriers of BRCA1/2 pathogenic variants are shown in red. Histotype 1, non-low-grade serous; histotype 2, low-grade serous; histotype 3, high-grade endometrioid; histotype 4, low-grade endometrioid; histotype 5, clear cell; histotype 7, mesonephric-like adenocarcinoma. Adj., adjusted; BC, breast cancer; FIGO, Federation of Gynaecology and Obstetrics; OC, ovarian cancer.

Among those in node 2, histotype produced a split into three nodes: one including patients with histotypes 4 (low-grade endometrioid), 5 (clear cell) or 7 (mesonephric-like adenocarcinoma), who had a 0.0% risk (node 6), one including those with histotypes 2 (low-grade serous) and 3 (high-grade endometrioid), who had a 9.4% risk (node 7), and the last including patients with histotype 1 (non-low-grade serous—node 5), in whom the risk was 45% (node 8) in the presence of a personal history of BC and 19.7% in the absence (node 9). The overall accuracy of the model was 80.9%, the positive predictive value (PPV) was 75.0% (42/56) and the sensitivity was 27.6% (Supplementary Table S4, available at https://doi.org/10.1016/j.esmoop.2025.105300).

Development of the risk score

In the revised multiple logistic regression model, including a new categorization of histotype, all variables except FIGO stage were significantly associated with the presence of germline BRCA1/2 pathogenic variants (Table 3). The categories ‘non-low-grade serous histotype’ and ‘low-grade serous + high-grade endometrioid histotypes’ were found to be associated with 48.5-fold and 17.3-fold risk compared with their reference category (‘other histotypes’), but with large confidence intervals (CIs). The categories ‘family history of OC’, ‘family history of BC’ and ‘personal history of BC’ were instead associated with similar increases in risk with those in the previous model.

Table 3.

Revised multiple logistic regression model and weight assignment (variables significantly associated with germline BRCA1/2 pathogenic variants are in boldface)

Variables	b	Standard error (b)	P value	OR	95% confidence interval		Relative OR	Weight
					Low	High
Age at diagnosis <50 years	0.937	0.249	<0.001	2.551	1.565	4.160	1.437	1
Histotype 1 versus Histotypes 4/5/7	3.881	1.044	<0.001	48.453	6.257	375.206	27.294	3
Histotype 2/3 versus Histotypes 4/5/7	2.849	1.112	0.010	17.269	1.953	152.661	9.728	2
FIGO stage	0.099	0.283	0.726	1.104	0.635	1.921	—	—
OC family history	2.286	0.294	<0.001	9.839	5.533	17.495	5.543	2
BC family history	0.574	0.216	0.008	1.775	1.163	2.710	1.000	1
BC personal history	1.300	0.318	<0.001	3.669	1.968	6.840	2.067	1

Histotype 1, non-low-grade serous; histotype 2, low-grade serous; histotype 3, high-grade endometrioid; histotype 4, low-grade endometrioid; histotype 5, clear cell; histotype 7, mesonephric-like adenocarcinoma.

BC, breast cancer; FIGO, FIGO, International Federation of Gynecology and Obstetrics; OC, ovarian cancer; OR, odds ratio.

To develop the risk score, all variables were used except ‘FIGO stage’, which was not statistically significant, for a total of six variables. Relative ORs were obtained as a division by the OR of the family history of BC.

The variable ‘family history of BC’ was assigned a weight of 1, as it was the variable associated with the lowest OR. Weights of one were assigned when the relative OR was <2, weights of 2 when the relative OR was from 2 to 5 and weights of 3 afterwards.

The only variable with a weight of 3 was ‘non-low-grade serous histotype’; the categories ‘low-grade serous + high-grade endometrioid histotypes’ and ‘family history of OC’ had a weight of 2, while all others a weight of 1. In this way, the score potentially ranges from 0 to 10.

The ROC analysis showed that the score had a higher AUC (0.778, 95% CI 0.738-0.818) compared with each of its components (Supplementary Figure S1 and Table S5, available at https://doi.org/10.1016/j.esmoop.2025.105300).

As shown in Supplementary Table S6, available at https://doi.org/10.1016/j.esmoop.2025.105300, the optimal cut-off maximising Youden’s index and balancing sensitivity and specificity was ≥4 points.

Logistic regression analysis, using the risk score as the only predictor, indicated that for each additional point of the score, the OR increased by 2.76 (95% CI 2.27-3.37). Figure 2 shows the estimated probability of being a carrier of germline BRCA1/2 pathogenic variants as a function of the risk score in the logistic regression model: patients with a risk score of 0 had a negligible chance (0.6%) to be carriers of a germline BRCA1/2 pathogenic variant, while those with a score ≥4 had a 26% probability, that increased up to 88% for those scoring 7 or more.

Figure 2.

Probability of being a carrier of germline BRCA1/2 pathogenic variants as a function of the risk score.

Discussion

In the era of precision medicine, BRCA1/2 genetic testing is being increasingly incorporated in the clinical pathway of OC patients due to its significant clinical and familial implications,^6,7 and rapid identification of carriers of germline pathogenic variants is currently crucial for proper management. However, the identification of patients who a priori are more likely to carry these variants may have a significant impact on individualized pretest counselling and test prioritization.

In order to develop an approach to assess the chance of testing positive for germline BRCA1/2 pathogenic variants, we analysed a cohort of 1009 OC patients who underwent BRCA testing at the Laboratory of the Medical Genetics Unit of the IRCCS Policlinico Sant’Orsola of Bologna, Italy, from 2012 to 2022, of whom 216 (21.4%) were found to carry germline BRCA1/2 pathogenic/likely pathogenic variants.

In the study population, germline BRCA1/2 pathogenic variants were significantly more frequent in patients with non-low-grade serous/undifferentiated tumours and advanced stage, age at diagnosis ≤50 years, a personal history of BC and a family history of BC/OC (Table 1), in line with the findings of our previous meta-analysis.¹⁹

Among the features established as predictive in the meta-analysis, a personal history of BC was the strongest predictor in the present study, followed by age at diagnosis and histotype: the latter, however, only reached statistical significance when comparing low-grade endometrioid tumours with non-low-grade serous tumours; the comparison of the other histotypes with non-low-grade serous tumours failed to reveal statistically significant differences, mainly due to sparse data (in the case of low-grade serous and high-grade endometrioid) or because there were no BRCA1/2 pathogenic variant carriers with those tumour types (clear cell and mesonephric-like adenocarcinoma), so that their effect could not be estimated in the model. Among other features, family history of OC and BC were both significantly correlated with the presence of germline BRCA1/2 pathogenic variants, with the former being the most strongly associated (OR 9.8, more than twice that of personal history of BC).

The logistic regression model achieved high specificity (95.2%) but low sensitivity (35.3%), as expected based on the imbalance between BRCA1/2 carriers and noncarriers (216 versus 793), which, however, is in line with previous reports.3, 4, 5

The decision tree classification algorithm provided a model of potential clinical value: based on only four, widely available, features (family history of OC, tumour stage, tumour histotype and personal history of BC). It allowed the stratification of OC patients into six homogeneous and well-defined groups, whose probability of carrying a germline BRCA1/2 pathogenic variant ranged from 75.0% (patients with a family history of OC and advanced-stage tumours) to 0 (patients without a family history of OC and with low-grade endometrioid, clear cell or mesonephric-like tumours). Again, sensitivity was low (27.6%), but overall accuracy and PPV were satisfactory (80.9% and 75.0%, respectively).

Finally, the revised logistic regression, recalibrated according to the decision tree classification, allowed the generation of a risk score that, on the basis of a limited number of clinical–pathological features (age at diagnosis, histotype, family history of BC/OC and personal history of BC), stratified OC patients into 10 risk levels associated with increasing probabilities of carrying germline BRCA1/2 pathogenic variants: almost null, in the case of a score of 0, up to almost 90% in the case of a score ≥7. Although the AUC of the score (0.778) does not reach the threshold of 0.8 generally adopted to establish diagnostic accuracy, for a probabilistic predictor, unlike for a diagnostic tool, an AUC close to 0.8 with a narrow confidence interval (0.738-0.818) may be considered adequate to support the value of the model.

This risk score is expected to be easy to use in the physician’s practice. Indeed, accurate tools to predict the presence of BRCA variants, such as the CanRisk Tool (https://www.canrisk.org)²⁵ are already available, but require the entry of a lot of information that is not easy to obtain, such as a comprehensive family history, the collection of which goes beyond the efforts that nongeneticist health care providers may make in pretest assessment. Therefore, these tools are only used by cancer genetics health care professionals in the context of ‘traditional’ genetic counselling procedures, while other specialists usually provide patients with standard information on the goal of testing, irrespective of their prior probability of being BRCA1/2 carriers. A simple tool enabling the incorporation of a limited number of variables that are quickly and easily available would be more accessible to them; the assessment of the chance for the patient to be a carrier would raise awareness among patients and professionals, increasing the effectiveness of pretest counselling and enabling oncologists and gynaecologists to make more personalised and tailored recommendations.

Beyond adding value to individual assessment and counselling, the opportunity to quantify the probability of detecting BRCA variants may allow the prioritizing of testing or strengthen the recommendations for those OC patients who are most likely to benefit from the result. Indeed, although guidelines recommend universal BRCA1/2 testing in patients with epithelial OC because of therapeutic, as well as preventive, implications,^8,9 genetic testing remains underused worldwide.^13,14 Among barriers, there are societal or individual economic constraints, racial disparities, insurance coverage issues, and gaps in the knowledge of patients and physicians. Setting a threshold and selecting for testing the subgroup of patients at higher risk, whenever needed, would reduce costs in constrained contexts, allow tailored insurance coverage, and support patient decision making if they have to pay themselves.

Finally, the risk score may contribute to the classification of variants of unknown clinical significance (VUS), which are detected in 6%-7% of BRCA tests and pose relevant challenges for counselling and management of carriers.²⁶ Indeed, the a priori probability of identifying BRCA1/2 pathogenic variants is one of the factors included in multifactorial likelihood modelling,²⁷ the result of which, combined with other evidence, is used to classify the variant.²⁴

A potential limitation of this study is that it is based on retrospective data from a single institution. To confirm its performance, the score needs to be validated in multicentre prospective studies.

In conclusion, we developed a score for predicting the risk of being carriers of germline BRCA1/2 pathogenic variants in OC patients, helping patients to receive a more personalized pretest genetic counselling, and allowing health services to prioritize genetic testing based on the probability of clinical benefit.