Clinical Impact of Polygenic Risk Score for Breast Cancer Risk Prediction in 382 Individuals with Hereditary Breast and Ovarian Cancer Syndrome

Simple Summary Breast cancer (BC) is the major cause of cancer-related deaths in women worldwide. In addition to genetic diagnostics for variants in high-risk genes, there is a need for better risk stratification to target high-risk individuals. The polygenic risk score (PRS) has emerged as a valuable addition to help sorting women into different risk categories for BC development. This study aimed to evaluate the impact of adding a PRS, based on 313 genetic variants, to standard genetic testing for 382 German women with BC or a family history of the disease. By incorporating the PRS into risk prediction models, meaningful changes in 10-year risks were observed in 13.6% of individuals. Additionally, the inclusion of the PRS led to clinically significant changes in prevention recommendations for 12.0% of cases, supporting the use of the PRS for BC risk assessment in genetic counselling. Abstract Single nucleotide polymorphisms are currently not considered in breast cancer (BC) risk predictions used in daily practice of genetic counselling and clinical management of familial BC in Germany. This study aimed to assess the clinical value of incorporating a 313-variant-based polygenic risk score (PRS) into BC risk calculations in a cohort of German women with suspected hereditary breast and ovarian cancer syndrome (HBOC). Data from 382 individuals seeking counselling for HBOC were analysed. Risk calculations were performed using the Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm with and without the inclusion of the PRS. Changes in risk predictions and their impact on clinical management were evaluated. The PRS led to changes in risk stratification based on 10-year risk calculations in 13.6% of individuals. Furthermore, the inclusion of the PRS in BC risk predictions resulted in clinically significant changes in 12.0% of cases, impacting the prevention recommendations established by the German Consortium for Hereditary Breast and Ovarian Cancer. These findings support the implementation of the PRS in genetic counselling for personalized BC risk assessment.


Introduction
Breast Cancer (BC) is the most prevalent form of cancer in women and the leading cause of cancer-related death among women worldwide [1]. To facilitate early detection of BC, improve chances of recovery, and decrease mortality rates, screening programs have been established [2,3]. However, current screening strategies, such as clinical examinations and mammograms, are vulnerable to overdiagnosis and overtreatment [4,5].
Personalized risk estimations of BC could help to improve prevention and screening programs by identifying women in risk categories who are most likely to benefit [6].

Study Cohort
We collected data from women from families with suspected HBOC and from families in which a PV in the causative genes for HBOC is already known from December 2020 to January 2023. All participants sought counselling for HBOC at the Institute of Human Genetics of the University of Leipzig, Germany, which is part of the Centre for Hereditary Breast and Ovarian Cancer Leipzig. A total of 382 women met the following eligibility criteria for this study: (1) Availability of genotyping data; (2) Availability of personal medical information (e.g., cancer diagnosis); (3) Available information on family history (e.g., family members with a cancer diagnosis); (4) Age range of 18 to 69 years; (5) The presence of a PV for which a risk calculation using CanRisk was possible in case of a positive carrier status.
Women with BC or OC were included if they met additional criteria: (1) Available information on tumour pathology; (2) Available clinical data such as age of first onset of disease; (3) Unilaterality in women with BC.
For individuals with unilateral BC, the risk of developing a contralateral tumour was calculated. Women with bilateral BC, ductal carcinoma in situ (DCIS), or pancreatic cancer could not be considered as they did not meet the requirements for BOADICEA computation [32]. All participants provided informed consent for their data to be used for research purposes. All individuals were anonymized.

Pedigree Collection and Initial Risk Assessment
The pedigrees included in this work were generated using the PhenoTips webtool [33] during the initial consultation. All participants' pedigrees include at least three generations. The following data were available for each individual in the pedigrees: (1) Life status; (2) Year of birth and current age; (3) Cancer status.
The age of first onset of disease was known for all index patients. All participants underwent an additional risk assessment using the checklist of The German Cancer Society [34,35]. The checklist is based on associated cases of HBOC in the maternal and paternal line, which are included with either single, double, or triple weighting while also considering the age of first onset of disease and the hormone receptor status in BC cases. Individuals with a score of ≥3 are considered to have a risk of carrying a PV of at least 10% and are thus eligible for genetic testing [36][37][38].

Molecular Genetics
All women from families with suspected HBOC underwent panel diagnostics including PRS testing. Some of the women, in whom a PV was known in the family, received the targeted testing within the framework of panel diagnostics including PRS testing, the rest received targeted testing by means of Sanger sequencing and additional PRS testing.
Genomic DNA was extracted from whole blood using MagCore Kit 101 and MagCore ® instrument (RBC Bioscience, New Taipei City, Taiwan). DNA concentration was measured using NanoDrop™ 2000 (Thermo Scientific™, Waltham, MA, USA) and Qubit (Thermo Sci-entific™,Waltham, MA, USA). Next generation sequencing (NGS) after sample preparation using Twist Library Preparation EF Kit Twist Library Preparation EF Kit1, 2.0, and Twist Universal Adapter System-TruSeq Compatible, 96 Samples Plate A-D, enrichment using Twist Custom Panel, design name: Cancer_PRS_HUGV6; Twist Design ID: TE-96674869 (Twist Bioscience, South San Francisco, CA, USA), and sample identification using the Nimagen RC-PCR assay. Sequencing was conducted on a NextSeq500/550 Mid Output v2.5 kit (Illumina, San Diego, CA, USA; Sequencer: Illumina NextSeq550 (Illumina, San Diego, CA, USA). Mean coverage was at least 300×, and all target regions were covered 20×. Single nucleotide variants (SNVs) as well as copy number variants (CNVs) were detected within this setting. This analysis can be established in all labs familiar with NGS settings.

Variant Classification and Panel Sequencing
Analysis of the raw data was performed using the software Varfeed (Limbus, Rostock, Germany), and the variants (SNVs and CNVs) were annotated using the software Varvis (Limbus, Rostock, Germany). All variants were described in regard to GRCh37 (NM_000492.4) and classified according to the latest ACMG criteria [39]. The databases ClinVar [40], HGMD [41], and HerediCare [42] were used for classification based on the following considerations: gene and variant attributes, frequency in the general population (gnomAD [43]), (assumed) effect on protein function, in silico prediction tools (mainly CADD [44], SpliceAI-lookup [45]), conservation, and phenotype.

Targeted Sequencing for PRS Calculation
Raw reads were quality checked using fastqc [46], and remaining adapter sequences and bad quality data were removed using trimmomatic [47]. Processed data was aligned to hg19 using minimap2 [48], visual duplicates were marked with samtools [49]. Haplotypes for the BCAC-313 PRS model were called using freebayes, and positions with a coverage of <20× or conflicting haplotype signal were imputed using twice the allele frequency described in the BCAC-313 model [50]. The resulting haplotypes were used as an input to calculate the normalized z-score via the CanRisk API [29]. Resulting z-scores were combined with the related BOADICEA file to calculate the 5-year, 10-year, and lifetime risk for each participant.

Statistical Testing
The PRS values of the study cohort were tested for normal distribution using the Shapiro-Wilk test from the scipy package [51].
The study cohort was compared to the BCAC-313 model [52] cohort using the mean (mu) and standard deviation (sigma) provided by CanRisk. PRS distribution of the BCAC-313 cohort was modelled using the numpy package, and the comparison between the cohorts was conducted using the individual t-test from the scipy package (Supplementary Figure S1).
To test the effect of the PRS value on the 10-year risk calculation, the study cohort was filtered for women under the age of 50 without any known PVs in the 11 core genes, the cohort was divided in to high and low PRS individuals by comparison to the mean PRS value described for the BCAC-313 model. The effect of including or excluding the PRS value in the 10-year BC risk was calculated using the Mann-Whitney U test from the scipy package (Supplementary Figure S2).

Results
We included 382 women in our cohort. The mean age at time of evaluation was 45 years with an age range between 18 and 69 years. A total of 233 women (60.9%) were aged 18-49 years and 149 (39.1%) were aged 50 and older. At the time of analysis, 48.7% of women (186) had an invasive breast tumour, 7.9% (30) had ovarian cancer (OC), and 0.5% (2) were affected by both BC and OC. The average age at diagnosis for women affected by BC was 46 years with a range between 27 and 67 years. For OC cases, the mean age of diagnosis was 54 with a range between 22 and 67 years.

Risk Calculations
The PRS (z-score) in our cohort is normally distributed with a mean of 0.45 (SD = 1.02, Shapiro-Wilk test p-value = 0.79) (Supplementary Figure S1).
Through the inclusion of the PRS, the threshold of 5% in 10-year risk calculations was observed to be either surpassed or subordinated in 13.6% of all individuals. Among them, 8.1% surpassed the 5% threshold, defining them as women at high risk of developing BC according to the GC-HBOC. The basis for establishing the threshold is that the 10-year risk of ≥5% is approximately double the value of a 50-year-old woman from the general population [8]. A total of 5.5% fell below the 5% threshold, indicating a relevantly lower risk ( Figure 2).

Risk Calculations
The PRS (z-score) in our cohort is normally distributed with a mean of 0.45 (SD = 1.02, Shapiro-Wilk test p-value = 0.79) (Supplementary Figure S1).
Through the inclusion of the PRS, the threshold of 5% in 10-year risk calculations was observed to be either surpassed or subordinated in 13.6% of all individuals. Among them, 8.1% surpassed the 5% threshold, defining them as women at high risk of developing BC according to the GC-HBOC. The basis for establishing the threshold is that the 10-year risk of ≥5% is approximately double the value of a 50-year-old woman from the general population [8]. A total of 5.5% fell below the 5% threshold, indicating a relevantly lower risk ( Figure 2). For 153 BC cases without any PVs, the inclusion of the PRS in 10-year risk calculations led to different risk stratification for 20.3%, with 11.1% surpassing the 5% threshold and 9.2% falling below. A total of 52.9% of cases exceeding the 5% threshold were under the age of 50.
No changes in risk stratification after including the PRS were observed in the 27 BC cases with PVs in high-risk genes (BRCA1, BRCA2, and PALB2) (Figure 3). For 153 BC cases without any PVs, the inclusion of the PRS in 10-year risk calculations led to different risk stratification for 20.3%, with 11.1% surpassing the 5% threshold and 9.2% falling below. A total of 52.9% of cases exceeding the 5% threshold were under the age of 50.
No changes in risk stratification after including the PRS were observed in the 27 BC cases with PVs in high-risk genes (BRCA1, BRCA2, and PALB2) (Figure 3). The effect of including or excluding the PRS value in the 10-year BC risk calculation showed that individuals with lower PRS values than the mean described for the BCAC-313 model exhibit a significantly lower 10-year risk after incorporating the PRS into the analysis (p-value = 0.0011). In contrast, women with PRS values above the mean PRS value do not show a significantly higher 10-year risk (p-value = 0.31). However, it is important to note that for a specific individual, a higher PRS can still translate to a higher 10-year risk (Supplementary Figure S2).

Change in Prevention Management
The GC-HBOC considers risk calculations of the following groups for inclusion in the intensified breast cancer surveillance program [8,54,55]: (1) Healthy women under the age of 50 with unremarkable predictive multigene panel diagnostic results; (2) Healthy women under the age of 50 with unremarkable targeted genetic testing in moderate-risk genes; (3) Ovarian cancer patients under the age of 50 with unremarkable multigene panel diagnostic results; (4) Relatives of index patients with a PV detected by the multigene panel diagnostic. We did not include this group in our analyses.
If individuals included in one of these groups have a 10-year risk of ≥5% for developing BC, they are eligible for intensified screening to timely detect a potential breast tumour.
A total of 30 participants in our cohort met the characteristics of the first group. By including the PRS in 10-year risk estimations, 16.7% of these participants either exceeded The effect of including or excluding the PRS value in the 10-year BC risk calculation showed that individuals with lower PRS values than the mean described for the BCAC-313 model exhibit a significantly lower 10-year risk after incorporating the PRS into the analysis (p-value = 0.0011). In contrast, women with PRS values above the mean PRS value do not show a significantly higher 10-year risk (p-value = 0.31). However, it is important to note that for a specific individual, a higher PRS can still translate to a higher 10-year risk (Supplementary Figure S2).

Change in Prevention Management
The GC-HBOC considers risk calculations of the following groups for inclusion in the intensified breast cancer surveillance program [8,54,55]: (1) Healthy women under the age of 50 with unremarkable predictive multigene panel diagnostic results; (2) Healthy women under the age of 50 with unremarkable targeted genetic testing in moderate-risk genes; (3) Ovarian cancer patients under the age of 50 with unremarkable multigene panel diagnostic results; (4) Relatives of index patients with a PV detected by the multigene panel diagnostic. We did not include this group in our analyses.
If individuals included in one of these groups have a 10-year risk of ≥5% for developing BC, they are eligible for intensified screening to timely detect a potential breast tumour.
A total of 30 participants in our cohort met the characteristics of the first group. By including the PRS in 10-year risk estimations, 16.7% of these participants either exceeded or fell below the 5% threshold, indicating a change in clinical management. There are 49 individuals that can be attributed to the second group with relevant changes in 10-year risk predictions observed in 10.2% of cases. For the four women attributed to the third group, there were no changes. Overall, 12.0% of participants that can be assigned to one of the groups would require a change in prevention management (Figure 4). group, there were no changes. Overall, 12.0% of participants that can be assigned to one of the groups would require a change in prevention management (Figure 4).

Figure 4.
Scatter plot of the change in individual breast cancer 10-year risk after inclusion of the PRS for healthy women age < 50 with unremarkable predictive multigene panel diagnostic results, healthy women age < 50 with unremarkable targeted genetic testing in moderate-risk genes, and ovarian cancer patients age < 50 with unremarkable multigene panel diagnostic results. Individuals surpassing the 5%-risk threshold and consequent changes in clinical management are shown as blue dots. Individuals falling below the 5%-risk threshold and consequent changes in clinical management are shown as yellow dots; BC = breast cancer, and PRS = polygenic risk score.

Discussion
In this study, we demonstrated the impact of the 313-variant-based PRS on BOADI-CEA-based BC risk calculations in a cohort of 382 women, resulting in changes of BC risk stratification in 13.6% of all participants. These variations in risk assessment might have significant implications, e.g., for making informed decisions regarding preventive surgeries.
The majority of BC patients without a PV who have exceeded the 5% threshold after including the PRS in 10-year risk calculations are younger than 50 years (52.9%), indicating that this particular group would benefit from a more precise risk assessment due to the incorporation of the PRS. The association between PRS and BC risk decreasing with age was shown by Mavaddat et al. [13]. This is the first study conducted on the clinical application of the PRS in Germany. In a Dutch study, the impact of incorporating the PRS into risk calculations for 1331 non-BRCA1/2 carriers was investigated regarding screening procedures aligned with Dutch . Scatter plot of the change in individual breast cancer 10-year risk after inclusion of the PRS for healthy women age < 50 with unremarkable predictive multigene panel diagnostic results, healthy women age < 50 with unremarkable targeted genetic testing in moderate-risk genes, and ovarian cancer patients age < 50 with unremarkable multigene panel diagnostic results. Individuals surpassing the 5%-risk threshold and consequent changes in clinical management are shown as blue dots. Individuals falling below the 5%-risk threshold and consequent changes in clinical management are shown as yellow dots; BC = breast cancer, and PRS = polygenic risk score.

Discussion
In this study, we demonstrated the impact of the 313-variant-based PRS on BOADICEAbased BC risk calculations in a cohort of 382 women, resulting in changes of BC risk stratification in 13.6% of all participants. These variations in risk assessment might have significant implications, e.g., for making informed decisions regarding preventive surgeries.
The majority of BC patients without a PV who have exceeded the 5% threshold after including the PRS in 10-year risk calculations are younger than 50 years (52.9%), indicating that this particular group would benefit from a more precise risk assessment due to the incorporation of the PRS. The association between PRS and BC risk decreasing with age was shown by Mavaddat et al. [13]. This is the first study conducted on the clinical application of the PRS in Germany. In a Dutch study, the impact of incorporating the PRS into risk calculations for 1331 non-BRCA1/2 carriers was investigated regarding screening procedures aligned with Dutch IKNL [56], UK NICE [57], and US NCCN [58] BC screening guidelines. The results revealed clinically significant shifts in 32.4%, 36.0%, and 25.7% of individuals (with 30% BC lifetime risk cut-off levels based on the IKNL and NICE BC screening guidelines) [59].
In our cohort, the inclusion of the PRS in BC risk predictions resulted in clinically relevant shifts leading to changes in prevention recommendations established by the GC-HBOC in 12.0% of participants that can be assigned to one of the groups defined by the consortium. This presents a different patient stratification from current clinical practice, in which solely family history is included in risk prediction.
The current susceptibility SNPs account for around 44% of the familial relative risk associated with common low-risk variants [60]. Recent genome-wide association studies have discovered new BC susceptibility loci [61,62], leaving a more extensive PRS to be expected in the future. By incorporating an expanded PRS in BC risk prediction, more accurate risk stratification can be achieved, resulting in a higher percentage of women transitioning to different risk categories, and leading to improved screening measures.
Based on the considerable utility of the PRS, a potential application beyond HBOC families as part of general cancer screening should be discussed. A meta study was conducted to assess the cost effectiveness of implementing the PRS for three prevalent cancer types (prostate, colorectal, and breast cancer). Out of the ten studies analysed, eight demonstrated cost effectiveness in the utilization of the PRS [63]. However, further prospective studies with a larger cohort and case control studies are needed to quantify the effect of incorporation of the PRS in general screening among the general population programs for the health care system.
The strengths of this study include good representation of families that are seen in genetic counselling. All participants provided thorough family and personal medical history, ensuring comprehensive data collection. The utilization of BOADICEA as a wellvalidated and comprehensive risk model, allows for accurate risk predictions in a familial setting [22].

Limitations
We had to assume the European ancestry based on the participants' names. So far, the PRS is only validated for people of European ancestry [21], though some studies have indicated associations between a subset of the PRS and BC in Asian [64] and Latina/x/o [65] populations.

Conclusions
In summary, our findings support implementation of the PRS in genetic counselling, although it might present logistical challenges. By utilizing a reliable and comprehensive risk prediction model such as BOADICEA, pedigree-based family history, individual PRS, and molecular genetic analysis results can be combined easily, enabling the calculation of personalized BC risks, and therefore allowing for an improved clinical management in BC prevention.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cancers15153938/s1, Figure S1: The distribution of the PRS in our cohort and that of the UK cohort. The PRS is normally distributed; PRS = polygenic risk score, Figure  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data that support the findings of this study are available on request from the corresponding author.