What is the appropriate genetic testing criteria for breast cancer in the Chinese population?—Analysis of genetic and clinical features from a single cancer center database

Abstract Background Genetic testing plays an important role in guiding screening, diagnosis, and precision treatment of breast cancer (BC). However, the appropriate genetic testing criteria remain controversial. The current study aims to facilitate the development of suitable strategies by analyzing the germline mutational profiles and clinicopathological features of large‐scale Chinese BC patients. Methods BC patients who had undergone genetic testing at the Sun Yat‐sen University Cancer Center (SYSUCC) from September 2014 to March 2022 were retrospectively reviewed. Different screening criteria were applied and compared in the population cohort. Results A total of 1035 BC patients were enrolled, 237 pathogenic or likely pathogenic variants (P/LPV) were identified in 235 patients, including 41 out of 203 (19.6%) patients tested only for BRCA1/2 genes, and 194 out of 832 (23.3%) received 21 genes panel testing. Among the 235 P/LPV carriers, 222 (94.5%) met the NCCN high‐risk criteria, and 13 (5.5%) did not. While using Desai's criteria of testing, all females diagnosed with BC by 60 years and NCCN criteria for older patients, 234 (99.6%) met the high‐risk standard, and only one did not. The 21 genes panel testing identified 4.9% of non‐BRCA P/LPVs and a significantly high rate of variants of uncertain significance (VUSs) (33.9%). The most common non‐BRCA P/LPVs were PALB2 (11, 1.3%), TP53 (10, 1.2%), PTEN (3, 0.4%), CHEK2 (3, 0.4%), ATM (3, 0.4%), BARD1 (3, 0.4%), and RAD51C (2, 0.2%). Compared with BRCA1/2 P/LPVs, non‐BRCA P/LPVs showed a significantly low incidence of NCCN criteria listed family history, second primary cancer, and different molecular subtypes. Conclusions Desai's criteria might be a more appropriate genetic testing strategy for Chinese BC patients. Panel testing could identify more non‐BRCA P/LPVs than BRCA1/2 testing alone. Compared with BRCA1/2 P/LPVs, non‐BRCA P/LPVs exhibited different personal and family histories of cancer and molecular subtype distributions. The optimal genetic testing strategy for BC still needs to be investigated with larger continuous population studies.


| INTRODUCTION
Breast cancer (BC) has become the most common cancer globally. 1 China has the largest number of BC cases, which accounts for 18.4% (416,000) of the global ones. 2 Approximately 15%-20% of BC have a family history, and 5%-10% are considered to have hereditary breast cancer (HBC). 3 Due to the increasing influence of genetic factors on BC surveillance, prevention, and treatment decision, genetic testing is rapidly expanding in clinical practice. 4 Besides the most common BRCA1/2 genes, HBCassociated genes are emerging, including other high to moderate penetrance genes such as PALB2, TP53, PTEN, STK11, CDH1, CHEK2, ATM, and a variety of low penetrance genes. 5,6 However, the criteria of germline testing for BC remain controversial. The National Comprehensive Cancer Network (NCCN) criteria recommend genetic testing only for high-risk patients, which may miss half of the cases who do not meet this standard. 7,8 Other guidelines such as the American Society of Breast Surgeons (ASBrS) propose genetic testing for all BC patients, which would detect more variants at the cost of testing a large number of patients. [9][10][11] Especially in the population unfit for the NCCN criteria, the frequency of variants in high-risk BC genes is only 0.8%. 8 Therefore, this universal testing strategy will undoubtedly increase the cost and burden of genetic testing. Another important issue is which genes should be detected. Compared to testing for BRCA1/2 alone, multigene panel testing can improve the detection rate of HBC. 12 However, a diversity of gene panels were applied in different studies, varying from 6 to more than 100 genes. 13,14 The expanded panels contain many genes with low penetrance or even unrelated to BC, which will lead to a series of problems, including excessive patient anxiety, difficulty in variants interpretation, and unnecessary screening and prevention strategies. 8,15 As a heterogeneous group of diseases, different subtypes of BC might have different genetic backgrounds and characteristics. 16 Current studies mainly focus on HER2negative patients, especially triple-negative breast cancer (TNBC), while the genetic profile of HER2-positive BC is unclear. 17 Moreover, there are significant racial differences in HBC. 18 As the largest BC country, little genetic data concerning BC in the Chinese population is disclosed. Our current study aims to analyze the clinicopathological features and genetic data in a large cohort of Chinese patients diagnosed with HBC, which will facilitate making suitable criteria for germline testing of HBC.

| Study population
BC patients received germline counseling and testing at the Sun Yat-sen University Cancer Center (SYSUCC) from September 2014 to March 2022 were retrospectively analyzed via electronic medical record review. Variables obtained included age of cancer diagnosis, personal and family history of cancer, histopathological characteristics, and genetic testing profiles. The criteria for high-risk BC are as follows: (1) diagnosed with BC at the age ≤45 years; (2) TNBC diagnosed at the age ≤60 years; (3) multiple primary BCs; (4) male BC; (5) ≥1 close blood relative with BRCA-related cancer, including BC, ovarian cancer (OC), pancreatic cancer (PaC), and prostate cancer (PrC). 19 All patients signed informed consent for genetic testing. The study was approved by the Ethical Committee of SYSUCC and the Ministry of Science and Technology for human genetic resource collection.

| Genetic testing
Ethylenediaminetetraacetic acid (EDTA)-anticoagulated peripheral blood samples from all patients were collected.

| Genetic variant classification and analysis
Variants were named based on the rules suggested by the Human Genome Variation Society (HGVS) (http:// varno men.hgvs.org/) according to the criteria developed by the International Agency for Research on Cancer (IARC) and the American College of Medical Genetics and Genomics (ACMG). 20,21 The detected genetic variants were classified into five categories: benign variant (class I), likely benign variant (class II), variant of uncertain significance (VUS, class III), likely pathogenic variant (LPV, class IV), and pathogenic variant (PV, class V). Several databases were used to identify and classify the pathogenicity of genetic variants, such as Clin Var (https://www.clini calge nome. org/data-shari ng/clinv ar/), Leiden Open Variation Database (LOVD) (https://github.com/LOVDnl), BRCA Exchange (https://brcae xchan ge.org/), and Human Gene Mutation Database (HGMD) (http://www.hgmd. cf.ac.uk/ac/index.php). Pathogenic and likely pathogenic variants (P/LPVs) were defined together as deleterious variants for analysis.

| Statistical analysis
The genetic and clinicopathologic characteristics of the enrolled patients were summarized by descriptive statistics. T-tests were used for continuous variables. Chi-square tests or Fisher's exact tests were used to compare the categorical variables between P/LPV carriers and non-P/LPV carriers, and between subgroups of different P/LPV carriers. All p values were two-sided, and p values <0.05 was considered statistically significant. Statistical analysis was performed using IBM SPSS Statistics (V25; SPSS).

| Clinicopathological characteristics of the study population
The flowchart of screening and genetic testing for patients was shown in Figure 1. After reviewing 1616 participants, 1035 pathologically confirmed BC patients were finally enrolled in the analysis. The clinicopathologic characteristics of enrolled patients were summarized in Table 1 including 70 with secondary primary and 4 with third primary cancers.
Notably, a number of novel germline P/LPVs which has not reported in public databases (ClinVar, LOVD) and in the literature before were found and highlighted in the Tables S3-S5, including seven BRCA1 variants, eighteen BRCA2 variants, and nine non-BRCA variants. Rare P/ LPVs were also highlighted. In addition, 352 VUSs were detected in 289 patients, including 7 (3.4%) patients who had BRCA1/2 testing alone and 282 (33.9%) had panel testing. The most common genes had VUSs were ATM (n = 44), BRCA2 (n = 42), and MUTYH (n = 34) (Figure 3). All VUSs were summarized in Table S6.

| Comparison of different criteria for HBC
We applied different screening criteria available to our population cohort and compared the number of patients who met the criteria and had P/LPV (Table 4). At last, we found using Desai's criteria of testing all females diagnosed with BC by the age of 60 years and NCCN criteria for older patients could find most P/LPVs by testing the least number of patients.

| DISCUSSION
The current study demonstrated the genetic and clinicopathological characteristics in a large cohort of Chinese HBC patients. Among the screened 1035 patients, 906 (87.5%) met the NCCN high-risk criteria, and 235 were identified to carry at least one P/LPV in 15 BC susceptibility genes, with an overall P/LPVs rate of 22.7%, and 24.5% in the high-risk population. Similar frequencies of P/LPVs have been reported in other studies of the Chinese population with high-risk characteristics. 22,23 Even though, P/ LPVs were detected for 13 out of 129 (10.1%) patients who did not meet the NCCN criteria for testing, suggesting NCCN criteria will miss a significant number of patients with HBC in the Chinese population.
The high missing rate of NCCN criteria has been demonstrated in studies of other ethnicities. 8,24 This is mainly due to the restrictive criteria for testing. On the other hand, universal genetic testing will create many other challenges such as high costs and genetic testing burdens. Given the limitations of these two genetic testing strategies, several other criteria have been proposed. The main difference among these criteria focused on the appropriate screening age. The Mayo Clinic hybrid approach by Yadav et al. reported that testing all female BC by the age of 65 years and using NCCN criteria for older patients could miss fewer variants than with NCCN criteria alone and spared 21% of patients for testing compared with universal screening. 24 A subsequent report by Desai et al. demonstrated that lowering the age from 65 to 60 years maintained the detection sensitivity of Yadav's criteria >90% while sparing testing for an additional 10% of the population. 25 Boddicker et al. showed that the frequency of BRCA1/2 or PALB2 variants in TNBC older than 65 years was 3.0%, thus supporting genetic testing for TNBC at any age. 26 When we applied these criteria to our patients, we found using Desai's criteria of testing all females diagnosed with BC by the age of 60 years and NCCN criteria for older patients could find most P/LPVs by testing the least number of patients. However, due to the highly selective patients in the current study, further studies enrolling larger-scale consecutive populations are needed to establish the appropriate criteria for Chinses patients.
Nowadays, multigene panel testing is increasingly used for HBC screening. 14 In our cohort, patients who were diagnosed before 2017 received BRCA1/2 testing alone, after then most patients underwent the 21 genes panel testing. Compared with BRCA1/2 testing alone, panel testing identified 4.9% of non-BRCA P/LPVs and a significantly high rate of VUSs (33.9%). The incidence of non-BRCA P/LPVs varied dramatically in different studies, ranging from 1% to 12%. Moreover, an even large difference of VUSs was reported, ranging from 0.6 to 88%. 14,27,28 The main reason probably due to the diversity of gene panels applied in different studies. However, more genes do not mean better. Many genes included in the panels are low-risk or even unrelated to BC. Therefore, it will lead to the detection of more P/LPVs and VUSs without clinical significance. 29 Besides BRCA1/2, 19 non-BRCA genes were included in our panel. And 13 non-BRCA variants were found, including PALB2   Bold P values indicate significant differences between the two variables.

T A B L E 2 (Continued)
T A B L E 3 P/LPV distribution in different molecular subgroups based on HR and HER2 status. 0.1%). PALB2 is the most prominent non-BRCA gene, and testing PALB2 is cost-effective. 15 PALB2 works as the partner and localizer of BRCA2, and associates with an overall increased risk of BC of five to nine-fold. 30 Two large cohort studies demonstrated a greater association between PALB2 P/LPVs and ER-negative BC, whereas our data found PALB2 variant was more common in HR-positive patients, accounting for 9 of 11 cases. 26 variants used to classify VUS are pathogenic mutations. 36 Therefore, these VUS should be investigated further in a large number of patients. A case of CDH1 PV was found in one female patient diagnosed with ILC at 40 years. She had a strong family history of gastric cancer, including her grandmother, father, and aunt. Notably, we found non-BRCA P/LPVs showed a low incidence of NCCN criteria listed family history of BC, OC, PaC, and PrC, as well as a low frequency of secondary primary cancer of OC, PaC, and PrC, suggesting testing criteria for non-BRCA or panel testing should be not only based on these personal and family histories. In addition, different molecular subtypes exhibited different variant profiles. Consistent with other reports, TNBC had the highest proportion of BRCA1 variants, and HR + HER2-BC had higher rates of BRAC2 and PALB2 variants. While HER2-positive BC patients had the highest non-BRCA P/LPVs, especially TP5 variants. Therefore, future criteria probably should take into consideration of molecular factors, but not just focus on HER2-negative or TNBC patients.
The main limitation of our current study is that it enrolled the highly selective high-risk patients. Larger scale studies in consecutive populations should be initiated to further explore optimal genetic testing strategies for BC.

| CONCLUSION
Our current study suggested Desai's criteria of testing all females diagnosed with BC by the age of 60 years and using NCCN criteria for older patients, might be a more suitable genetic testing criteria for Chinese BC patients. Panel testing could identify more non-BRCA P/LPVs than BRCA1/2 testing alone. Non-BRCA P/LPVs showed different personal and family histories of cancer and molecular subtype distributions compared with BRCA1/2 P/LPVs. The optimal genetic testing strategy for BC still needs to be investigated by larger-scale consecutive population studies.