Germline Cancer Susceptibility Gene Testing in Unselected Patients With Colorectal Adenocarcinoma: A Multicenter Prospective Study

Germline Cancer Susceptibility Gene Testing in Unselected Patients With Colorectal Adenocarcinoma: A Multicenter Prospective Study Pedro L. S. Uson Jr,* Douglas Riegert-Johnson,‡,§,k Lisa Boardman, John Kisiel, Luke Mountjoy,* Neej Patel, Blanca Lizaola-Mayo, Mitesh J. Borad,*,k Daniel Ahn,* Mohamad B. Sonbol,* Jeremy Jones,kk Jonathan A. Leighton, Suryakanth Gurudu, Harminder Singh,** Margaret Klint, Katie L. Kunze, Michael A. Golafshar, Ed D. Esplin, Robert L. Nussbaum, A. Keith Stewart,*,§,k Tanio S. Bekaii-Saab,* and Niloy Jewel Samadder§,k,#


Incremental Genetic Findings
Germline cancer susceptibility gene testing in unselected patients with colorectal adenocarcinoma

BACKGROUND & AIMS:
Hereditary factors play a role in the development of colorectal cancer (CRC). Identification of germline predisposition can have implications on treatment and cancer prevention. This study aimed to determine the prevalence of pathogenic germline variants (PGVs) in CRC patients using a universal testing approach, association with clinical outcomes, and the uptake of family variant testing.

METHODS:
We performed a prospective multisite study of germline sequencing using a more than 80-gene next-generation sequencing platform among CRC patients (not selected for age or family history) receiving care at Mayo Clinic Cancer Centers between April 1, 2018, and March 31, 2020.
H ereditary factors play a key role in the risk of developing several cancers including colorectal cancer (CRC). 1,2 Identification of genes related to the development or inheritance of cancer can have important implications in the selection of targeted treatments and in the screening of affected populations. 1,2 The prevalence of germline genomic alterations related to the development of colorectal carcinoma has been assessed in multiple studies with estimates of at least 10% of CRC patients. However, many of these studies have focused on registry populations, genetic testing company databases, high-risk cancer clinics, and the use of relatively small gene panels, which may have biased results or underestimated the true prevalence. [3][4][5][6][7] The emergence of targeted therapies, including the use of poly-adenosine diphosphate ribose polymerase (PARP) inhibitors for use in several tumors with germline homologous recombination deficiency (HRD) and the superior outcomes with anti-programmed cell death protein 1 pembrolizumab compared with chemotherapy in a randomized trial in metastatic microsatellite instability high CRC 2,[8][9][10][11] has raised the question of whether more broad-based universal germline genetic testing has clinical implications in solid tumors.
We previously performed a prospective multisite study of germline genetic alterations among solid-tumor cancer patients at the Mayo Clinic Cancer Center (Rochester, MN; Jacksonville, FL; and Phoenix, AZ) and a community oncology practice (Eau Claire, WI) between April 1, 2018, and March 31, 2020. Patients were unselected for cancer type, stage of disease, family history of cancer, ethnicity, or age, and underwent germline testing using a more than 80gene next-generation sequencing (NGS) platform. In this report, we evaluated only the patients with colorectal cancer recruited from the 3 Mayo Clinic Cancer Center sites (Rochester, MN; Jacksonville, FL; and Phoenix, AZ). We describe their clinical characteristics, prevalence, and the incremental yield of pathogenic genomic alterations, and the impact on clinical outcomes.

Patient Selection
From April 1, 2018, through March 31, 2020, there were 2984 adult (age, 18-85 y) patients with a new or active cancer diagnosis, confirmed pathologic diagnosis of carcinoma, seen in medical oncology, radiation oncology, dermatology, or surgical oncology clinics at any of the 3 Mayo Clinic destination Cancer Centers in Phoenix, AZ, Jacksonville, FL, or Rochester, MN, and English-speaking enrolled (Mayo Clinic Interrogating Cancer Etiology using Proactive Genetic Testing-Interrogating Cancer using Proactive genetic Testing [INTERCEPT] Program). Exclusion criteria were patients with hematologic malignancies and those undergoing surveillance after curative cancer. Patients were recruited using central lists of daily oncology clinic visits by research coordinators at each site. Patients were not selected based on stage of disease, family history of cancer, ethnicity, age at diagnosis, multifocal tumors, or personal history of multiple malignancies. Details of the INTERCEPT study and full cohort have been described previously. 12 The current study describes the cohort of 361 patients with a diagnosis of colorectal adenocarcinoma (2 cases were excluded from the original INTERCEPT cohort owing to a lack of complete medical records allowing for comprehensive in-house review of the pathologic diagnosis). Enrollment for each major cancer site was capped at approximately 300 to allow for sufficiently powered comparisons across cancer sites for the primary analysis.
All patients viewed a standardized pretest education video and were offered additional pretest genetic counseling if desired. Germline sequencing using an NGS panel of 83 genes (84 genes as of July 2019) on the Invitae Multi-Cancer panel was offered at no cost (Appendix 1). All test results, along with available pedigree information, were reviewed by a certified genetic counselor, disclosed to the patient, and those with pathogenic germline variants (PGVs) were invited for genetic counseling. Patients with a variant of unknown significance (VUS) or negative test results were informed of their findings via electronic medical record portal message, telephone, or certified letter. Genetic counselors provided general cancer prevention screening recommendations based on family history if genetic test results were nondiagnostic.
Clinical, demographic, family history data, and pathologic information were collected on all patients from either medical records or self-administered electronic questionnaires. Race/ethnicity (ancestry) was determined by patient self-report. Family history information was collected using an electronic pedigree tool (Cancer-Gene Connect, Invitae Labs, San Francisco, CA).
This study was approved by the Mayo Clinic Institutional Review Board (18-000326). All patients provided written informed consent. Data were de-identified except to investigators of the study.

Comparison of Guideline-Based Testing
The National Comprehensive Cancer Network (NCCN), National Society of Genetic Counselors, and American College of Medical Genetics guidelines (2018 and 2020) were used to determine whether genetic testing was indicated for a patient. 16 A PGV was considered incremental if it was detected based on the universal testing performed in this study and would not have been identified based on genetic testing/referral criteria of the 2018/2020 NCCN/National Society of Genetic Counselors/American College of Medical Genetics guidelines or was a gene outside those recommended on a guideline-based panel, such as the Invitae colorectal cancer guideline-based panel (APC, AXIN2, BMPR1A, CHEK2, EPCAM, GREM1, MLH1, MSH2, MSH3, MSH6, MUTYH, NTHL1, PMS2, POLD1, POLE, PTEN, RPS20, SMAD4, STK11, and TP53).

Family Variant Testing
Cascade family variant testing (FVT) at no cost was offered to all blood relatives of affected participants with PGV within a 90-day window of their finalized test result report. At the time of genetic counseling, patients were informed of the FVT program offered by Invitae and assisted in communicating this information to their relatives using a standardized template letter and an online video describing the risks and benefits of genetic testing.

Statistical Analysis
Demographic and clinical characteristics of the cohort are presented using descriptive statistics. The prevalence of PGV and VUS were reported in the cohort. Categoric variables were compared using the Pearson chi-square test. Rates of incremental findings were compared between subgroups by the Pearson chi-square test, and proportions of germline findings were compared by stage of disease using the chi-squared test. P values less than .05 were considered statistically significant. All statistical tests were 2-sided. Univariate logistic regression models were used to predict pathogenic germline mutation. Recurrence-free survival was considered the time from surgery to first disease recurrence and overall survival was considered the time from diagnosis to death or last date of follow-up evaluation.

Cohort Characteristics
The distribution of sex, age, comorbidities, stage, and location of disease in these patients is shown in

Findings
This was a multicenter prospective study of universal germline genetic testing in CRC patients not selected for age at diagnosis, cancer stage, or family history of cancer. Among 361 patients with CRC, 15.5% had pathogenic mutations identified, of which 25% would not have been detected using standard clinical criteria for genetics evaluation. Eleven percent of patients had modifications in their treatment based on genetic findings, however, there was no difference in overall survival between those with or without a pathogenic germline mutation Implications for patient care Our data support the use of broad universal multigene panel testing in CRC patients over a targeted approach based on clinical guidelines. One in 10 patients will have modifications to their clinical management plan based on the genetic findings. Cascade family variant testing is underused and can be improved to increase the potential for cancer prevention in the population.

Variant Detection and Clinical Outcomes
Of the 361 patients undergoing germline analysis, 56 patients (15.5%) harbored 62 pathogenic/likely pathogenic germline variants, of whom 5 patients had more than 1 PGV detected ( Figure 1). PGVs could be stratified into those with high (n ¼ 21), moderate (n ¼ 21), or low (n ¼ 7) penetrance, and 7 were carriers of variants associated with recessive syndromes (Supplementary  Tables 2 and 8   PALB2, RAD50, RAD51C, RAD51D, and WRN). More than half (54%) of Lynch syndrome cases were observed in patients with right-sided colon cancer. Figure 2 shows the distribution of PGVs by gene and tumor site. Table 2 shows the distribution of sex, age, ancestry, location, and stage stratified by mutation type. A higher proportion of patients with PGVs had no history of smoking, diabetes, or hypertension, and had a lower body mass index (<30). The rate of PGVs was higher in patients with rectal cancer than those with colon cancer (P ¼ .008  Figure 3 shows the specific type of PGV (deletion, insertion, duplication, or missense) listed by gene and tumor location. The majority of PGVs were missense mutations (67%), followed by deletion (24.2%), duplication (6.5%), and insertion (1.6%).
Overall survival (OS) was similar for those with and without a PGV (P ¼ .2). Similarly, no differences in OS were noted in patients by tumor location (colon vs rectum) or presence of PGVs in Lynch or HRD genes. Early stages of disease (stages 1-3 vs 4) were associated with improved survival rates (P < .001). Kaplan-Meier plots for overall survival are shown in Supplementary  Figure 1.
Sex, stage of cancer, family history of cancer in a firstdegree relative, and CEA level were not associated with a higher likelihood of having a PGV ( Table 3).

Application of Clinical Genetic Referral Criteria
Thirty-four cases had incremental clinically actionable findings that would not have been detected by phenotype or family history-based testing criteria using the 2018 NCCN guidelines (Supplementary Tables 3 and  4), or were outside of the genes recommended on a 20gene panel of CRC-associated genes (Supplementary Table 5). 10 This represents 9.4% of the 361 cases overall and 60.7% of the 56 patients with PGV. Of the 34 patients with incremental findings, 8 (23.5%), 15 (44.1%), 5 (14.7%), and 6 (17.6%) patients carried high-, moderate-, or low-penetrance variants, and recessive alleles, respectively. Incremental mutations were found more often in patients with colorectal cancer diagnosed at an older age (50 vs <50 y; P < .0017) (Supplementary Table 6). Use of the 2020 NCCN guidelines resulted in the same 34 patients with incremental findings.

Clinical Implications of Pathogenic Germline Variants
Of the 56 patients found to have a PGV, 10.7% had clinically actionable management and treatment changes after detection of the PGV (Supplementary Table 7). These can be broadly categorized as need for surgery (n ¼ 2) and targeted therapy (n ¼ 4). One patient with a CHEK2 mutation underwent surgical management involving hysterectomy and bilateral salpingooophorectomy, which was not consistent with guideline management, however, because of extenuating circumstances specific to the patient's tumor and treatment plan, it was performed after multidisciplinary discussion at the local tumor board.

Family Variant Cascade Testing
No-cost FVT was offered to all blood relatives of affected participants. Only 9 (16%) patients with PGVs had family members undergo FVT within a 3-month window of their test result. The median number of family members tested was 2.0, with a range of 1 to 4. Almost half (46.4%) of family members tested were relatives of patients with stage IV disease.

Discussion
In this multisite prospective study of 361 unselected CRC patients, universal germline genetic testing found that approximately 1 in 6 CRC patients (15.5%) harbor a PGV of which more than 50% would not have been detected using standard practice guidelines or a guideline-specific gene panel. No differences in diseasespecific outcomes were observed among patients with or without a PGV. Nearly 10% of patients had modifications in their treatment based on the findings, and cascade testing was underused even when cost was not a barrier.
The prevalence of pathogenic germline mutations in CRC has ranged between 5% and 18% in the literature. 1,3,5,17 The difference in the prevalence of germline PGV could be owing to many factors. First, the median age of patients can directly influence the percentage of positive results; a retrospective analysis of 430 patients younger than age 50 years resulted in a PGV rate of 18%. 17 Similarly, in our study, 22% of patients younger than age 50 years were detected with a PGV. Second, the number of genes evaluated in the panel can omit genes that are not associated commonly with colon malignancy/polyposis. Germline testing using a panel of 25 genes was evaluated in 1058 unselected patients with CRC and found nearly 10% had a PGV. 3 Finally, prior retrospective studies can suffer from selection bias, considering that germline testing would be more common in patients with a family history of cancer, younger age, and recruitment bias from a cancer genetics clinic or research registry. These are inherent strengths of our study, including prospective recruitment from multiple clinics and cancer centers, and additionally the use of a broader NGS panel.
Equally important to the discovery of a PGV in a CRC patient is the potential to share these findings with their relatives. Targeted testing may allow for earlier disease detection and prevention in those who are positive for the familial mutation. Family cascade testing remained low (16%) even though it was available at no cost. This result is consistent with several other studies that have found low adherence to cascade family testing. In a study conducted in Singapore in 183 patients, the uptake of free cascade testing was 21.6%. 18 A low rate of family testing also was observed in 2 studies examining patients with hereditary gynecologic cancer and Lynch syndrome. 19,20 Although the majority (87%) of patients were comfortable sharing genetic test results with close relatives, only 40% of first-degree relatives underwent genetic testing. 19 An online approach to low-cost cascade testing also was impacted by low rates of participation. 21 There likely are several factors associated with the low uptake outside of financial barriers, including poor understanding of the importance of testing or difficulties interpreting the reports, distance between relatives, family dynamics including communication barriers, shortage of genetics professions and long wait times, variable insurance coverage for genetic counseling, other social/economic restraints, fear of discrimination, and fear of eventual procedures related to the findings. The reliance on patient-driven cascade testing as a result of privacy laws with little support and follow-up evaluation from providers likely is a major barrier in the United States. To help support and standardize the disclosure of results to relatives, clinician-provided education material such as letters, brochures, websites, and videos can be considered. 19 In addition, providing patients with an annotated copy of their family tree indicating which members should receive genetic testing information may help ensure this information is shared with those who may benefit. 20 Finally, options that empower the clinician or testing laboratory to reach out to relatives directly may be fruitful, 21 however, this is contrary to current US privacy laws. The future also may include artificial intelligence-powered, Health Insurance Portability and Accountability Act-compliant chatbots to screen for heritable CRC syndromes and engage family members to increase cascade testing uptake. A total of 11 (3.1%) patients had PGV in mismatch repair genes (MMR), confirming a molecular diagnosis of Lynch syndrome. Most of these patients presented with right-sided colon cancer (54%). The prevalence of Lynch syndrome in Western and Eastern populations ranges between 1% and 5% and is similar to our study. 22,23 Interestingly, 6.4% of patients with CRC had a mutation in genes related to HRD, with a higher prevalence seen in patients with rectal cancer (70%). In our cohort, we were able to evaluate disease outcomes including overall survival based on the presence of PGV and more specifically the type of mutation (HRD-associated gene or Lynch syndrome). No differences in OS were observed between patients with or without a PGV, including in those with MMR or HRD-associated gene mutations. Some studies have found that patients with advanced CRC carrying germline mutations on HRD genes were associated with better disease outcomes or survival. 24,25 The interpretation of these results should be taken with caution because a longer follow-up period likely is required because the vast majority of our cohort did not experience disease recurrence or death.
The discovery of PGV in a CRC patient will have therapeutic implications. For example, the presence of a MMR may help guide treatment decision making for the use of programmed cell death protein 1/PDL1 inhibitors in those with metastatic CRC. 2 Pembrolizumab improved the response rate and progression-free survival compared with chemotherapy in metastatic microsatellite instability high CRC in first-line therapy, in a phase III trial, with fewer treatment-associated adverse events. 2 Moreover, HRD-associated gene defects as a biomarker may have therapeutic implications including the use of topoisomerase inhibitors and PARP inhibitors. 5,[26][27][28] There have been recent reports of metastatic CRC patients refractory to standard-line chemotherapy who harbor germline mutations in DNA damage response genes-including CHEK2 and ATM-showing preliminary clinical response to PARP inhibitors and platinum therapy. 26,29 Currently, multiple tumor agnostic trials including solid tumors with mutations in genes related to DNA damage response are being developed (NCT04276376, NCT04657068, and NCT04564027). In our cohort of 58 CRC patients with P/LP variants, 35 (60%) potentially were eligible for approved precision therapy and/or clinical treatment trials. We identified nearly 11% of CRC patients with PGV in whom clinical management and treatment changes were made based on the detected PGV, including 7% who were started on targeted therapies or clinical trials. This is consistent with a prior study from New York that reported discussion of initiation of targeted therapy in 18% of their advanced cancer patients with PGV. 12 In addition to phenotypic characteristics (which may be muted in some patients) and tumor testing results, universal genetic testing will allow us to ensure all cancer patients who qualify for targeted therapy are identified.
Multidisciplinary care between surgeons, oncologists, and geneticists is critical to ensure guideline-adherent modifications in care are recommended. This study had limitations including the demographic characteristics of patients seen at the multiple Mayo Clinic sites participating, which may not completely reflect those in other regions of the country or other countries. 30 Family history of cancer was obtained through patient self-report or review of medical records and was not complete. Because of the relatively short follow-up period, most patients are still alive, liming the utility of survival analysis to address outcomes related to PGV in the MMR and HRD genes and response to targeted therapies. The study was not able to track if blood relatives underwent cascade testing outside of Invitae Laboratories or if they underwent full panel testing at another laboratory separately. The interpretation of the study and applicability of the results in wider heterogeneous populations should be taken with caution. Furthermore, long-term follow-up evaluation has not been conducted and will be necessary to address the real impact of germline PGV in disease-specific outcomes, on cancer decision making, and morbidity of screening and other invasive procedures.

Conclusions
In this multisite prospective cohort study of patients diagnosed with CRC, unselected for family cancer history, we found that universal multigene panel testing was associated with an increased detection of clinically actionable, heritable mutations over that predicted based on clinical guidelines. Ten percent of patients with PGV had modifications in their treatment. Family variant testing remained low, although the financial barrier of such testing was not present in this study, suggesting the importance of nonfinancial barriers to this process. In addition to Lynch syndrome, more than 6% of patients had mutations in genes related to homologous recombination repair, particularly in rectal tumors, suggesting the possibility of additional therapeutic trials for those with advanced cancer. The application of broader germline testing in CRC patients is feasible and can add valuable insights into the development of personalized treatment strategies.

Supplementary Material
Note: To access the supplementary material accompanying this article, please click here.