Rate of hereditary cancer predisposition identified in eGT of patients with solid tumors

The patient cohort consisted of 17,523 patients with a broad range of solid tumors who received eGT. In comparison with the incidence rates reported by the National Cancer Institute Surveillance, Epidemiology, and End Results Program [39], our cohort was particularly enriched for pancreatic, ovarian, endometrial/cervical, CNS cancers, and sarcomas, while having a relatively lower proportion of lung, head/neck, thyroid, breast cancers, and melanomas (Additional file 4: Fig. S1). P/LP variants were identified in 16.7% (2930/17,523) (95%CI 16.2–17.3%) of patients overall, with 10.6% (1865/17,523) (95%CI 10.2–11.1%) having P/LP variants in a high- or moderate-penetrance gene with autosomal dominant inheritance (Fig. 1). In cancer types with > 1000 patients tested, ovarian cancer had the highest rate of patients with P/LP variants (24.2%), followed by pancreatic cancer (19.4%) and breast cancer (17.5%), with 18.1%, 13.7%, and 13% having P/LP variants in high/moderate-penetrance genes, respectively. In other cancer types represented by a smaller number of patients in our eGT cohort, the highest rates of germline P/LP variants were identified in gastrointestinal stromal tumors (30.6%), non-small cell lung cancer (19.6%), small bowel cancer (19.2%), esophagogastric cancer (17.9%), and mesotheliomas (17.2%), with 29.4%, 16%, 12.1%, 16.4%, and 11.6% having P/LP variants in high/moderate-penetrance genes, respectively (Fig. 1).

Fig. 1

Distribution of pathogenic/likely pathogenic variants identified in eGT of 17,523 cancer patients across genes and cancer types. The distribution of P/LP variants identified in patients with each cancer type, grouped based on gene penetrance and inheritance pattern, is presented. The number of patients in each category is in parentheses. Specific variants that show different penetrance or inheritance pattern from typical variants in the gene were plotted separately: APC^: APC p.Ile1307Lys; CHEK2^: CHEK2 p.Ile157Thr; FH^: FH p.Lys477dup; and VHL^: VHL p.Arg200Trp. Percentage of high/moderate-penetrance (%H/M pen) variants identified in each cancer type and overall percentage of patients with P/LP variants (%Overall) are presented with 95% confidence intervals in parentheses

Positive results include findings in genes that are known to be associated with the patient’s cancer type and those in genes that have not been associated with the patient’s current disease, which likely represent secondary findings. P/LP variants in genes that confer increased risk for the individual’s tumor type were also identified in patients with cancer types that are not frequently interrogated in traditional targeted GT models, such as 8.2% (6/73) of small bowel cancer patients having MLH1, MSH2, or PMS2 [40], 4.1% (6/145) of mesothelioma patients having BAP1 [41], 3.1% (3/96) of osteosarcoma patients having RB1 [42], and 2.5% (11/433) of soft tissue sarcoma patients having TP53 [43,44,45] P/LP variants.

A significant proportion of our cohort (1.2%) had one of the three BRCA1/BRCA2 Ashkenazi Jewish founder variants [46,47,48], due to the prevalence of individuals with Ashkenazi Jewish ancestry in our patient population (16% of patients receiving MSK-IMPACT [49]).

Diagnostic yield of eGT for breast, colorectal, ovarian, pancreatic, and prostate cancer

For patients with breast, colorectal, ovarian, pancreatic, and prostate cancer, GT is often pursued in a guideline-dependent manner, either by targeting a group of genes that are strongly associated with the particular cancer type or testing larger panels of hereditary cancer predisposition genes including those that are not known to increase the risk for the patient’s current disease. We assessed the rate of positive results (identification of P/LP variants) in each gene for these five common cancers that are most frequently interrogated for hereditary predisposition in the current practice and evaluated the rate of additional findings in genes that are not known to be associated with the patient’s cancer type (Fig. 2).

Fig. 2

Rate of positive results identified in each gene in patients with five common cancer types. Percentage of patients with breast cancer (n = 2243), colorectal cancer (n = 2060), ovarian cancer (n = 1122), pancreatic cancer (n = 1648), and prostate cancer (n = 2114) who had P/LP variants in genes that have been associated with each cancer type. Error bars represent 95% confidence intervals

In breast cancer patients (n = 2243), the overall yield of eGT was 17.5% (392/2243). BRCA1 and BRCA2 P/LP variants were identified in 2.9% (n = 51) and 2.3% (n = 65) of the patients, respectively, and accounted for 26.9% of all positive results in these patients. Three other commonly targeted genes, CHEK2, ATM, and PALB2 [29, 50,51,52], had a diagnostic yield of 2.5% (n = 56), 1.4% (n = 31), and 0.7% (n = 15), respectively. High-penetrance genes that implicate breast cancer risk and are often targeted in the presence of additional features in the patient’s personal and/or family history [29, 50,51,52] and had positive results in our cohort include NF1 with 0.2% (n = 5), TP53 with 0.2% (n = 5), and CDH1 with 0.09% (n = 2) yield. While these three genes added a minor increase in the diagnostic yield, all five patients with NF1 variants had features of neurofibromatosis type 1, both of the two patients with CDH1 variants had lobular breast carcinoma, and one of the two patients with TP53 variants had a history of sarcoma and breast cancer at 29 years of age. One patient with the TP53 variant had breast cancer at 44 years of age and did not meet the current TP53 GT criteria [43, 53]. Other genes with moderate, low, or uncertain penetrance that have been implicated in breast cancer [29, 50,51,52, 54,55,56], RECQL, RAD51B, ERCC3, MRE11A, RAD51D, BARD1, and NBN, had a yield of 0.3% (n = 7), 0.3% (n = 7), 0.3% (n = 6), 0.3% (n = 6), 0.2% (n = 4), 0.2% (n = 4), and 0.1% (n = 3), respectively.

In colorectal cancer patients (n = 2060), the overall yield of eGT was 15.3% (316/2060). The highest rate of positive results was in APC, with the low-penetrance p.Ile1307Lys variant identified in 1.8% (n = 38) and other APC variants in 0.2% (n = 5), followed by monoallelic MUTYH variants in 1.7% (n = 36), and Lynch syndrome-associated variants in MSH2, MLH1, MSH6, and PMS2 [30] identified in 1.2% (n = 25), 1.1% (n = 23), 0.8% (n = 16), and 0.6% (n = 13) of the patients, respectively. P/LP variants in other genes that have been associated with colorectal cancer [30, 57], CHEK2, NTHL1 (monoallelic variants), MSH3 (monoallelic variants), POLD1, BMPR1A, and SMAD4, were identified in 1.6% (n = 34), 0.3% (n = 7), 0.1% (n = 2), 0.05% (n = 1), 0.05% (n = 1), and 0.05% (n = 1) of the patients, respectively. Of note, the POLD1 carrier had hyper-mutated colon adenocarcinoma, the BMPR1A carrier had a hamartomatous polyp, and the SMAD4 carrier had a history of a juvenile polyp, consistent with the identified genes, although the patients with BMPR1A and SMAD4 variants do not meet the current GT criteria for the respective genes [30].

In ovarian cancer patients (n = 1122), the overall yield of eGT was 24.2% (272/1122). BRCA1 and BRCA2 P/LP variants were identified in 7.2% (n = 81) and 3.9% (n = 44) of the patients and accounted for 42% of all positive results in these patients. Other genes implicated in ovarian cancer [29, 58,59,60], BRIP1, RAD51D, PALB2, and RAD51C, had a yield of 1.1% (n = 12), 0.8% (n = 9), 0.3% (n = 3), and 0.09% (n = 1), respectively. MSH2, PMS2, and MSH6 variants were identified in 0.3% (n = 3), 0.5% (n = 5), and 0.09% (n = 1), with a total of 0.9% of ovarian cancer patients having Lynch syndrome-associated variants, and 78% (7/9) of them had endometrioid, clear cell, or mixed ovarian carcinoma/carcinosarcoma, whereas two had high-grade serous ovarian carcinoma [61, 62]. Microsatellite instability (MSI) and/or loss of the mutated protein’s expression by immunohistochemistry (IHC) in the tumors were detected in five patients, who were considered to meet Lynch syndrome GT criteria based on their MSI/mismatch repair-deficient tumor profiles [62], whereas four patients with MSH2 or PMS2 variants had microsatellite stable/indeterminate tumors with retained mismatch repair protein expression. Additionally, SMARCA4 variants were identified in three patients with small cell carcinoma of the ovary, hypercalcemic type, accounting for 0.3% of our ovarian cancer patient cohort.

In pancreatic cancer patients (n = 1648), the overall yield of eGT was 19.4% (319/1648). BRCA2, ATM, and BRCA1 [29] variants were identified in 4.2% (n = 70), 2.2% (n = 36), and 1.8% (n = 29) of the patients, respectively. PALB2 and CDKN2A [29] had a yield of 0.7% (n = 12) and 0.4% (n = 7), respectively. Variants in PMS2, MSH2, MSH6, and MLH1 [29] were identified in 0.2% (n = 4), 0.2% (n = 3), 0.2% (n = 3), and 0.1% (n = 2), respectively, with 0.7% of pancreatic cancer patients having Lynch syndrome-associated variants overall.

In prostate cancer patients (n = 2114), the overall yield of eGT was 15.9% (337/2114). BRCA2, CHEK2, ATM, BRCA1, PALB2, and HOXB13 [63] variants were identified in 3.5% (n = 75), 2.5% (n = 54), 1.4% (n = 30), 0.6% (n = 12), 0.5% (n = 11), and 0.3% (n = 7) of the patients, respectively. Additionally, MSH2, MSH6, and PMS2 [63] variants were identified in 0.4% (n = 8), 0.2% (n = 5), and 0.2% (n = 4), with 0.8% of prostate cancer patients having Lynch syndrome-associated variants overall.

Additional findings discovered in eGT

For individuals with breast, colorectal, ovarian, pancreatic, and prostate cancer, we next sought to characterize the additional P/LP variants in genes other than those that are associated with the patient’s current cancer type, as described above. Overall, 765 additional P/LP variants in genes not known to be associated with the patient’s current cancer type were identified in 8% (736/9187) of the patients with five common cancer types, with 0.3% (29/9187) having multiple such variants (Fig. 3). Additional findings were identified in 7% (156/2243) of breast, 6.8% (140/2060) of colorectal, 11.2% (125/1122) of ovarian, 10% (164/1648) of pancreatic, and 7.2% (151/2114) of prostate cancer patients. Additionally, 1.7% of breast, 1.5% of colorectal, 2.2% of ovarian, 1.4% of pancreatic, and 1.1% of prostate cancer patients had multiple P/LP variants identified in eGT, including those in genes that are associated with their cancer type.

Fig. 3

Pathogenic/likely pathogenic variants identified in genes not associated with the patient’s cancer type. Genes were grouped based on inheritance pattern, and autosomal dominant genes were further grouped based on penetrance as high, moderate, low, and uncertain. In three genes, only specific variants were targeted: HOXB13 p.Gly84Glu, MITF p.Glu318Lys, and YAP1 p.Arg331Trp. Certain variants were considered as having different penetrance or inheritance pattern from typical variants in the gene: APC p.Ile1307Lys and CHEK2 p.Ile157Thr as having uncertain penetrance; FH p.Lys477dup, VHL p.Arg200Trp, and EGFR loss-of-function variants as being AR. The percentage of carriers within each cancer type is presented in the upper panels. The number of P/LP variants identified in each gene and cancer type are presented in the lower panels

Overall, 3.3% (299/9187) of patients had an additional finding that indicated early or additional surveillance, and 0.2% (17/9187) had a finding that indicated prophylactic surgery recommendations to reduce future cancer risks for the patient and their carrier family members (Fig. 3, Additional file 2: Table S2). Monoallelic variants in AR genes conferring carrier status, which are not expected to increase disease risk but may have reproductive planning implications, were identified in 3% (278/9187) of the patients.

A total of 69 patients (0.8%) had a P/LP variant in a high-penetrance gene that is not associated with their cancer type (Table 1). We retrospectively reviewed the detailed clinical and family histories of these patients to assess whether they had any clinical features or history that was consistent with these findings and if they met the traditional GT criteria for the identified genes per current NCCN guidelines. Of the 69 patients, 18 (26%) met the current criteria to receive GT for the additional gene identified in eGT based on their personal and/or family histories. These include four colorectal cancer patients with BRCA1/BRCA2 and a history of breast cancer, one breast cancer patient with MLH1 and a history of endometrial cancer, one colorectal cancer patient with RB1 and a history of retinoblastoma, one colorectal cancer patient with NF1 and features of neurofibromatosis type 1, one prostate cancer patient with FLCN and fibrofolliculomas and lung cysts, and one pancreatic cancer patient with TSC1 and angiomyolipoma, brain lesions, and bilateral renal cysts, which were discovered upon receiving eGT results (Table 1). Nine patients met the GT criteria based on their family histories.

Table 1 Additional high-penetrance P/LP variants identified via eGT in genes not associated with the patient’s cancer type

Fifty-one patients (74% of patients with high penetrance additional findings) did not meet the current criteria to receive GT for the additional gene identified in eGT. These include patients with P/LP variants identified in BRCA1/BRCA2 (n = 9), MSH6/PMS2 (n = 7), FLCN (n = 4), SDHB/SDHC/SDHAF2 (n = 3), TP53 (n = 3), BAP1 (n = 3), CDKN2A (n = 3), DICER1 (n = 3), PALB2 (n = 3), RET (n = 3), CDH1 (n = 1), ETV6 (n = 1), PTCH1 (n = 1), VHL (n = 1), and NF1 (n = 1 (mosaic)). Additionally, six patients had FH P/LP variants (p.Gln376Pro (n = 3), p.His402Tyr (n = 2), p.Gly397Arg (n = 1)) that have been reported in homozygous and compound heterozygous patients with fumarate hydratase deficiency, but have not, to our knowledge, been reported in patients with HLRCC. Five of the six patients with these variants had no known features of HLRCC and one of them had uterine fibroids. Therefore, although these variants were classified as P/LP for AR fumarate hydratase deficiency, whether they confer increased risk for HLRCC is currently uncertain.

Variants of uncertain significance (VUSs) identified in eGT

One of the main concerns restricting the use of eGT is the potential burden of assessing VUSs by laboratories performing the test. To understand the impact of VUSs in variant interpretation and reporting processes of eGT, we analyzed the number of variants classified as VUS in patients with one of the five common cancer types. Overall, 57% (5238/9187) of the patients had at least one VUS identified, with 56.8% (1275/2243), 59.4% (1223/2060), 52.5% (589/1122), 54.5% (898/1648), and 59.3% (1253/2114) of breast, colorectal, ovarian, pancreatic, and prostate cancer patients having at least one VUS, respectively. The number of VUSs identified ranged from zero to nine, with a median of one VUS per patient.