Advertisement

Clinical testing with a panel of 25 genes associated with increased cancer risk results in a significant increase in clinically significant findings across a broad range of cancer histories

Open AccessPublished:September 25, 2017DOI:https://doi.org/10.1016/j.cancergen.2017.09.003

      Highlights

      • Multi-gene hereditary cancer testing detected >1 pathogenic variants (PVs) in 6.7% of individuals.
      • PVs were most common in BRCA1 and BRCA2 (42.2%) and 5 additional breast cancer-risk genes (32.9%).
      • Up to 50% of all clinically significant findings would have been missed by single-syndrome testing.
      Genetic testing for inherited cancer risk is now widely used to target individuals for screening and prevention. However, there is limited evidence available to evaluate the clinical utility of various testing strategies, such as single-syndrome, single-cancer, or pan-cancer gene panels. Here we report on the outcomes of testing with a 25-gene pan-cancer panel in a consecutive series of 252,223 individuals between September 2013 and July 2016. The majority of individuals (92.8%) met testing criteria for Hereditary Breast and Ovarian Cancer (HBOC) and/or Lynch syndrome (LS). Overall, 17,340 PVs were identified in 17,000 (6.7%) of the tested individuals. The PV positive rate was 9.8% among individuals with a personal cancer history, compared to 4.7% in unaffected individuals. PVs were most common in BRCA1/2 (42.2%), other breast cancer (BR) genes (32.9%), and the LS genes (13.2%). Half the PVs identified among individuals who met only HBOC testing criteria were in genes other than BRCA1/2. Similarly, half of PVs identified in individuals who met only LS testing criteria were in non-LS genes. These findings suggest that genetic testing with a pan-cancer panel in this cohort provides improved clinical utility over traditional single-gene or single-syndrome testing.

      Keywords

      Introduction

      Genetic assessment and testing for inherited cancer risk is now a widely used tool in cancer prevention and treatment. Identification of individuals carrying pathogenic variants in hereditary cancer genes allows for targeted interventions for the prevention of cancer through lifestyle and environmental modification, chemoprevention, and/or preventative surgeries. These individuals can also benefit from interventions aimed at early detection of cancer through screening initiated at younger ages, more frequent intervals, and with more sensitive technologies than would be recommended for individuals in the general population (
      • Provenzale D.
      • Gupta S.
      • Ahnen D.J.
      • et al.
      Genetic/familial high-risk assessment: colorectal. NCCN clinical practice guidelines in Oncology.
      ,
      • Daly M.
      • Pilarski R.
      • Berry M.
      • et al.
      Genetic/familial high-risk assessment: breast and ovarian. NCCN clinical practice guidelines in Oncology.
      ). Additionally, there is growing evidence that many cancers arising as a result of hereditary cancer syndromes are candidates for targeted therapies (
      • Goyal G.
      • Fan T.
      • Silberstein P.T.
      Hereditary cancer syndromes: utilizing DNA repair deficiency as therapeutic target.
      ).
      Genetic testing strategies across all areas of medical genetics are evolving in response to expanded knowledge about gene associations and the widespread availability of technologies that allow for cost-effective, high throughput screening. This has led to the development of multi-gene, multi-syndrome panel tests for the assessment of inherited cancer risk as an alternative to the historic strategy of testing a single or limited set of genes based on an analysis of the individual's personal and family history. Panel testing provides a mechanism to address overlap in the clinical presentation of numerous hereditary cancer conditions as well as growing awareness of the limitations of family history as a predictor of genetic risk.
      Although panel testing is becoming more common in clinical practice, there is still considerable debate surrounding the best strategies for utilization and design. One option is to use panel testing as the front-line test for all individuals. Alternatively, panels may be selectively utilized for individuals whose personal and family histories are not a good fit with a single gene or syndrome, or as a second-line option for those who have tested negative for a pathogenic variant as part of single-syndrome testing but whose histories remain highly suspect for an inherited condition. Additionally, questions remain regarding the clinical utility of different approaches to panel design. Many of the panels now available for clinical use are targeted at specific cancers, i.e. breast, ovarian or colorectal cancer. Other “pan-cancer” panels take a broader approach and include genes associated with multiple cancer types, usually focusing on those that are substantial contributors to disease burden in the population and are known to have a significant hereditary component. Although there is an ongoing debate about the best strategy for panel design and the choice of individual genes to be included, there is limited evidence available to support an objective evaluation based on outcomes.
      Beginning in September of 2013, our laboratory has offered a single pan-cancer panel test targeted mainly, but not exclusively, to individuals at risk for the two most common hereditary cancer syndromes: Hereditary Breast and Ovarian Cancer (HBOC) and the hereditary colorectal/endometrial cancer condition, Lynch syndrome (LS). The 25 genes included are known to be significant contributors to risk for one or more of the following eight cancers: breast, ovarian, colorectal, endometrial, pancreatic, gastric, melanoma, and prostate. The panel is heavily weighted toward genes for which findings have concrete clinical relevance. The National Comprehensive Cancer Network (NCCN) and other professional societies currently provide medical management guidelines for individuals with PVs in all but one of the 25 panel genes (
      • Provenzale D.
      • Gupta S.
      • Ahnen D.J.
      • et al.
      Genetic/familial high-risk assessment: colorectal. NCCN clinical practice guidelines in Oncology.
      ,
      • Daly M.
      • Pilarski R.
      • Berry M.
      • et al.
      Genetic/familial high-risk assessment: breast and ovarian. NCCN clinical practice guidelines in Oncology.
      ).
      Here we report on the outcomes of testing with this 25-gene panel for the first 252,223 individuals for whom results were reported. This summary provides insight into the distribution of pathogenic variants identified among the 25 genes and clinically significant findings. This is the largest study to date reporting on the outcomes of clinical testing for hereditary cancer risk in a diverse population with a single pan-cancer panel. This analysis provides valuable information to inform comparisons of panel testing versus the targeted single-syndrome or single-cancer testing, as well as the evaluation of different strategies for panel design.

      Materials and methods

      Cohort characteristics

      This analysis includes a consecutive series of the first 252,223 individuals tested with a 25-gene hereditary cancer panel from September 2013 through July 2016 (Myriad Genetic Laboratories, Inc., Salt Lake City, UT). Testing was performed in a Clinical Laboratory Improvement Amendments (CLIA) and College of American Pathology (CAP) approved laboratory. All individuals provided informed consent for clinical testing. Only data collected as part of clinical testing is utilized here. Ordering providers indicated that the majority of individuals were ascertained for suspicion of Hereditary Breast and Ovarian Cancer (HBOC-Panel) or for suspicion of Lynch syndrome (LS-Panel), and 92.8% of the tested individuals met NCCN criteria for one or both of those conditions. The same panel was run for all individuals.
      For the purposes of this analysis, only results of testing with the full 25-gene panel were included. Specifically excluded were: 1) single-site tests for known familial gene mutations and 2) tests ordered for individuals who previously had genetic testing for inherited cancer risk, including comprehensive testing for mutations in BRCA1 and BRCA2 or the Lynch syndrome genes, or previous testing for the three common Ashkenazi Jewish founder mutations in BRCA1 and BRCA2. Ashkenazi Jewish individuals for whom the 25-gene panel was ordered as the initial test were included in the analysis.

      Panel composition and categorization of genes

      The 25 genes included in the panel are listed in Table 1 along with all of the cancers for which there is sufficient evidence to support a significant association as of July 2016 (
      • Myriad myRisk
      Hereditary cancer gene table.
      ). To facilitate analysis, the genes are grouped into seven categories, based on their primary cancer/syndrome associations, focusing on the cancers widely regarded as most distinctly associated with each gene. Table 1 also includes the source of professional society recommendations for the management of individuals with findings in each gene. BARD1 is the only panel gene for which management guidelines are not yet available.
      Table 1List of genes, associated cancers, and professional society management guidelines. The cancers considered to have a strong association with each gene, and which are most likely to contribute to ascertainment of individuals for testing, are shown in black. Additional cancers for which there is sufficient evidence to support a significant association are shown in gray
      Table thumbnail cgen554-fig-5001

      NGS assay and variant classification

      The details of the Next Generation Sequencing (NGS) assay used have been described previously (
      • Judkins T.
      • Leclair B.
      • Bowles K.
      • et al.
      Development and analytical validation of a 25-gene next generation sequencing panel that includes the BRCA1 and BRCA2 genes to assess hereditary cancer risk.
      ,
      • Tung N.
      • Battelli C.
      • Allen B.
      • et al.
      Frequency of mutations in individuals with breast cancer referred for BRCA1 and BRCA2 testing using next-generation sequencing with a 25-gene panel.
      ,
      • Yurgelun M.B.
      • Allen B.
      • Kaldate R.R.
      • et al.
      Identification of a variety of mutations in cancer predisposition genes in patients with suspected Lynch syndrome.
      ). Briefly, this assay consists of sequencing and large rearrangement detection, followed by data review and reporting. All tests were performed on genomic DNA extracted from whole blood or saliva by QIAsymphony using the DSP DNA Midi kit (Qiagen, Venlo, The Netherlands). Sequencing was performed on an Illumina HiSeq2500 or MiSeq platform (Illumina, Inc., San Diego, CA) and long-range PCR was incorporated to address the highly homologous pseudogenes in the CHEK2 and PMS2 genes. Large rearrangements identified with quantitative dosage analysis of the NGS data were confirmed with microarray CGH and multiplexed ligation-dependent probe amplification (MLPA) analysis. All data were reviewed to assess zygosity and quality metrics. Only those variants detected at allele frequencies between 30% and 70% were regarded as germline in nature, as allele frequencies outside of this range are highly suspicious for representing somatic mosaicism rather than germline inheritance.
      Variants were classified according to current guidelines from the American College of Medical Genetics and Genomics (
      • Richards S.
      • Aziz N.
      • Bale S.
      • et al.
      Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.
      ), as previously described (
      • Eggington J.M.
      • Bowles K.R.
      • Moyes K.
      • et al.
      A comprehensive laboratory-based program for classification of variants of uncertain significance in hereditary cancer genes.
      ). For the purposes of the analyses performed here, variants with a laboratory classification of Deleterious or Suspected Deleterious were considered to be a Pathogenic Variant (PV). Variants with a laboratory classification of Polymorphism or Favor Polymorphism were considered to be Benign (clinically insignificant). Variants for which the clinical significance could not be determined were classified as a Variant of Uncertain Significance (VUS). A small proportion of large rearrangement variants were not counted as PVs if they were reported as Inconclusive findings after the initial NGS findings could not be confirmed with microarray CGH and/or MLPA. This analysis is based on the classification of all variants as of July 2016, regardless of whether or not they had a different classification at the time when the report was issued. Multiple PVs in the same gene established to be in cis (on the same allele) were counted as a single PV in that gene.
      Two special considerations apply to the data analyses presented below. For individuals with variants in the MUTYH gene, responsible for the recessive condition MUTYH-associated Polyposis, only those individuals with biallelic MUTYH PVs are counted as positive for a MUTYH PV. Carriers of a single PV in MUTYH are believed to have a small increased risk for colorectal cancer for which the NCCN has recently provided medical management guidelines (
      • Provenzale D.
      • Gupta S.
      • Ahnen D.J.
      • et al.
      Genetic/familial high-risk assessment: colorectal. NCCN clinical practice guidelines in Oncology.
      ). NCCN also provides management guidelines for individuals with the Ashkenazi Jewish I1307K founder variant in APC (c.3920T >A), which is also believed to be associated with a small increased risk for colorectal cancer in the Ashkenazi Jewish population (
      • Provenzale D.
      • Gupta S.
      • Ahnen D.J.
      • et al.
      Genetic/familial high-risk assessment: colorectal. NCCN clinical practice guidelines in Oncology.
      ). Due to the relatively low level of cancer risk associated with these findings, and the NCCN panel's recognition that data to support special screening are “evolving at this time” (
      • Provenzale D.
      • Gupta S.
      • Ahnen D.J.
      • et al.
      Genetic/familial high-risk assessment: colorectal. NCCN clinical practice guidelines in Oncology.
      ), individuals with a single MUTYH PV or the APC I1307K variant are not counted as positive for a PV in the analyses below and are considered separately.

      Analysis of clinical data

      All clinical and demographic data for participants were obtained from Test Request Forms (TRFs) submitted by ordering healthcare providers. Ancestry, personal history of breast, endometrial/uterine, ovarian, colon/rectal cancer, and personal history of colon/rectal adenomas were documented with check boxes. Personal history of other cancers, family cancer history, and age of diagnosis were documented as free text. No additional information was acquired from tested individuals or providers for this analysis. The ancestry of tested individuals was evaluated for those who indicated one or more ancestries on the TRF, which included check-boxes for the following categories: Central/Eastern European, Northern/Western European, Ashkenazi Jewish, Latin American/Caribbean, African, Asian, Native American, Near/Middle Eastern, Other (write-in field available). Unless otherwise specified, individuals who were identified as Central/Eastern European, Northern/Western European, or Ashkenazi Jewish were categorized as European. The 2010 US census data (
      Generated by Krystal Brown using American FactFinder.
      ) was used to approximate the distribution of ancestries among the general population in these same categories.
      Test results were analyzed according to whether individuals met testing guidelines for HBOC and/or LS. HBOC testing criteria were as defined by NCCN in 2013 (
      • Daly M.
      • Pilarski R.
      • Axilbund J.E.
      • et al.
      Genetic/familial high-risk assessment: breast and ovarian. Version 1.2013. NCCN clinical practice guidelines in Oncology.
      ), excluding the contribution from prostate cancer, as we do not collect the necessary Gleason score information. This may have resulted in an underestimate of the number of men and women meeting HBOC criteria. Lynch syndrome testing criteria were defined as an individual, or one or more of their first- or second-degree relatives, meeting Bethesda criteria or having a diagnosis of endometrial cancer under age 50 (
      • Burt R.W.
      • Barthel J.S.
      • Cannon J.
      • et al.
      Genetic/familial high-risk assessment: colorectal. NCCN clinical practice guidelines in oncology.
      ). These criteria may have overestimated the number of individuals meeting Lynch testing criteria in cases where we were unable to determine if all of the affected second-degree relatives are on the same side of a family. For the purposes of this analysis, individuals with more than five colorectal adenomas were categorized as affected by a cancer diagnosis and individuals for whom five or fewer colorectal adenomas were reported were considered unaffected.

      Results

      Characteristics of tested individuals

      The clinical characteristics of the 252,223 individuals included in this analysis, and the outcomes of their testing broken down by gender, are presented in Table 2. The overwhelming majority of tested individuals were female, comprising 96.9% of the total group. Overall, 60.9% of women and 37.9% of men were unaffected with cancer at the time of testing. The most common diagnoses among women were breast cancer (28.2%), ovarian cancer (4.1%), colorectal cancer (2.3%), and endometrial cancer (2.2%). The most common cancer diagnoses among men were colorectal cancer (31.3%), >5 colorectal adenomas (11.6%), and breast cancer (10.2%).
      Table 2Personal cancer history and demographic information for all individuals tested
      FemaleMaleAll
      N% of Tested IndividualsIndividuals with PVs% of Individuals with PVsN% of Tested IndividualsIndividuals with PVs% of Individuals with PVsN% of Tested IndividualsIndividuals with PVs% of Individuals with PVs
      Total244,31996.9%15,8876.5%79043.1%111314.1%252,223100.0%17,0006.7%
      Personal Cancer Dx
       Breast68,85828.2%64909.4%80510.2%12115.0%69,66327.6%66119.5%
       Ovarian99864.1%136313.6%0099864.0%136313.6%
       Colorectal55392.3%63711.5%247731.3%45518.4%80163.2%109213.6%
       Endometrial54572.2%62611.5%0054572.2%62611.5%
       Gastric2030.1%2311.3%951.2%1717.9%2980.1%4013.4%
       Pancreatic4590.2%5812.6%1842.3%3619.6%6430.3%9414.6%
       Melanoma27511.1%2047.4%1702.2%3017.6%29211.2%2348.0%
       Prostate004585.8%7115.5%4580.2%7115.5%
       C. Adenomas (>5)18760.8%26314.0%91611.6%17819.4%27921.1%44115.8%
       Other12,3455.1%9097.4%80110.1%15319.1%13,1465.2%10628.1%
       Unspecified3110.1%247.7%180.2%211.1%3290.1%267.9%
      Total Affected95,47939.1%90369.5%491062.1%82516.8%100,38939.8%98619.8%
      Unaffected148,84060.9%68514.6%299437.9%2889.6%151,83460.2%71394.7%
      Ancestry
       European124,30350.9%85186.9%453557.4%66514.7%128,83851.1%91837.1%
       Ashkenazi30251.2%1846.1%2292.9%3917.0%32541.3%2236.9%
       African15,3696.3%9446.1%4055.1%4912.1%15,7746.3%9936.3%
       Asian61752.5%4186.8%1942.5%3116.0%63692.5%4497.0%
       Latin Amer/Carib19,2427.9%12396.4%5537.0%8715.7%19,7957.8%13266.7%
       Native American34391.4%1895.5%720.9%1013.9%35111.4%1995.7%
       Near/Middle East17160.7%1136.6%851.1%1315.3%18010.7%1267.0%
       Other15440.6%905.8%580.7%610.3%16020.6%966.0%
       Multiple15,1756.2%9306.1%3614.6%4011.1%15,5366.2%9706.2%
       Not Specified57,35623.5%34466.0%164120.8%21212.9%58,99723.4%36586.2%
      The ancestry of tested individuals for whom a single or multiple ancestries was specified on the TRF is compared to the general US population in Figure 1. This excludes the 23.4% of tested individuals for whom no ancestry was specified on the TRF. European ancestry was reported by 66.7% of tested individuals (Figure 1A), which is only slightly higher than the US population (63.7 %, Figure 1B). The proportion of most of the non-European ancestries is smaller among individuals who underwent hereditary cancer testing compared to their representation in the general population. This appears to be partially due to the 8.0% of individuals in the testing cohort who reported multiple ancestries, compared to only 1.9% of the US population. This multiple ancestry group does not include individuals with exclusively European ancestry, so it is composed of individuals who reported at least one non-European ancestry. If these individuals self-identified as non-European in the census data this could indicate that the ancestry profile of the tested individuals tested is not dissimilar to that of the overall US population. There were no substantial differences in the overall rate at which PVs were detected in individuals according to ancestry, with the exception of ancestries with very few male individuals who had testing (Table 2).
      Figure 1
      Figure 1Comparison of ancestry distribution in A) the tested population and B) the US population based on the 2010 census. Individuals for whom no ancestry was indicated are not included.

      PVs and VUSs detected in 252,223 individuals

      Among the 252,223 tested individuals, 17,000 (6.7%) were found to carry one or more PVs in at least one of the 25 genes included on the panel (Table 2). The overall positive rate in individuals affected by cancer was 9.8%, which is about twice as high as the 4.7% positive rate in those who were unaffected at the time of testing (Table 2). The positive rate was substantially higher in men than women when comparing affected individuals (16.8% versus 9.5%) as well as unaffected individuals (9.6% versus 4.6%). Among women, those with a personal diagnosis of >5 adenomas, ovarian cancer, or pancreatic cancer were the most likely to test positive for a PV. Among men, the highest positive rates were found in those with a personal diagnosis of pancreatic cancer, >5 adenomas, or colorectal cancer.
      A total of 17,340 PVs were detected in this cohort, with the distribution of PVs by gene given in Table 3 (columns A–D) and Figure 2A. The largest percentages of PVs were found in the HBOC genes BRCA1 and BRCA2 (42.2%), breast cancer (BR) genes (32.9%), the LS genes (13.2%), and ovarian cancer (OV) genes (6.8%). Relatively small fractions of PVs were identified in non-LS colorectal cancer (CRC) genes (2.5%), melanoma/pancreatic cancer (M/P) genes (1.0%) and mixed syndrome (MS) genes (1.6%). The high proportion of PVs identified in genes related to breast and ovarian cancer risk (81.8%) is consistent with the finding that 85.7% of the sample met testing criteria for HBOC.
      Table 3Distribution of PVs by gene and gene-associated cancers
      Gene Grouping

      [A]
      Gene

      [B]
      Total PVs by Gene N

      [C]
      Total PVs in Group N (%)

      [D]
      PVs by Gene in Patients with no Personal or Family History of Gene Associated Cancer
      Gene associated cancers are any of the cancers listed in Table 1.


      N (%)

      [E]
      PVs by Group in Patients with no Personal or Family History of Gene Associated Cancer
      Gene associated cancers are any of the cancers listed in Table 1.


      N (%)

      [F]
      Hereditary Breast & Ovarian Cancer (HBOC)BRCA134887311 (42.2%)37 (1.1%)97 (1.3%)
      BRCA2382360 (1.6%)
      Breast Cancer (BR)ATM16135707 (32.9%)203 (12.6%)675 (11.8%)
      CHEK22045260 (12.7%)
      PALB2121068 (5.6%)
      NBN479103 (21.5%)
      BARD136041 (11.4%)
      Ovarian Cancer (OV)BRIP16921174 (6.8%)372 (53.8%)630 (53.7%)
      RAD51C341181 (53.1%)
      RAD51D14177 (54.3%)
      Lynch Syndrome (LS)MLH13762281 (13.2%)7 (1.9%)254 (11.1%)
      MSH252211 (2.1%)
      EPCAM90 (0%)
      MSH666476 (11.4%)
      PMS2710160 (22.5%)
      Non-Lynch Colorectal (non-LS CRC)APC272426 (2.5%)31 (11.4%)56 (13.1%)
      MUTYH (biallelic)12120 (16.5%)
      BMPR1A162 (12.5%)
      SMAD4173 (17.6%)
      Mixed Syndrome (MS)STK1118272 (1.6%)1 (5.6%)12 (4.4%)
      TP531031 (1.0%)
      PTEN576 (10.5%)
      CDH1944 (4.3%)
      Melanoma Pancreatic (M/P)CDKN2A168169 (1.0%)95 (56.5%)96 (56.8%)
      CDK411 (100%)
      * Gene associated cancers are any of the cancers listed in Table 1.
      Figure 2
      Figure 2Distribution of 17,340 PVs in 17,000 tested individuals according to gene category and testing criteria.
      Two PVs in different genes were identified in 338 individuals, with 93 unique gene combinations (Supplemental Table S1). A single individual had PVs identified in 3 different genes; BRCA1, BRIP1 and NBN. There were four individuals with PVs in both copies of CHEK2, all of whom were homozygous for the common CHEK2 variant c.1100del.
      Individuals with a single MUTYH PV or the Ashkenazi Jewish founder APC variant I1307K are not included in Table 3, as these findings are believed to be associated with relatively small colorectal cancer risks. APC I1307K was found in 739 individuals, of whom only 281 (38.0%) reported full or partial Ashkenazi Jewish ancestry and 102 (13.8%) did not specify any ancestry. A single PV in the MUTYH gene was identified in 4287 individuals, which constitutes 1.7% of the tested population. If APC I1307K and monoallelic MUTYH results are included in the overall PV count, then the overall positive rate for one or more PVs with the panel test increases from 6.7% to 8.7%.
      Table 4 shows the number of individuals whose reports included one or more variants of uncertain significance (VUS). About 30% of all tests results include at least one VUS finding. This is lower than the approximately 40% VUS rate reported for the same 25-gene panel in the first year of testing (
      • Tung N.
      • Battelli C.
      • Allen B.
      • et al.
      Frequency of mutations in individuals with breast cancer referred for BRCA1 and BRCA2 testing using next-generation sequencing with a 25-gene panel.
      ,
      • Yurgelun M.B.
      • Allen B.
      • Kaldate R.R.
      • et al.
      Identification of a variety of mutations in cancer predisposition genes in patients with suspected Lynch syndrome.
      ,
      • Langer L.R.
      • McCoy H.
      • Kidd J.
      • et al.
      A study of patients with ovarian cancer tested with a 25-gene hereditary cancer panel.
      ). The remaining 64.5% of individuals had no PVs or VUSs.
      Table 4Distribution of variants identified during pan-cancer panel testing
      Variant FindingsIndividuals (%)
      ≥1 PV17,000 (6.7%)
      ≥1 VUS72,444 (28.7%)
       1 VUS58,213
       2 VUS11,905
       3 VUS1976
       4 VUS287
       5 VUS52
       6 VUS10
       ≥7 VUS1
      No PV or VUS162,779 (64.5%)
      Total252,223

      Clinical impact of test findings

      In order to evaluate the clinical context of the PVs identified as part of pan-cancer panel testing, the distribution of PVs was evaluated according to NCCN testing criteria for HBOC and LS. Overall, 177,717 (70.5%) of tested individuals met testing criteria for only HBOC, 11,839 (6.7%) of whom were found to carry a PV (Table 5). In total, 12,060 PVs were identified in this subgroup. The vast majority (n = 11,000, 91.2%) of PVs were identified in the HBOC, BR and OV genes, with most of the remaining PVs found in the LS genes (n = 690, 5.7%; Figure 2B). The pan-cancer panel identified about twice as many PVs among individuals who met only HBOC testing criteria relative to testing for BRCA1 and BRCA2 alone. Virtually all non-BRCA1 and BRCA2 PVs were detected in genes having some association with breast and ovarian cancer risk.
      Table 5Distribution of tested individuals by testing criteria
      Criteria MetN (% of sample)N (% with PV)
      HBOC Only177,717 (70.5%)11,839 (6.7%)
      LS Only17,743 (7.0%)1393 (7.9%)
      HBOC and LS38,513 (15.3%)2860 (7.4%)
      Neither HBOC or LS18,250 (7.2%)908 (5.0%)
      An additional 17,743 (7.0%) of tested individuals met testing criteria for only LS and 1393 (7.9%) were found to carry a PV (Table 5). There were a total of 1423 PVs identified in this subgroup. Over half (n = 752, 52.8%) of all detected PVs in this group were in the LS genes, with an additional 12.3% (n = 175) in other genes associated with colorectal cancer risk (Figure 2C). An additional one-third (n = 466, 32.7%) of mutations were found in genes thought to be associated primarily with breast and/or ovarian cancer risk, which is a higher proportion of “unexpected” findings (findings inconsistent with the suspected syndrome) in this group relative to individuals who met criteria only for HBOC. These findings are weighted toward the moderate penetrance BR genes, rather than the more highly penetrant HBOC genes, BRCA1 and BRCA2 (Figure 2C). This is consistent with the lower breast cancer risk and lack of ovarian cancer risk associated with these genes, which increases the likelihood of incidentally finding a PV in ATM, CHEK2 or PALB2 in an individual tested for reasons other than suspicion of HBOC.
      Individuals who met testing criteria for both HBOC and LS constituted 15.3% of the testing population and 7.4% of this group were found to carry a PV (Table 5). In this group, 2938 PVs were identified. The majority of PVs were in genes associated with one of these syndromes, with 34.9% (n = 1024) in BRCA1 and BRCA2, 36.0% (n = 1056) in the BR and OV genes, and 23.7% (n = 696) in the LS genes (Figure 2D). This is consistent with the more varied personal and family histories that would have been required to meet both syndrome testing criteria.
      A relatively small proportion (7.2%) of the tested population did not meet testing criteria for either HBOC or LS, only 5.0% (n = 908) of whom were found to carry a PV (Table 5). A total of 919 PVs were identified in this group, diversely distributed among all gene groups. The largest proportion of PVs (n = 360, 39.2%) was identified in the BR genes (Figure 2E). About a quarter of all PVs in this group were in BRCA1 and BRCA2 (n = 235, 25.6%), with another quarter in the LS or non-LS CRC genes (n = 233, 25.4%).
      As an alternative way to establish the extent to which panel testing yields unexpected findings, we determined the proportion of PVs identified in individuals who had no reported personal or family history of any gene-associated cancers, as defined in Table 1. This ranged from a low of 1.3% for the HBOC genes, to a high of over 50% for the M/P and OV genes (Table 3, columns E and F). An inverse relationship between the likelihood of an unexpected finding and the penetrance of a gene was apparent within some of the gene groups. For example, approximately 2% of PVs in the high penetrance LS genes (MLH1, MSH2 or EPCAM) were found in individuals lacking a personal or family history of an LS associated cancer; however, this increased greatly for PVs in the less penetrant LS genes to 11.4% for MSH6 and 22.5% for PMS2. A similar pattern is apparent within the BR genes, comparing the highest penetrance gene, PALB2, to the other genes in this group.

      Discussion

      Here we summarize the results of clinical testing in over 250,000 individuals using a single 25-gene hereditary cancer panel designed to identify inherited risks for breast, ovarian, colorectal, endometrial, gastric, pancreatic, prostate, and melanoma cancers. While this very large dataset can be utilized to investigate a wide range of clinical questions, this analysis focused on providing a broad overview of outcomes and data relevant to evaluating the utility of a pan-cancer strategy for clinical testing.
      Overall, 6.7% of individuals in this cohort were positive for one or more PVs linked to an increased risk for cancer. The 9.8% positive rate in individuals with a personal diagnosis of cancer was just over twice as high as the 4.7% positive rate in those who were unaffected at the time of testing. This is consistent with an unaffected individual being half as likely to carry a pathogenic variant relative to an affected first degree relative. This is also consistent with NCCN recommendations for the testing of unaffected individuals from families in which no affected relatives have been tested or in which no PVs have yet been identified.
      We believe that the data presented here support the utilization of multi-gene panels for the majority of individuals undergoing assessment for inherited cancer risk, and demonstrate the potential value of a pan-cancer approach to panel design. Over half of the PVs detected among individuals who met only HBOC testing criteria were in genes other than BRCA1 and BRCA2, with the largest contribution from the BR genes ATM, CHEK2, PALB2, NBN and BARD1, followed by the OV genes, BRIP1, RAD51C and RAD51D. This is consistent with findings that have been reported in other studies (
      • Tung N.
      • Battelli C.
      • Allen B.
      • et al.
      Frequency of mutations in individuals with breast cancer referred for BRCA1 and BRCA2 testing using next-generation sequencing with a 25-gene panel.
      ,
      • Tung N.
      • Lin N.U.
      • Kidd J.
      • et al.
      Frequency of germline mutations in 25 cancer susceptibility genes in a sequential series of breast cancer patients.
      ). Since there is no reliable way to distinguish those carrying a PV in BRCA1 and BRCA2 from those with a PV in one of the BR or OV genes based on personal and family history, it is difficult to imagine a situation where it would be advantageous to limit testing to only BRCA1 and BRCA2 for individuals meeting HBOC testing criteria.
      Furthermore, we found that 5.7% (n = 690) of the PVs in individuals meeting HBOC testing criteria were in one of the LS genes. This represents a disproportionately large fraction of all the patients identified with LS through the panel test, since 70% of all tested individuals met only HBOC testing criteria. Most of these findings were in the lower penetrance, but clinically important, MSH6 and PMS2 genes, and close to 50% of all the PVs in these two genes were found in individuals who did not meet LS testing criteria (data not shown). Therefore, targeting LS gene testing to only that portion of the sample meeting LS testing criteria would have missed half of the individuals with LS due to PVs in these two genes. This is consistent with the recent study by Espenschied et al, which showed that individuals with pathogenic mutations in MSH6 or PSM2 were less likely to meet LS testing criteria compared to MLH1 and MSH2 mutation carriers (
      • Espenschied C.R.
      • LaDuca H.
      • Li S.
      • et al.
      Multigene panel testing provides a new perspective on Lynch syndrome.
      ).
      A similar trend was observed among individuals who exclusively met LS testing criteria, where only 52.8% of the PVs detected were in one of the LS genes. In this group, 8.6% of the PVs detected were in the HBOC genes, and 19.5% in the BR genes. Overall, a third of the PVs identified were in genes with no established association to either colorectal or endometrial cancer. Here again, panel testing identified a substantial proportion of clinically actionable PVs that would not have been identified with syndrome- or cancer-specific testing.
      The MS genes, TP53, PTEN, STK11 and CDH1, contributed only 1.6% of all the PVs detected in this sample. This is not surprising, considering the rarity of the syndromes with which they are associated, and the possibility that many affected individuals have been targeted to syndrome-specific testing based on distinctive clinical features. Despite these distinctive clinical features, which include high risks for breast, colorectal and other cancers, there are numerous anecdotal reports of unexpected findings in these genes. We found that the proportion of PVs in these genes identified in individuals with no personal or family history of any gene-associated cancer ranged from 1.0% to 10.5%. It remains to be determined if these unexpected findings are a sign of gaps in our understanding of the clinical implications of PVs in these genes, or if they can be explained by testing that occurred in the context of an incomplete personal and family history. This is an important area for future investigation.
      For many of the genes on the panel, including the LS, BR and OV genes, the data presented here show a clear relationship between the magnitude of the cancer risks and the likelihood of finding a PV in a patient who either does not meet testing criteria for that gene, or does not have any personal or family history of a gene-associated cancer. This is illustrated by comparing the data in Figure 2B and 2C, which show that PVs in BRCA1 and BRCA2 outnumber those in the lower penetrance BR genes in patients who meet HBOC testing criteria, whereas the reverse is true in patients who do not meet the HBOC testing criteria. Looking at Columns E and F in Table 3, there is a striking trend toward higher percentages of PVs in genes with lower—albeit clinically significant— cancer risks in individuals lacking any personal or family history of a gene-related cancer. Among the LS genes, the percentage of unexpected findings is highest for PMS2, lowest in MLH1/MSH2/EPCAM, and intermediate in MSH6. This trend mirrors the relative penetrance of each gene. The same pattern is apparent when comparing the five BR genes, either in aggregate against BRCA1 and BRCA2, or individually against each other. Over half of the findings in the OV genes are in individuals with no personal or family history of ovarian cancer, which is not surprising considering that the ovarian cancer risks for these genes are estimated to be between 5.8% and 14.8%, and are only able to manifest in female relatives who carry the PV.
      Our testing population is dramatically weighted toward women with a personal diagnosis of breast cancer and/or individuals who meet current testing criteria for HBOC. However, a striking gender disparity in testing is apparent even for cancers that are not gender-specific, including colorectal, gastric, melanoma and pancreatic. Men constitute only 2% of the over 150,000 unaffected individuals tested, even though men are just as likely as women to have a family history that meets testing criteria for HBOC or LS. Overall, the men who were tested were substantially more likely to have a PV detected, which was observed in every sub-category based on affected status, specific cancer diagnosis, or ancestry. Collectively, this suggests that the index of suspicion may be higher for a man to be tested. It is important to address this apparent gender disparity in testing, considering that findings in most of the 25 genes included on this panel have medical management implications for men as well as women.
      It has been widely reported that genetic testing for inherited cancer risk is also under-utilized in non-European populations (
      • Levy D.E.
      • Byfield S.D.
      • Comstock C.B.
      • et al.
      Underutilization of BRCA1/2 testing to guide breast cancer treatment: black and Hispanic women particularly at risk.
      ,
      • McCarthy A.M.
      • Bristol M.
      • Domchek S.M.
      • et al.
      Health care segregation, physician recommendation, and racial disparities in BRCA1/2 testing among women with breast cancer.
      ). Here we found that the percentage of tested individuals reporting exclusively European ancestry was only slightly higher than that reported in the 2010 US Census data. A direct comparison with the census data for individual non-European ancestries is complicated by the 8% of individuals who listed multiple ancestries in our sample, and we have to consider the possibility of bias in the distribution of individuals for whom ancestry information was not provided (23.5%). However, analysis of this sample raises the possibility that significant disparities in access to testing are less of a problem than previously estimated. In recent years, the NCCN and other professional groups have provided relatively straightforward guidelines for the ascertainment of individuals at risk for HBOC and LS. The consistent application of these guidelines by providers in a wide range of specialties may explain a trend toward more equitable access to genetic testing.
      The relatively high percentage of individuals with one or more VUS is consistent with other reports of panel testing and is an expected outcome of testing that includes large number of genes (
      • Yurgelun M.B.
      • Allen B.
      • Kaldate R.R.
      • et al.
      Identification of a variety of mutations in cancer predisposition genes in patients with suspected Lynch syndrome.
      ,
      • Tung N.
      • Lin N.U.
      • Kidd J.
      • et al.
      Frequency of germline mutations in 25 cancer susceptibility genes in a sequential series of breast cancer patients.
      ,
      • Frey M.K.
      • Kim S.H.
      • Bassett R.Y.
      • et al.
      Rescreening for genetic mutations using multi-gene panel testing in patients who previously underwent non-informative genetic screening.
      ,
      • Lincoln S.E.
      • Kobayashi Y.
      • Anderson M.J.
      • et al.
      A systematic comparison of traditional and multigene panel testing for hereditary breast and ovarian cancer genes in more than 1000 patients.
      ). The high VUS rate is also attributable to the limited availability of clinical data for individual variants in genes for which testing was not widely utilized in the past. This situation is expected to improve as data is gathered and utilized to reclassify prevalent VUSs, which has a large impact on the VUS rate for each gene. The 30% VUS rate in this study already represents a substantial decline from the approximately 40% VUS rate reported for this same 25-gene panel in previous studies (
      • Tung N.
      • Battelli C.
      • Allen B.
      • et al.
      Frequency of mutations in individuals with breast cancer referred for BRCA1 and BRCA2 testing using next-generation sequencing with a 25-gene panel.
      ,
      • Yurgelun M.B.
      • Allen B.
      • Kaldate R.R.
      • et al.
      Identification of a variety of mutations in cancer predisposition genes in patients with suspected Lynch syndrome.
      ,
      • Langer L.R.
      • McCoy H.
      • Kidd J.
      • et al.
      A study of patients with ovarian cancer tested with a 25-gene hereditary cancer panel.
      ).
      One limitation to this analysis is the reliance on provider completed TRFs for the collection of personal and family cancer history. Although some follow-up with providers was performed as part of normal testing operations to resolve errors or missing information, no systematic verification was performed. In addition, the majority of tested individuals were ascertained for HBOC and LS. As such, there may be some reporting bias introduced by ordering providers focusing on one of these syndromes when listing the personal and family history. This may be particularly relevant to the high proportion of unexpected findings in the melanoma/pancreatic cancer genes, CDKN2A and CDK4, since melanoma diagnoses are not a component of either the HBOC or LS testing criteria. Concerns regarding the accuracy and completeness of clinical information submitted on a TRF most likely also reflect problems with the way in which personal and family history information is collected and documented in clinical practice. In this case, our data may be a relatively accurate representation of outcomes in a real world setting. An additional limitation is uncertainty regarding the exact cancer risk spectrum for each gene included on the panel, as displayed in Table 1. Although we believe that Table 1 represents an accurate assessment of the current literature and professional society guidelines, it is not possible to rule out additional associations for which there is as yet insufficient data, i.e., ovarian cancer for PALB2, breast cancer risk for BRIP1 and RAD51C, or colorectal cancer risk for CHEK2.
      Strengths of the study include the large sample size, the diversity of individuals tested, and the use of a single 25-gene panel as the testing modality. Compared with other studies to date on panel testing, and hereditary cancer testing in general, our sample included a high proportion of unaffected individuals. This enhances the relevance of the findings to the cancer prevention setting, and it reduces the bias toward high penetrance findings in samples composed primarily of affected individuals.
      Genetic testing for hereditary cancer risk provides an opportunity to appropriately implement screening and prevention strategies for tested individuals and their families. The data presented here show that genetic testing with a 25-gene pan-cancer panel is more effective at identifying hereditary cancer risks than traditional single-gene or single-syndrome testing. This is demonstrated by the large increase in the detection of clinically significant findings, many of which can be characterized as “unexpected” in the context of the individual's personal and family history. Perhaps it should not be surprising that unexpected findings are relatively common for genes conferring moderate cancer risks, and the data presented here add to a growing body of evidence demonstrating the limitations of genetic testing narrowly tailored to an individual's personal and family history.

      Funding

      This work was supported by Myriad Genetic Laboratories, Inc .

      Conflict of interest

      All authors are employees of Myriad Genetic Laboratories, Inc., and as their compensation, receive salaries and stock options.

      Author contributions

      E. Rosenthal contributed to project conception, data interpretation, manuscript drafting and revisions. R. Bernhisel and J. Kidd contributed to data analysis and manuscript revision. K. Brown contributed to data interpretation, manuscript drafting and revision. S. Manley contributed to project conception and manuscript revision.

      Supplementary data

      The following is the supplementary data to this article:

      References

        • Provenzale D.
        • Gupta S.
        • Ahnen D.J.
        • et al.
        Genetic/familial high-risk assessment: colorectal. NCCN clinical practice guidelines in Oncology.
        (Version 2. 2016; Available at:) (Accessed 26 September 2016)
        • Daly M.
        • Pilarski R.
        • Berry M.
        • et al.
        Genetic/familial high-risk assessment: breast and ovarian. NCCN clinical practice guidelines in Oncology.
        (Version 1.2017; Available at:)
        • Goyal G.
        • Fan T.
        • Silberstein P.T.
        Hereditary cancer syndromes: utilizing DNA repair deficiency as therapeutic target.
        Fam Cancer. 2016; 15: 359-366
        • Myriad myRisk
        Hereditary cancer gene table.
        (Available at:)
        • Judkins T.
        • Leclair B.
        • Bowles K.
        • et al.
        Development and analytical validation of a 25-gene next generation sequencing panel that includes the BRCA1 and BRCA2 genes to assess hereditary cancer risk.
        BMC Cancer. 2015; 15: 215
        • Tung N.
        • Battelli C.
        • Allen B.
        • et al.
        Frequency of mutations in individuals with breast cancer referred for BRCA1 and BRCA2 testing using next-generation sequencing with a 25-gene panel.
        Cancer. 2015; 121: 25-33
        • Yurgelun M.B.
        • Allen B.
        • Kaldate R.R.
        • et al.
        Identification of a variety of mutations in cancer predisposition genes in patients with suspected Lynch syndrome.
        Gastroenterol. 2015; 149: 604-613
        • Richards S.
        • Aziz N.
        • Bale S.
        • et al.
        Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.
        Genet Med. 2015; 17: 405-424
        • Eggington J.M.
        • Bowles K.R.
        • Moyes K.
        • et al.
        A comprehensive laboratory-based program for classification of variants of uncertain significance in hereditary cancer genes.
        Clin Genet. 2014; 86: 229-237
      1. Generated by Krystal Brown using American FactFinder.
        (Available at:)
        http://factfinder.census.gov
        Date: 2015
        Date accessed: August 31, 2015
        • Daly M.
        • Pilarski R.
        • Axilbund J.E.
        • et al.
        Genetic/familial high-risk assessment: breast and ovarian. Version 1.2013. NCCN clinical practice guidelines in Oncology.
        (Available at:)
        • Burt R.W.
        • Barthel J.S.
        • Cannon J.
        • et al.
        Genetic/familial high-risk assessment: colorectal. NCCN clinical practice guidelines in oncology.
        (Version 1.2011)2012 (Available at:) (Accessed 2016)
        • Langer L.R.
        • McCoy H.
        • Kidd J.
        • et al.
        A study of patients with ovarian cancer tested with a 25-gene hereditary cancer panel.
        J Community Support Oncol. 2016; 14: 314-319
        • Tung N.
        • Lin N.U.
        • Kidd J.
        • et al.
        Frequency of germline mutations in 25 cancer susceptibility genes in a sequential series of breast cancer patients.
        J Clin Oncol. 2016; 34: 1460-1468
        • Espenschied C.R.
        • LaDuca H.
        • Li S.
        • et al.
        Multigene panel testing provides a new perspective on Lynch syndrome.
        J Clin Oncol. 2017; 35: 2568-2575
        • Levy D.E.
        • Byfield S.D.
        • Comstock C.B.
        • et al.
        Underutilization of BRCA1/2 testing to guide breast cancer treatment: black and Hispanic women particularly at risk.
        Genet Med. 2011; 13: 349-355
        • McCarthy A.M.
        • Bristol M.
        • Domchek S.M.
        • et al.
        Health care segregation, physician recommendation, and racial disparities in BRCA1/2 testing among women with breast cancer.
        J Clin Oncol. 2016; 34: 2610-2618
        • Frey M.K.
        • Kim S.H.
        • Bassett R.Y.
        • et al.
        Rescreening for genetic mutations using multi-gene panel testing in patients who previously underwent non-informative genetic screening.
        Gynecol Oncol. 2015; 139: 211-215
        • Lincoln S.E.
        • Kobayashi Y.
        • Anderson M.J.
        • et al.
        A systematic comparison of traditional and multigene panel testing for hereditary breast and ovarian cancer genes in more than 1000 patients.
        J Mol Diagn. 2015; 17: 533-544