VOLUME 34 • NUMBER 34 • DECEMBER 1, 2016 JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T Conflicting Interpretation of Genetic Variants and Cancer Risk by Commercial Laboratories as Assessed by the Prospective Registry of Multiplex Testing Judith Balmaña, Laura Digiovanni, Pragna Gaddam, Michael F. Walsh, Vijai Joseph, Zsofia K. Stadler, Katherine L. Nathanson, Judy E. Garber, Fergus J. Couch, Kenneth Offit, Mark E. Robson, and Susan M. Domchek See accompanying editorial on page 4061 Judith Balmaña, Hospital Vall d’Hebron and Universitat Autònoma de Barcelona, A B S T R A C T Barcelona, Spain; Laura Digiovanni, Susan M. Domchek, and Katherine L. Purpose Nathanson, University of Pennsylvania, Massively parallel sequencing allows simultaneous testing of multiple genes associated with cancer Philadelphia, PA; Pragna Gaddam, susceptibility. Guidelines are available for variant classification; however, interpretation of these Michael F. Walsh, Vijai Joseph, Zsofia K. guidelines by laboratories and providers may differ and lead to conflicting reporting and, potentially, Stadler, Kenneth Offit, and Mark E. Robson, Memorial Sloan Kettering Cancer to inappropriate medical management. We describe conflicting variant interpretations between Center and Weill Cornell Medical College, Clinical Laboratory Improvement Amendments–approved commercial clinical laboratories, as re- New York, NY; Judy E. Garber, Dana- ported to the Prospective Registry of Multiplex Testing (PROMPT), an online genetic registry. Farber Cancer Institute, Boston, MA; and Fergus J. Couch, Mayo Clinic, Rochester, Methods MN. Clinical data and genetic testing results were gathered from 1,191 individuals tested for inherited cancer Published online ahead of print at susceptibility and self-enrolled in PROMPT between September 2014 and October 2015. Overall, 518 www.jco.org on September 12, 2016. participants (603 genetic variants) had a result interpreted by more than one laboratory, including at least Processed as a Rapid Communication one submitted to ClinVar, and these were used as the final cohort for the current analysis. manuscript. Results Authors’ disclosures of potential conflicts Of the 603 variants, 221 (37%) were classified as a variant of uncertain significance (VUS), 191 (32%) as of interest are found in the article online at pathogenic, and 34 (6%) as benign. The interpretation differed among reporting laboratories for 155 (26%). www.jco.org. Author contributions are Conflicting interpretations were most frequently reported for CHEK2 and ATM, followed by RAD51C, PALB2, found at the end of this article. BARD1, NBN, and BRIP1. Among all participants, 56 of 518 (11%) had a variant with conflicting interpretations Corresponding author: Judith Balmaña, ranging from pathogenic/likely pathogenic to VUS, a discrepancy that may alter medical management. MD, PhD, Medical Oncology Department, Hospital Vall d’Hebron and Vall d’Hebron Conclusions Institute of Oncology, Paseo Vall Conflicting interpretation of genetic findings from multiplex panel testing used in clinical practice is d’Hebron 119-129, Barcelona 08035, frequent and may have implications for medical management decisions. Spain; e-mail: jbalmana@vhio.net. © 2016 by American Society of Clinical J Clin Oncol 34:4071-4078. © 2016 by American Society of Clinical Oncology Oncology 0732-183X/16/3434w-4071w/$20.00 makes it difficult to obtain reliable estimates of the INTRODUCTION DOI: 10.1200/JCO.2016.68.4316 associated cancer risk.1 Variants of uncertain significance are fre- Advances in next-generation sequencing tech- quent in panel testing and can be challenging to nology have led to the development of multiplex resolve. Despite the availability of public databases panel testing for the molecular diagnosis of inherited for sharing genetic variants, the development of cancer susceptibility. Commercially available panels differ in their exact composition but usually include prediction models based on protein structure and moderate-penetrance and high-penetrance genes function, and the potential for laboratory-based (with mutations reported in the literature to be functional analyses to determine the pathogenicity associated with a relative risk [RR] of cancer between of some variants, discordant interpretation of the 2 and 5, or greater than 5, respectively). Genes in clinical pathogenicity of variants remains a frequent which mutations are associated with susceptibility to problem.2-4 Different standardized classification inherited cancer have been rapidly incorporated into systems for interpretation of sequence-based results these panels, often before robust evidence of the have been developed.5,6 The ClinVar database7,8 is magnitude of the association is known. For some a publicly available database that has allowed clinical genes, the relatively low prevalence of mutations laboratories to submit their identified variants and © 2016 by American Society of Clinical Oncology 4071 Balmaña et al share their interpretation. All accessions of the same genetic variant who had undergone tumor testing, who did not have any genetic findings, provided by different submitters are maintained in ClinVar, which or whose unique finding was a BRCA result, were excluded (n = 112), as allows tracking of the changes and sharing of the evidence used for were individuals with self-reported results for which a ClinVar submission or test report was not available for confirmation (n = 410). From the 669 interpretation. However, submission to ClinVar is voluntary and not remaining participants, 518 (43%) had results interpreted by more than all laboratories choose to submit data. Moreover, resolution of the one laboratory (including at least one in ClinVar) or findings with multiple diverse submissions is a voluntary activity for experts and a Herculean submissions reported in ClinVar, and these participants were used as the task. From a clinical perspective, it is relevant to quantitate the fre- final cohort for the current analysis (Fig 1). Because commercial labo- quencies and describe the patterns of variants with conflicting in- ratories are not required to submit to ClinVar, if there were known dif- terpretations, particularly those that may impact medical management. ferences in classification of a genetic test result among laboratories Unrelated individuals or even members from the same family observed through reporting in the PROMPT registry, then this result was classified as a conflicting interpretation of pathogenicity. That is, a patient with the same genetic variant tested by different clinical labo- may have provided a report from a non-ClinVar submitting laboratory and ratories may be given a different clinical assessment of variant the same finding may have been entered into ClinVar by another pathogenicity. The ordering provider may not be aware that a dif- laboratory. ferent laboratory testing a patient’s relative has provided a different All genetic test results were checked in the ClinVar public archive interpretation. It is crucial to identify variants with conflicting in- (http://www.ncbi.nlm.nih.gov/clinvar/) and their clinical significance was terpretations among laboratories, the frequencies and types of assigned according to the submissions by different laboratories through clinical testing or submissions from research and literature curation. In discrepancies, and the underlying reasons for these discrepancies to ClinVar, if differences in interpretation among submitters are observed, the enable specific guidance for variant curation and to increase the genetic test results are classified as conflicting interpretation of patho- consistency of variant interpretation among the laboratories. genicity.7 A search for any update in reclassification was performed at the Here, we describe the frequencies and types of genetic findings time of data analysis lock, with no change. Descriptive statistics were used with conflicting interpretations in non-BRCA genes tested as to describe the study population. part of panels assembled by Clinical Laboratory Improvement Amendments–approved commercial laboratories. Data were col- lected from individuals who underwent clinical testing and en- RESULTS rolled in a prospective registry. This registry includes patients with results from some commercial laboratories that currently do not Overall, 518 participants enrolled into PROMPT were considered deposit data in ClinVar. eligible for this analysis. Their median age was 52 years (range, 44-61 years) and 95% were female. Overall, 427 (82%) were white, and 350 (68%) had a cancer diagnosis, mostly breast cancer PATIENTS AND METHODS (n = 188; 36%). Thirty-one percent had multiple primary tumors. A total of 419 participants (81%) reported being invited into Overall, 1,191 individuals tested for cancer susceptibility genes as part of PROMPT by their health-care provider or by the laboratory where commercial multiplex panels self-enrolled in the ongoing Prospective the testing had been performed (Table 1). Registry of Multiplex Testing (PROMPT) between September 1, 2014, and These 518 participants reported 603 genetic variants with September 30, 2015, the period for which data were ascertained for this analysis. multiple interpretations by several commercial laboratories and/or The PROMPT registry partnered with several clinical laboratories, submissions to ClinVar. Of the 518 participants included in this including Ambry Genetics, Color Genomics, GeneDx, Invitae, Myriad analysis, 165 provided information from testing done in a labo- Genetics, Pathway Genomics, and Quest Diagnostics, to recruit individuals ratory that does not submit to ClinVar. The most frequent gene who had undergone genetic panel testing and had at least one variant in with sequence alterations reported through PROMPT was CHEK2 their report. Participating laboratories advertised PROMPT within the (n = 117), followed by ATM (n = 105) (Fig 2). packet of test report forms sent to individuals. Health-care providers also Regarding the type of results according to their clinical in- were educated about PROMPT at academic and industry meetings, and through e-mails. An informational Web site and video were created for terpretation in ClinVar, 220 (36%) were consistently classified as participants and providers. An informational article was communicated VUS, 191 (32%) as pathogenic/likely pathogenic, and 34 (6%) as via the Dr Susan Love Research Foundation e-newsletter to individuals benign/likely benign, while 155 (26%) were classified as conflicting enrolled in the Army of Women. Some participants found the study di- interpretation. Among these 155, 26% of them were in CHEK2; rectly through Internet. 20% in ATM; 8% in RAD51C; 7% in PALB2; 5% in BARD1; 4% in An enrollment site for PROMPT was built on a platform maintained NBN and APC; 3% in RAD50, PMS2, TP53, and MUTYH; 2% in by a partner organization, PatientCrossroads. To enroll, participants BRIP1 and FANCC; and the remainder were distributed among the created an account with PatientCrossroads and consented to participate in an online genetic registry in either an identifiable (contact information other genes at approximately 1% each (Fig 3). Of the 155 dis- available to research team) or de-identified (contact information not cordant findings, 56 (36%) were reported as pathogenic/likely available to research team) manner. All participants completed the baseline pathogenic by at least one laboratory but not by all laboratories questionnaire with personal and family cancer history of cancer, genetic (ie, clinically significant; Table 2). testing, and demographics. Participants were given the opportunity to Of 117 findings with multiple interpretations for CHEK2, 41 upload their genetic testing report to the PatientCrossroads portal or to (35%) were conflicting, and the majority (n = 36; 88%) would be send it directly to PROMPT registry staff. From the initial 1,191 participants assessed for eligibility, individuals characterized as clinically significant. Eighteen were c.470T.C considered for this analysis were required to have verifiable genetic data by (p.Ile157T) variant and 12 were c.1283C.T (p.Ser428Phe), both PROMPT staff (ie, a copy of a test report submitted, or a detailed self- classified as either pathogenic/likely pathogenic or as VUS reported test result in the registry that was found in ClinVar). Participants (Table 3). Other variants in this gene with a conflicting 4072 © 2016 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY Conflicting Interpretations in Commercial Panel Testing Assessed for eligibility for PROMPT (N = 1,191) Total excluded participants (n = 112) BRCA1/2 findings only (n = 100) Somatic findings only (n = 3) No findings on germline test reports (n= 3) Single site testing (n = 6) Total excluded (n = 410) Assessed verified data Patient-reported findings (gene and (n = 1,079) variant provided) not in ClinVar (n = 111) Incomplete information provided (n = 299) Fig 1. CONSORT diagram showing the flow of participants and genetic variants per participant from the PROMPT registry until inclusion for current analysis. PROMPT, Total with verifiable data (test report directly Prospective Registry of Multiplex Testing. reviewed or patient-reported finding present in ClinVar) Participants (n = 669) Findings (n = 736) Excluded if only one verified report and, therefore, no comparison available Participants (n = 151) Findings (n = 197) Total with multiple submissions Participants (n = 518) Findings (n = 603) interpretation between pathogenic/likely pathogenic or VUS are variants with conflicting interpretations were in CHEK2, followed described in Table 2. In addition, 32 of 105 variants (30%) in the by ATM, RAD51C, and PALB2. Many conflicting interpretations ATM gene with multiple submissions were classified as conflicting. are of low clinical significance because the discrepancy ranged However, all of these ATM variants ranged from benign/likely between an interpretation of benign/likely benign or VUS; benign or VUS (Appendix Fig A1, online only; Table 2) and, therefore, medical management should default to personal and therefore, should not alter medical management. In PALB2, nine of family history. However, the identification of a VUS can cause 49 findings (18%) reported with multiple submissions were a great deal of uncertainty for patients and providers alike and conflicting; one, c.3113G.A (p.Trp1038Ter), which was reported increase the risk for inappropriate medical management.5 For three times, was classified as either pathogenic or VUS, and the example, it is inappropriate to recommend oophorectomy based remainder were interpreted as either benign or VUS. In BRIP1, on a VUS finding alone. Of greater concern, 36% of conflicting three of 33 findings (9%) were discordant; one of them was re- results appeared to be clinically relevant, because they were either ported twice [c.139C.G (p.Pro47Ala)] and was classified as either reported as pathogenic/likely pathogenic or as a VUS by different likely pathogenic or VUS. In RAD51C, 13 of 26 (50%) findings clinical laboratories. In this regard, CHEK2, PALB2, and BRIP1 (50%) were conflicting. Twelve of these findings corresponded to were most frequently identified as having discordant interpretation the variant c.790G.A (p.Gly264Ser), which was submitted as between these two levels of pathogenicity. As these genes are being either benign or VUS, and one, the c.1026+5_1026+7delGTA incorporated into clinical practice as part of cancer risk assessment, variant, was interpreted as both likely pathogenic and VUS. In the tailored screening and cancer prevention recommendations, or NBN, six of 20 findings (30%) reported were conflicting. Two of within tumor panel sequencing for potential targeted therapy, it them, c.511A.G (p.Ile171Val) and c.643C.T (p.Arg215Trp), will be critical to standardize their curation and clinical classifi- were reported twice each, and had a two-step difference in con- cation of variants to provide appropriate management for mu- flicting interpretation in ClinVar (pathogenic, VUS, benign/likely tation carriers and their families. benign). Other genes with conflicting interpretation are listed in A strength of this study is the inclusion of individuals who Table 2. underwent testing at laboratories that do not submit data to ClinVar. Our data highlight several specific variants of interest related to differential reporting. Of 117 of the CHEK2 findings DISCUSSION reported in PROMPT to date, 41 (35%) were conflicting. The c.470T.C (p.Ile157Thr, I157K) and the c.1283C.T (p.Ser428Phe, One quarter of the clinical genetic results from commercially S428F) variants were the most common with conflicting in- available multiplex cancer panels and reported at the PROMPT terpretation between pathogenic/likely pathogenic and VUS. The registry had conflicting interpretations within ClinVar. Most of the putative pathogenicity of I157T has long been studied. Several www.jco.org © 2016 by American Society of Clinical Oncology 4073 Balmaña et al of clinical interpretations from VUS to pathogenic (low pene- Table 1. Study Population Enrolled in the Prospective Registry of Multiplex Testing Registry trance), likely pathogenic, and pathogenic (Table 3). Differences in interpretation of the evidence for the pathoge- Eligibility, Final Cohort, Cohort No. (%) No. (%) nicity of CHEK2 S428F are less pronounced but still lead to con- No. (%) 1,191 518 flicting reports of pathogenic, pathogenic/low penetrance, and VUS. Median age, (range), years 52 (43-60) 52 (44-61) All laboratories agree that the S428F variant is located in a well- Sex established functional domain16; and all but one agree that there Male 59 (5) 27 (5) is evidence of an increased prevalence in affected individuals v Female 1,111 (93) 491 (95) Unknown 21 (2) — controls.9 However, three laboratories report that functional Race, no. (%) studies are supportive of the damaging effect of CHEK2 S428F, Black 10 (0.8) 4 (0.77) and one laboratory concludes that the impairment is variable. Asian 29 (2) 7 (1) Finally, two laboratories report the literature evidence from White 824 (69) 427 (82) Multiple 45 (4) 22 (4) cosegregation of the CHEK2 variant with the disease; only one Other/unknown 283 (24) 58 (11) laboratory reports the prediction of pathogenicity by in silico Cancer algorithms, and another laboratory emphasizes the contribution Yes 774 (65) 350 (68) of this variant to tumorigenesis by loss of heterozygosity (Table 3). No 363 (30) 162 (31) Unknown 54 (5) 6 (1) The examples of the discrepancies involving CHEK2 I157T and Cancer diagnoses S428F also may reflect the challenges of describing a so-called low- Blood/lymphoma 4 (0.5) 4 (1) penetrance susceptibility allele (RR , 2) in a format designed for Brain 1 (0.1) 1 (0.2) Breast 397 (51) 188 (54) high-penetrance alleles.17 Cervical/uterine 15 (2) 7 (2) Another interesting variant with conflicting interpretation Ovarian/fallopian tube 50 (6) 21 (6) between likely pathogenic and VUS is the c.139C.G substitution Colon/rectal 25 (3) 7 (2) (p.Pro47Ala) in the BRIP1 gene, which was observed in two un- Kidney/renal pelvis 2 (0.3) 1 (0.2) Melanoma 8 (1) 1 (0.2) related participants from PROMPT. This variant was first described Nonmelanoma skin 11 (1) 4 (1) in 2001,18 when the BRIP1 protein was found to interact with Pancreas 3 (0.4) 1 (0.2) BRCA1 and contribute to its DNA repair function. This variant was Prostate 6 (0.8) 4 (1) initially identified in an individual with early-onset breast cancer Thyroid 8 (1) 2 (0.4) Multiple primaries 232 (30) 108 (31) and a family history of breast and ovarian cancer; segregation Not specified/other 12 (1) 1 (0.2) analysis was not available and loss of heterozygosity in the tumor Referral source was not demonstrated. Since then, it has been reported in the Family/friend 14 (1) 4 (0.77) literature several times as a breast or ovarian cancer susceptibility Professional/laboratory 775 (65) 419 (81) Social media/Susan Love Foundation 45 (4) 6 (1) gene in affected individuals, but has also been observed in control Online/Internet 18 (2) 7 (1) subjects.19,20 Two clinical laboratories providing a summary of Other/unsure/no response 339 (28) 82 (16) their clinical interpretation in ClinVar differ in their assessment of the meaning of loss of function observed in the experimental studies in regard to cancer susceptibility; one laboratory highlights reports analyze its association with breast cancer risk or other that it is present in population databases, albeit as a very rare allele tumors.9-13 Functional analyses of this variant have also been (0.04%); and only one endorses the supporting evidence provided published14-16 and in silico predictions are available. The frequency by the prediction of in silico algorithms. In phase 2 of the PROMPT of this variant in the population is 0.4% in the National Heart, project, enrollment of family members will be encouraged and Lung, and Blood Institute Exome Variant Server and approxi- cosegregation analysis of this BRIP1 variant may be undertaken, mately 5% are found in some Northern European populations.9 which may help provide one more piece of supporting evidence to Despite these data, there was discordant interpretation of this clarify its pathogenicity. evidence by the different clinical genetic laboratories. For instance, Findings of discordant interpretation of results in genetic two laboratories found the functional analysis data compelling testing are not limited to oncologic settings nor ClinVar. Similar enough to suggest a damaging effect on protein function and to discrepancies also have been observed in hereditary connective- influence variant interpretation, whereas two other laboratories felt tissue disorders and are felt to be due to lack of submission of data that this variant had no effect on CHEK2 protein kinase activity to public databases, limited use of allele frequency data, and and the relationship between functional studies and cancer as- varying consideration of protein structure and function.3 In the sociation is unclear. Only one laboratory supported the association current study, the discrepancies between commercial laboratories with cancer risk as being significant, whereas the other three in interpretation of variants in cancer-related genes seem to be laboratories reported an increased prevalence in affected in- mostly based on differences in the interpretation of evidence. dividuals, but also documented high frequencies observed among Efforts have already been initiated among laboratories to resolve controls in diverse populations. All agreed that the CHEK2 variant these differences.21 The lack of a gold standard test for pathoge- is located in a well-established functional domain, but only one nicity, or a noncontroversial interpretation of functional analyses, laboratory used the supporting limited evidence provided from suggests that discrepant interpretation of challenging variants, predictions of in silico algorithms of this variant’s effect. Overall, particularly missense and splice-site variants, will persist for these discrepancies in interpretation of the evidence lead to a range some time. 4074 © 2016 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY Conflicting Interpretations in Commercial Panel Testing SMAD4, 1 STK11, 2 TP53, 15 APC, 10 SDHD, 1 TSC2, 1 RAD51D, 9 RAD50, 17 RAD51C, PTEN, 4 26 PMS2, 21 ATM, 105 PALB2, 50 BARD1, 23 BMPR1A, 2 Fig 2. Distribution of genetic variants with NF1, 4 BRCA1, 17 multiple submissions in ClinVar by gene (N = 603). NBN, 20 BRCA2, 23 MUTYH, 22 BRIP1, 33 MSH6, 20 CDH1, 22 MSH2, 13 MRE11A, 8 CDK4, 1 CHEK2, 117 MLH1, 7 CDKN2A, 5 FH, 1 FANCC, 3 Multigene testing for cancer susceptibility is a complex en- rates of discrepant interpretation reported herein support the need deavor characterized by challenges in curation and reporting of for initiatives focused on harmonizing variant interpretation in the variants; unfortunately, this study demonstrates that conflicting context of shared data. We encourage clinical laboratories to interpretation of those variants may be relatively frequent. The submit their findings to ClinVar and other public databases with 221, 37% 34, 5% 45 191, 32% 40 155, 26% 35 Fig 3. Distribution of genetic variants according to ClinVar interpretation (N = 603), 30 and the absolute number of variants with conflicting interpretation by gene (n = 155). 25 20 Pathogenic/likely pathogenic VUS 15 Benign/likely benign 10 Conflicting 5 0 TSC2 STK11 NF1 MSH6 MLH1 FH MRE11A CDKN2A CDH1 BMPR1A FANCC BRIP1 BRCA2 RAD50 PMS2 MUTYH TP53 NBN APC BARD1 PALB2 RAD51C ATM CHEK2 www.jco.org © 2016 by American Society of Clinical Oncology 4075 Balmaña et al Table 2. Genetic Variants With Conflicting Interpretation Between Pathogenic/Likely Pathogenic and Uncertain Significance (n = 57) Gene No. With Multiple Submissions Total No. (%) Description of Variants (No. identified) Type of Variant APC 10 4 (40) c.3920T.A (p.Ile1307Lys) (2) Missense c.1240C.T (p.Arg414Cys) (2) Missense BRIP1 33 2 (6) c.139C.G (p.Pro47Ala) (2) Missense CDKN2A 5 2 (40) c.9_32dup24 (2) (p. Duplication Pro11_Ser12insAlaAlaGlySerSerMetGluPro) CHEK2 117 36 (31) c.470T.C (p.Ile157Thr) (18) Missense c.1283C.T (p.Ser428Phe) (12) Missense c.190G.A (p.Glu64Lys) (2) Missense c.1111C.T (p.His371Tyr) Missense c.1427C.T (p.Thr476Met) Missense c.483_485delAGA (p.Glu161del) Deletion c.715G.A (p.Glu239Lys) Missense FH 1 1 c.1431_1433dupAAA (p.Lys477_Asn478insLys) Duplication MSH6 20 1 (5) c.2927G.A (p.Arg976His) Missense MUTYH 22 3 (14) c.821G.A (p.Arg274Gln) Missense c.934-2A.G (2) Splice acceptor NBN 20 4 (20) c.511A.G (p.Ile171Val) (2) Missense c.643C.T (p.Arg215Trp) (2) Missense PALB2 49 3 (6) c.3113G.A (p.Trp1038Ter) (3) Nonsense RAD51C 26 1 (4) c.1026+5_1026+7delGTA Deletion Between Benign/Likely Benign and Uncertain Significance (n = 99) APC 10 2 (20) c.3352A.G (p.Asn1118Asp) Missense c.6821C.T (p.Ala2274Val) Missense ATM 105 32 (30) c.1229T.C (p.Val410Ala) (7) Missense c.4388T.G (p.Phe1463Cys) (6) Missense c.6067G.A (p.Gly2023Arg) (5) Missense c.1810C.T(p.Pro604ser) (4) Missense c.2932T.C (p.Ser978Pro) (2) Missense c.4424A.G (p.Tyr1475Cys) (2) Missense c.1066-6T.G (IVS8-6 T.G) Missense c.1073A.G (p.Asn358Ser) Missense c.3925G.A (p.Ala1309Thr) Missense c.4709T.C (p.Val1570Ala) Missense c.7390T.C (p.Cys2464Arg) Missense c.8734A.G (p.Arg2912Gly) Missense BARD1 23 8 (35) c.2282G.A (p.Ser761Asn) (4) Missense c.33G.T (p.Gln11His) Missense c.1694G.A (p.Arg565His) Missense c.1977A.G (p.Arg659 = ) Synonymous c.2191C.G (p.Arg731Gly) Missense BMPR1A 2 2 (100) c.1243G.A (p.Glu415Lys) Missense c.1327C.T (p.Arg443Cys) Missense BRCA2 23 3 (12) c.6803G.A (p.Arg2268Lys) Missense c.10121C.T(p.Thr3374lle) Missense c.1514T.C (p.Ile505Thr) Missense BRIP1 33 1 (3) c.587A.G (p.Asn196Ser) Missense CDH1 19 2 (10) c.670C.T (p.Arg224Cys) Missense c.1774G.A (p.Ala592Thr) Missense CDKN2A 5 1 (20) c.361G.A (p.Ala121Thr) Missense CHEK2 117 5 (4) c.320-5T.A (2) Intron variant c.538C.T (p.Arg180Cys) (2) Missense c.707T.C (p.Leu236Pro) Missense FANCC 3 3 (100) c.345+6A.T Intron variant c.632C.G (p.Pro211Arg) Missense c.178G.A (p.Val60Ile) Missense MLH1 7 1 (17) c.2252A.G (p.Lys751Arg) Missense MRE11A 8 2 (25) c.1475C.A (p.Ala492Asp) (2) Missense MUTYH 22 1 (4) c.1276C.T (p.Arg426Cys) Missense NBN 20 2 (10) c.1720T.A (p.Leu574Ile) (2) Missense NF1 4 1 (25) c.8041A.G (p.Ile2681Val) Missense PALB2 49 6 (12) c.2228A.G (p.Tyr743Cys) Missense c.2816T.G (p.Leu939Trp) (3) Missense c.656A.G (p.Asp219Gly) (2) Missense PMS2 21 4 (19) c.86G.C (p.Gly29Ala) (2) Missense c.620G.A (p.Gly207Glu) (2) Missense (continued on following page) 4076 © 2016 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY Conflicting Interpretations in Commercial Panel Testing Table 2. Genetic Variants With Conflicting Interpretation (continued) Between Benign/Likely Benign and Uncertain Significance (n = 99) Gene No. With Multiple Submissions Total No. (%) Description of Variants (No. identified) Type of Variant RAD50 17 4 (24) c.3036+5G.A (3) Missense c.943G.T (p.Val315Leu) Missense RAD51C 26 12 (46) c.790G.A (p.Gly264Ser) (12) Missense STK11 2 1 c.1211c.T (p.Ser404Phe) Missense TP53 15 5 (33) c.248C.T (p.Ala83Val) Missense c.704A.G (p.Asn235Ser) Missense TSC2 1 1 (100) c.5378G.A (p.Arg1793Gln) Missense relevant clinical interpretations of findings, as well as the evidence for the Interpretation of Germline Mutant Alleles (ENIGMA), and on which interpretations are based. To better identify the un- International Society for Gastrointestinal Hereditary Tumors (In- derlying reasons for their discrepant interpretation, laboratories SIGHT) have demonstrated the utility of multidisciplinary collabo- could explicitly describe their weighting (from very strong to ration to curate and reclassify submitted VUS.22,23 Platforms such as moderate or supporting) for each pathogenic criterion used for the Leiden Open Source Variation Database (LOVD)24, which allows classification.6 This transparency in data sharing and collaboration for collection, curation, and display of phenotypes and DNA sequence between academic research consortia and commercial laboratories variants, and the BRCA Challenge (http://brcaexchange.org/; an in- may promote strategies to standardize clinical variant curation ternational effort to review and provide vetted data on BRCA1 and algorithms. Ongoing consortia such as Evidence-Based Network BRCA2 gene variants) are additional relevant resources. PROMPTuses Table 3. Criteria Used for Clinical Interpretation of CHEK2 Variants p.S428F and p.I157K Criteria for Classifying Pathogenic Variants Laboratory 1 Laboratory 2 Laboratory 3 Laboratory 4 CHEK2 p.S428F Functional studies, from literature review, x x x x* supportive of a damaging effect(very strong evidence) Increased prevalence in affected individuals v x x x x† control subjects (strong evidence) Location in a well-established functional x x x x domain (moderate evidence) Cosegregation with disease in multiple N/A x x N/A affected family members (supporting evidence) Prediction by in silico algorithms (supporting x N/A N/A N/A evidence) Contribution to tumorigenesis determined by N/A x N/A N/A LOH Final interpretation Pathogenic Pathogenic Pathogenic Uncertain clinical (low penetrance) significance CHEK2 p.I157K Functional studies, from literature review, x x‡ x x‡ supportive of a damaging effect (very strong evidence) Increased prevalence in affected individuals v x§ x x§ x§ control subjects (strong evidence) Location in a well-established functional x x x x domain (moderate evidence) Cosegregation with disease in multiple x§ N/A x¶ N/A affected family members (supporting evidence) Prediction by in silico algorithms (supporting Inconsistent N/A Limited value N/A evidence) Contribution to tumorigenesis determined by N/A N/A N/A N/A LOH Final interpretation Expected Pathogenic Pathogenic Uncertain clinical pathogenic (low penetrance) significance Abbreviations: LOH, loss of heterozygosity; N/A, not applicable. *Interpretation of variable impairment. †Interpretation of increased risk only in specific populations. ‡No effect on CHEK2 protein kinase activity, and relationship between experimental findings and cancer association is unclear. §Observed at high frequencies among controls in some particular populations. ¶Incomplete segregation. www.jco.org © 2016 by American Society of Clinical Oncology 4077 Balmaña et al both crowd-sourcing and direct participant enrollment; therefore, it collect data to inform efforts to standardize classification of genetic also can provide a platform for cohort formation and prospective variants, and can play an important role in efforts to minimize follow-up of individuals and their family members harboring genetic potential medical harms due to false alarm or false reassurance variants. However, at this time, health-care providers and patients following cancer genetic testing. need to be aware that there could be conflicting interpretations of variants and those variants may be reclassified. Our analysis has limitations. Because PROMPT is an elective, AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS patient-oriented registry, there may be significant ascertainment OF INTEREST bias. For example, individuals with VUS or those who were self- Disclosures provided by the authors are available with this article at aware of a variant with conflicting interpretation may have been www.jco.org. more likely to report their variant to PROMPT, leading to an overestimation of discrepant findings. On the other hand, a sig- nificant number of test results reported by participants were not AUTHOR CONTRIBUTIONS submitted to ClinVar by any laboratory and, therefore, were ex- cluded from the primary analysis, potentially leading to a mis- Conception and design: Judith Balmaña, Mark E. Robson, Susan M. estimate of conflicting variant interpretations. Finally, because Domchek some patients did not upload their clinical test report, it was not Collection and assembly of data: Judith Balmaña, Laura Digiovanni, possible to carry out more in-depth analysis in many cases. Efforts Pragna Gaddam, Judy E. Garber, Kenneth Offit, Mark E. Robson, Susan M. to obtain reports from all participants are ongoing. Domchek In conclusion, clinical interpretation of genetic testing for Data analysis and interpretation: Judith Balmaña, Laura Digiovanni, Michael F. Walsh, Vijai Joseph, Zsofia K. Stadler, Katherine L. Nathanson, increased cancer susceptibility as assessed by multiplex panels Fergus J. Couch, Kenneth Offit, Mark E. Robson, Susan M. Domchek hinges on accurate curation and interpretation of variants. Dis- Manuscript writing: All authors crepant interpretation of some genetic variants appears to be Final approval of manuscript: All authors common. Internet-based registries provide a powerful tool to Accountable for all aspects of the work: All authors 8. Landrum MJ, Lee JM, Benson M, et al: 17. Stoffel EM, Fearon ER: Germline sequence REFERENCES ClinVar: Public archive of interpretations of clinically variants and ovarian cancer: Known-knowns and relevant variants. Nucleic Acids Res, 44:D862-D868, known-unknowns. JAMA Oncol 2:491-492, 2016 1. Easton DF, Pharoah PD, Antoniou AC, et al: 2016 18. Cantor SB, Bell DW, Ganesan S, et al: BACH1, Gene-panel sequencing and the prediction of breast- 9. Desrichard A, Bidet Y, Uhrhammer N, et al: a novel helicase-like protein, interacts directly with cancer risk. N Engl J Med 372:2243-2257, 2015 CHEK2 contribution to hereditary breast cancer in BRCA1 and contributes to its DNA repair function. 2. Rehm HL, Berg JS, Brooks LD, et al: ClinGen: non-BRCA families. Breast Cancer Res 13:R119, Cell 105:149-160, 2001 the Clinical Genome Resource. N Engl J Med 372: 2011 19. Seal S, Thompson D, Renwick A, et al: Trun- 2235-2242, 2015 10. Liu C, Wang Y, Wang QS, et al: The CHEK2 cating mutations in the Fanconi anemia J gene BRIP1 3. Pepin MG, Murray ML, Bailey S, et al: The I157T variant and breast cancer susceptibility: A are low-penetrance breast cancer susceptibility al- challenge of comprehensive and consistent se- systematic review and meta-analysis. Asian Pac J leles. Nat Genet 38:1239-1241, 2006 quence variant interpretation between clinical labo- Cancer Prev 13:1355-1360, 2012 20. Ramus SJ, Song H, Dicks E, et al: Germline ratories. Genet Med 18:20-24, 2016 11. Han FF, Guo CL, Liu LH: The effect of CHEK2 mutations in the BRIP1, BARD1, PALB2, and NBN 4. Jurgens J, Ling H, Hetrick K, et al: Assess- variant I157T on cancer susceptibility: Evidence from genes in women with ovarian cancer. J Natl Cancer ment of incidental findings in 232 whole-exome se- a meta-analysis. DNA Cell Biol 32:329-335, 2013 Inst 107:djv214, 2015 quences from the Baylor-Hopkins Center for 12. Seppälä EH, Ikonen T, Mononen N, et al: 21. Harrison S, Dolinsky J, Vincent L, et al. Clinical Mendelian Genomics. Genet Med 17:782-788, 2015 CHEK2 variants associate with hereditary prostate laboratories collaborate to resolve variant in- 5. Plon SE, Eccles DM, Easton D, et al: Sequence cancer. Br J Cancer 89:1966-1970, 2003 terpretation differences in ClinVar. [abstract 2093T]. variant classification and reporting: Recommenda- 13. Cybulski C, Górski B, Huzarski T, et al: CHEK2 Baltimore, MD, American Society of Human Genet- tions for improving the interpretation of cancer sus- is a multiorgan cancer susceptibility gene. Am J Hum ics American Society of Human Genetics, 2015. ceptibility genetic test results. Hum Mutat 29: Genet 75:1131-1135, 2004 22. Thompson BA, Spurdle AB, Plazzer JP, et al: 1282-1291, 2008 14. Falck J, Mailand N, Syljuåsen RG, et al: The Application of a 5-tiered scheme for standardized 6. Richards S, Aziz N, Bale S, et al: Standards and ATM-Chk2-Cdc25A checkpoint pathway guards classification of 2,360 unique mismatch repair gene guidelines for the interpretation of sequence variants: against radioresistant DNA synthesis. Nature 410: variants in the InSiGHT locus-specific database. Nat A joint consensus recommendation of the American 842-847, 2001 Genet 46:107-115, 2014 College of Medical Genetics and Genomics and the 15. Schwarz JK, Lovly CM, Piwnica-Worms H: 23. Eccles DM, Mitchell G, Monteiro AN, et al: Association for Molecular Pathology. Genet Med 17: Regulation of the Chk2 protein kinase by oligomerization- BRCA1 and BRCA2 genetic testing-pitfalls and rec- 405-424, 2015 mediated cis- and trans-phosphorylation. Mol Cancer ommendations for managing variants of uncertain 7. Landrum MJ, Lee JM, Riley GR, et al: ClinVar: Res 1:598-609, 2003 clinical significance. Ann Oncol 26:2057-2065, 2015 Public archive of relationships among sequence 16. Roeb W, Higgins J, King MC: Response to DNA 24. Fokkema IF, Taschner PE, Schaafsma GC, variation and human phenotype. Nucleic Acids Res damage of CHEK2 missense mutations in familial breast et al: LOVD v.2.0: The next generation in gene variant 42:D980-D985, 2014 cancer. Hum Mol Genet 21:2738-2744, 2012 databases. Hum Mutat 32:557-563, 2011 n n n 4078 © 2016 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY Conflicting Interpretations in Commercial Panel Testing AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST Conflicting Interpretation of Genetic Variants and Cancer Risk by Commercial Laboratories as Assessed by the Prospective Registry of Multiplex Testing The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO’s conflict of interest policy, please refer to www.asco.org/rwc or jco.ascopubs.org/site/ifc. Judith Balmaña Judy E. Garber No relationship to disclose Consulting or Advisory Role: Pfizer, Sequenom, Novartis (I), Pfizer (I), SV Life Sciences (I) Laura Digiovanni Research Funding: Ambry Genetics, Myriad Genetics, Novartis (I), No relationship to disclose Pfizer (I) Pragna Gaddam Fergus J. Couch No relationship to disclose Travel, Accommodations, Expenses: Ambry Genetics Michael F. Walsh Kenneth Offit No relationship to disclose No relationship to disclose Vijai Joseph Mark E. Robson No relationship to disclose Honoraria: AstraZeneca Zsofia K. Stadler Consulting or Advisory Role: Bayer, Pfizer, McKesson Stock or Other Ownership: Alimera Sciences (I), Avalanche Research Funding: AstraZeneca (Inst), AbbVie (Inst), Myriad Genetics Biotechnologies (I) (Inst), Biomarin (Inst) Consulting or Advisory Role: Alimera Sciences (I), Avalanche Travel, Accommodations, Expenses: AstraZeneca, Biomarin Biotechnologies (I), Allergan (I), Bausch and Lomb (I), Roche (I), Susan M. Domchek Regeneron (I), Optos (I) Honoraria: EMD Serono Katherine L. Nathanson Research Funding: AstraZeneca (Inst), Clovis Oncology (Inst), No relationship to disclose Abbvie (Inst), Pharmamar (Inst) www.jco.org © 2016 by American Society of Clinical Oncology Balmaña et al Acknowledgment J.B. is a recipient of an award by Sociedad Española de Oncologı́a Médica. This project was supported also by the Robert and Kate Niehaus Center for Inherited Cancer Genomics at Memorial Sloan Kettering Cancer Center (M.E.R., M.F.W., K.O.), the Breast Cancer Research Foundation (S.M.D., M.E.R., K.O., F.C., J.G.), and the Susan G. Komen Foundation (S.M.D.). We thank Myriad, Ambry, Invitae, GeneDx, Color Genomics, Pathway Genomics, and Quest Diagnostics for their active collaboration with PROMPT Appendix ATM BRIP1 33% 67% 100% CHEK2 NBN 12% 43% 57% 88% PALB2 (Likely) Pathogenic v VUS (Likely) Pathogenic v VUS v (Likely) Benign 33% VUS v Benign 67% Fig A1. Type of conflicting interpretation by genes. © 2016 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY