Prevalence of Variant Reclassification Following Hereditary Cancer Genetic Testing

Key Points

Question  How often and for what types of variant reclassifications are amended reports issued as part of hereditary cancer genetic testing?

Findings  In this retrospective cohort study that included 1.45 million individuals and 1.67 million initial tests, 59 955 amended reports were issued due to variant reclassification. Among variants initially classified as uncertain significance, 7.7% were reclassified, of which 91.2% were downgraded to less severe classifications and 8.7% were upgraded to more severe classifications. Reclassification of variants initially classified as pathogenic or benign was rare.

Meaning  This study provides an estimate of the likelihood of variant reclassification following hereditary cancer genetic testing, but replication is required using other data sources.

Importance  Variant reclassification is an important component of hereditary cancer genetic testing; however, there are few published data quantifying the prevalence of reclassification.

Objective  Retrospective cohort study of individuals who had genetic testing from 2006 through 2016 at a single commercial laboratory.

Design, Setting, and Participants  A retrospective cohort of individuals who had genetic testing between 2006 and 2016 at a single commercial laboratory was assessed. Variants were classified as benign, likely benign, variant of uncertain significance, likely pathogenic, or pathogenic. Retrospective chart reviews were conducted for patients from the University of Texas Southwestern (UTSW) Medical Center.

Exposures  Hereditary cancer genetic testing.

Main Outcomes and Measures  Frequency of and time to amended reports; frequency and types of variant reclassification.

Results  From 2006 through 2018, 1.45 million individuals (median [interquartile range] age at testing, 49 years [40.69-58.31 years], 95.6% women) had genetic testing, and 56.6% (n = 821 724) had a personal history of cancer. A total of 1.67 million initial tests were reported and 59 955 amended reports were issued due to variant reclassification. Overall, 6.4% (2868 of 44 777) of unique variants were reclassified. Reclassification to a different clinical category was rare among unique variants initially classified as pathogenic or likely pathogenic (0.7%, 61 of 9112) or benign or likely benign (0.2%, 15 of 8995). However, 7.7% (2048 of 26 670) of unique variants of uncertain significance were reclassified: 91.2% (1867 of 2048) were downgraded to benign or likely benign (median time to amended report, 1.17 years), 8.7% (178 of 2048) were upgraded to pathogenic or likely pathogenic variants (median time to amended report, 1.86 years). Because most variants were observed in more than 1 individual, 24.9% (46 890 of 184 327) of all reported variants of uncertain significance were reclassified.

Conclusions and Relevance  Following hereditary cancer genetic testing at a single commercial laboratory, 24.9% of variants of uncertain significance were reclassified, which included both downgrades and upgrades. Further research is needed to assess generalizability of the findings for other laboratories, as well as the clinical consequences of the reclassification as a component of a genetic testing program.

Genetic testing results play a substantial role in medical management of hereditary cancers and may inform screening, surgery recommendations, and treatment decisions.^1,2 According to the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) guidelines and standards for variant classification, initial classifications should be based on all available information regarding variant pathogenicity, including population frequency, functional data, segregation analysis, and phenotype analysis.³ With the expanding use of genetic testing for hereditary cancer risk, the information available regarding variant pathogenicity is constantly evolving.^4-9 As such, genetic test results are based on the best scientific information at a given moment, which may change as scientific knowledge evolves.

Given the clinical implications of genetic testing, accurate and timely variant classification is increasingly important for appropriate long-term patient care. It is, therefore, critical that new information for variants in cancer-risk genes be reviewed by testing laboratories in order to evaluate whether reclassification is appropriate. This is done in an effort to ensure that patients with increased cancer risk receive appropriate screening, risk-reducing surgical interventions, or both while those with lower risk avoid unnecessary interventions.

Despite the clinical importance of variant reclassification, there is little published data to aid in clinical discussions on this topic.¹⁰ As a result, clinicians are faced with counseling patients and managing their disease with limited information about the frequency and clinical implications of variant reclassification. In this study, variant reclassifications and amended test reports for hereditary cancer testing at a single laboratory were assessed over a 10-year period. Specifically, the number and types of variant reclassifications, as well as the median time between the initial and amended report were evaluated. A subset of patients from a single academic center was further analyzed to review additional clinical follow-up.

Quiz Ref IDThis analysis included all individuals who had genetic testing for hereditary cancer risk in the course of routine practice and received an initial single-syndrome test report, pan-cancer panel test report, or both between 2006 and 2016 (see eTable 1 in Supplement 1 for test genes and their National Center for Biotechnology [NCBI] accession numbers). If individuals had more than 1 of these tests, then results for all completed tests were included. Patients who received single-variant or founder-mutation testing were not included in these analyses. Demographic information for all tested individuals was obtained from clinician-completed test request forms.

All genetic testing was performed by a single laboratory (Myriad Genetic Laboratories Inc). During the 10-year period of this analysis, the laboratory methods for genetic testing included the use of Next Generation Sequencing (NGS) (2013-2016), Sanger sequencing (2006-2016), or both to identify sequence variants. Large rearrangements were detected with NGS dosage analysis (2013-2016), microarray-based comparative genomic hybridization (2012-2016), multiplex ligation–dependent probe amplification analysis (2011-2016), Southern blot analysis (2006-2012), quantitative polymerase chain reaction (PCR) analysis (2007-2016), or long-range PCR analysis (2007-2016).

Chart review was performed only for the subset of patients seen at the University of Texas Southwestern (UTSW) Medical Center with institutional review board (IRB) oversight and waived patient consent. This included patients who received an amended report with a different test result (negative or positive) from the original test report. The information collected included cancer history and self-reported family history at the time of testing and at reclassification, genetic test results, and cancer risk management choices, and familial cascade testing. Such clinical outcome data were only available for the UTSW subset. The UTSW IRB determined that analysis of the remainder of the deidentified full cohort did not meet the definition of human subject research as defined by Title 45 of the Federal Regulations (45 CFR 46.102) because none of the investigators could reidentify these subjects.

The overall process for variant classification and reclassification is summarized in the eMaterials and eFigure 1 in Supplement 1. Although the specific methods used by the testing laboratory for variant classification evolved during the course of this analysis as new information and tools became available, the same fundamental steps were involved. During the full study period, variants were classified using a 5-tier classification system, as previously described.⁵ This classification system is similar to the current ACMG/AMP guidelines.³ In order of increasing severity, the classification categories were benign or polymorphism, likely benign or favor polymorphism, variant of uncertain significance, likely pathogenic or suspected deleterious, and pathogenic or deleterious. A small number of variants did not fit any of these categories and were classified as special interpretation (see eMaterial in Supplement 1 for detailed explanation). Test reports were sent to clinicians based on variant classifications at the time of testing. Reports were positive if they contained at least 1 pathogenic or likely pathogenic variant and were negative if they contained no variants that were pathogenic or likely pathogenic. When multiple variants were identified for an individual, the test report interpretation was based on the most clinically severe classification (eFigure 1in Supplement 1).

Quiz Ref IDThe availability of new evidence in the literature or methods specific to the testing laboratory (ie, statistical methods, functional data) was monitored daily by an automated system. Variant classification was reevaluated immediately upon the identification of additional information, and all new and existing evidence was reviewed. If variant reclassification was appropriate, the testing laboratory sent an amended report indicating the new classification to the clinician (eFigure 1 in Supplement 1). Amended reports included more than 1 reclassified variant if multiple variants were reclassified in close proximity. In some cases, amended reports were sent for reasons other than a variant reclassification (ie, a change in clinician, correction of a typographical error, or to highlight new gene-specific considerations). This included reclassification to report on new evidence suggesting that a proportion of apparent germline variants in TP53 were instead somatic mosaic variants, which do not carry the same cancer risks as germline variants.^11,12 Following this discovery, amended reports were issued for individuals with a pathogenic or likely pathogenic variant in TP53 to indicate that follow-up testing is necessary to determine whether the variant was truly in the germline. Only amended reports issued due to variant reclassification or gene-specific considerations were included.

This was a descriptive study. Cohort demographics and clinical information were summarized using descriptive statistics based on clinician-completed test request forms. This included demographic information (ie, age, sex), personal and family cancer history, and self-reported ancestry. Patients who were missing information were included in the analysis and noted in a separate category.

Reported variant classifications, reclassifications, and amended reports were evaluated. Variant reclassifications were considered downgrades if the variant was reclassified to a less severe category and upgrades if the variant was reclassified to a more severe category. Because the medical management for variants classified as special interpretation was complex, a reclassification to or from special interpretation did not definitively change the clinical category to more or less severe. Therefore, these reclassifications were assessed separately.

Unique variants were defined herein as a specific genetic variant. Total reported variants were defined as the total number of observed variants, which included multiple observations of the same unique variant. Because most variants were detected in more than 1 individual over the course of laboratory testing, analyses were performed for the number of unique variants as well as the total reported variants. Variant frequency was assessed as the total reported variants divided by the unique variants. The median and interquartile ranges (IQRs) for time to the amended report were calculated using the initial report issue date and the amended report issue date. In addition, the time to the amended report was assessed according to the year of the initial test report. Because all analyses were performed using laboratory data for variant classification and reporting, there were no missing data. All analyses were conducted using R version 3.4.1. Boxplot graphics were created using the R package ggplot2.

All analyses were performed for the subset of patients seen at UTSW Medical Center to assess a single institution’s experience and clinical follow-up after variant reclassification. The UTSW group was a subset of the full cohort and overlapped entirely with this patient population. Statistical comparisons between the UTSW subset and the full cohort were not possible due to overlap between the cohorts. Limited statistical analyses were performed for the exclusive cohorts (full cohort excluding the UTSW subset vs UTSW subset) to highlight clinically meaningful differences in patient demographics. Analyses were performed using Fisher exact test with Monte Carlo simulations (n = 10 000). P values <.05 were considered significant.

Between 2006 and 2016, genetic testing for hereditary cancer risk was performed for 1.45 million individuals. Overall, 95.6% (n = 1.39 million) of tested individuals were women, 51.9% (n = 752 793) indicated European ancestry, and the median age at testing was 49 years (IQR, 40.69-58.31; Table 1). At the time of testing, 56.6% (n = 821 724) of individuals reported a personal history of cancer (Table 1).

The subset of patients seen at UTSW (n = 8427) showed some differences from the full cohort, including differences in ancestry and personal cancer history (Table 1). Relative to the full cohort, a higher proportion of individuals seen at UTSW reported a Latin American or Caribbean ancestry (21.0% [1773 of 8427] vs 6.5% [94 918 of 1.45 million) or African ancestry (15.1% [1271 of 8427] vs 5.6% [81 616 of 1.45 million]). In addition, a higher proportion of patients from UTSW had a personal history of cancer at the time of testing, relative to the overall cohort (71.5% [6022 of 8427]) vs 56.6% [821 724 of 1.45 million]; Table 1). When the exclusive cohorts were assessed, these differences were statistically significant (P < .001, eResults in Supplement 1).

In this 10-year period, 1.67 million initial genetic test reports were issued (Table 2). Overall, 5.4% (n = 90 052) of all test reports were positive and 5.8% (n = 96 684) were negative with 1 or more variants of uncertain significance of 1 or more (Table 2). The distribution of variants of uncertain significance by gene is shown in eTable 2 in Supplement 1. The majority (88.7%, n = 1.48 million) of reports were negative based on the identification of only benign or likely benign variants (Table 2). Only 0.1% (n = 1841) of initial results were reported as special interpretation. Test results according to test type (panel or single-syndrome) are shown in Table 2. Within the UTSW subset, 9493 initial test reports were issued, of which 6.9% (n = 658) of patients had a positive test result and 9.4% (n = 897) had a negative test result with 1 or more variants of uncertain significance of 1 or more (Table 2). The distribution of variants of uncertain significance by gene within the UTSW subset is shown in eTable 3 in Supplement 1.

A total of 60 064 amended reports were issued in the full cohort (Table 2), 99.8% (n = 59 955) of which were due to the reclassification of at least 1 variant. Similarly, 99.5% (579 of 582) of amended reports issued in the UTSW subset were due to variant reclassification. For amended reports issued due to the reclassification of 1 or more variants, the Figure shows the time to the amended report as a function of the year the variant was initially reported. The median time to the amended report was highest in 2006 at 2.55 years. For variants reported between 2007 and 2014, the median time to the amended report fluctuated between 1.01 and 1.97 years. The median time to the amended report dropped to 0.77 years for variants initially reported in 2015 and 0.34 years for 2016. The substantial overlap in IQRs for the time to amended report regardless of when the initial report was issued indicates the time to reclassification did not decrease over time; however, there are caveats to consider such as the time available for reclassification would not be sufficient to catch the longer times in the later years and different genes tested and different testing volumes per year confound a simple analysis of these data. When the genes associated with hereditary breast and ovarian cancer syndrome (ie BRCA1, BRCA2) and Lynch syndrome (MLH1, MSH2, MSH6, PMS2, EPCAM) were analyzed, for which genetic testing was available for the full study period, the data were similar (eFigures 2 and 3 in Supplement 1).

Overall, 44 777 unique variants were detected by the testing laboratory. Because many variants were observed in multiple individuals, these unique variants were observed a total of 6.22 million times. This corresponded to a variant frequency of 138.9 observations per variant. In the UTSW subset, 3158 unique variants were detected a total of 187 033 times (variant frequency of 59.2 observations per variant). In this period, 6.4% (2861 of 44 777) of all unique variants identified were reclassified following review of additional information by the testing laboratory. Variant reclassification details are given by gene for both single-syndrome and pan-cancer panel testing in eTables 4 through 7 in Supplement 1, which includes detailed reclassification categories according to the initial classification and reclassification. The median time to the amended report due to variant reclassification was 1.10 years (IQR, 0.54-2.06 years). In the UTSW subset, 9.1% (287 of 3158) of unique variants were reclassified with a median time to the amended report of 1.06 years (IQR, 0.56-1.73 years).

In the full cohort, 8995 unique benign or likely benign variants were observed a total of 5.94 million times (variant frequency of 660.8 observations per variant; Table 3). Overall, 6.0% (n = 542) of unique variants initially classified as benign or likely benign were reclassified, nearly all of which were within the same clinical category (likely benign to benign). This type of reclassification occurred for 5.9% (527 of 8995) of unique and 0.1% (8099 of 5.94 million) of all detected benign to likely benign variants (Table 3). The remaining reclassifications were upgrades from benign or likely benign to variants of uncertain significance. However, this type of reclassification was extremely rare and accounted for only 0.2% (15 of 8995) of unique and less than 0.1% (34 of 5.94 million) of all detected benign to likely benign variants (Table 3). Data from the UTSW subset was similar, and the majority of benign or likely benign variant reclassifications occurred within the same clinical category (Table 3).

Pathogenic or likely pathogenic variants were rarer than benign or likely benign variants; 9112 unique pathogenic or likely pathogenic variants were observed a total of 91 685 times in the full cohort (variant frequency of 10.1 observations per variant; Table 3). Among variants initially classified as pathogenic or likely pathogenic, 3.1% (n = 278) of unique variants in the full cohort were reclassified (Table 3). Classification changes within the same clinical category (ie, likely pathogenic to pathogenic) occurred for 2.2% (202 of 9112) of unique and 1.1% (1503 of 91 685) of all reported pathogenic or likely pathogenic variants. Changes to a different clinical category were rare: only 0.7% (61 of 9112) of unique and 0.3% (278 of 91 685) of all reported pathogenic or likely pathogenic variants were downgraded to variants of uncertain significance (Table 3). Sixteen pathogenic or likely pathogenic variants were reclassified to special interpretation.

In the UTSW subset, 2.8% (12 of 427) of unique and 1.8% (12 of 671) of all reported pathogenic or likely pathogenic variants were reclassified in this period (Table 3). This included only 3 pathogenic or likely pathogenic variants that were downgraded to variants of uncertain significance (0.7% [3 of 427] of unique and 0.4% [3 of 671] of all reported pathogenic or likely pathogenic variants; Table 3).

In the full cohort, 26 670 unique variants of uncertain significance were initially detected a total of 184 327 times (Table 4). This corresponds with a variant frequency of 6.9 observations per variant, which was less frequent than both benign or likely benign and pathogenic or likely pathogenic variants. In this period, 7.7% (2048 of 26 670) of unique variants of uncertain significance were reclassified, affecting 24.9% (46 890 of 184 327) of all reported variants of uncertain significance. The majority of reclassified variants of uncertain significance were downgraded to benign or likely benign, accounting for 91.2% (1867 of 2048) of unique and 97.0% (44 509 of 46 890) of all reported reclassified variants of uncertain significance (Table 4). The median time to the amended report for downgraded variants of uncertain significance was 1.17 years (IQR, 0.58-2.20 years).

In the UTSW subset, 1189 unique variants of uncertain significance were detected a total of 1945 times (variant frequency of 1.6 observations per variant). A higher proportion of variants of uncertain significance were unique and therefore reclassified in the UTSW subset (19.6%, 233 of 1189). The majority of these reclassifications were downgrades to benign or likely benign (Table 4), with a median time to the amended report of 1.19 years (IQR, 0.93-3.40 years).

A smaller proportion of reclassified variants of uncertain significance were upgraded to pathogenic or likely pathogenic. In the full cohort, upgrades accounted for 8.7% (178 of 2048) of unique and 3.0% (1372 of 46 890) of all reported variants of uncertain significance that were reclassified (Table 4). The median time to an amended report for upgraded variants of uncertain significance was 1.86 years (IQR, 0.85-3.28 years). In the UTSW subset, 11 variants of uncertain significance were upgraded to pathogenic or likely pathogenic, accounting for 4.7% of unique and 2.0% of all reclassified variants of uncertain significance (Table 4). The median time to the amended report for these 11 variant of uncertain significance upgrades was 0.81 years (IQR, 0.35-2.00 years).

In the UTSW subset, 3 variants in BRCA1, TP53, and BRIP1 were downgraded from pathogenic or likely pathogenic to variants of uncertain significance. The BRCA1 variant (dup exons 1-22) was identified in a woman with unilateral breast cancer at age 58 years. Bilateral mastectomy was elected following the receipt of her test results (Table 5; eTable 8 in Supplement 1). The TP53 variant (c.542G>A [p.Arg181His]) was identified in a patient with breast cancer at age 39 years (Table 5; eTable 8 in Supplement 1). This patient pursued genetic testing at the time of diagnosis and chose bilateral mastectomy because of the TP53 variant. The BRIP1 variant (c.2992_2993del [p.Lys998Glu*3]) was identified in a woman with a family history of breast cancer (no ovarian cancer) who was unaffected with cancer at the time of testing (age 58 years; Table 5). She continued with high-risk breast cancer surveillance after the BRIP1 variant reclassification due to her family history of breast cancer.

In the UTSW subset, variants in BRCA1 (n = 4), BRCA2 (n = 3), CDH1 (n = 1), CHEK2 (n = 1), MLH1 (n = 1), and MSH6 (n = 1) were upgraded from variants of uncertain significance to pathogenic or likely pathogenic (Table 5). No known interim cancers were diagnosed in the patients prior to reclassification (eTable 8 in Supplement 1). Nine of the 11 individuals with upgraded variants of uncertain significance had a personal history of breast (n = 7) or colon cancer (n = 2). One 28-year-old with unilateral breast cancer chose bilateral mastectomy at the time of diagnosis due to her young age. All other individuals with breast cancer pursued lumpectomy or unilateral mastectomy at the time of diagnosis. One patient with breast cancer was also diagnosed with platinum-taxane–resistant ovarian cancer. The upgrade of the BRCA1 variant qualified this patient for olaparib therapy. Similarly, the 2 patients with colon cancer may now qualify for recently approved immunotherapy (eg, nivolumab) if their disease recurs due to the upgrade of their MLH1 and MSH6 variants of uncertain significance. Additional details regarding the reason for these reclassifications are given in the eResults section and eTable 8 in Supplement 1.

Quiz Ref IDIn this study, variant classification and reclassification were evaluated for 1.45 million individuals who received genetic testing from a single testing laboratory over a 10-year period. Overall, 6.4% of unique variants were reclassified in this period. Very few variants initially classified as benign or pathogenic were reclassified to a different clinical category. This indicates that once a variant reached a definitive classification (benign or pathogenic), it is unlikely that classification would change when a comprehensive classification program is used. However, variant reclassification in these cases may affect patient care, so it is important for patients to be aware that, while unlikely, this type of reclassification is a possibility.

The majority of reclassified variants of uncertain significance were downgraded to benign, which is consistent with other studies on variant reclassification.¹⁰ Although this type of reclassification should not affect medical management, downgrades may have clinical importance. Previous studies have demonstrated variability in clinical practice regarding patients with a variant of uncertain significance that may include increased surgical intervention,¹³ screening,¹⁴ and anxiety.¹⁵ Although improved education for patients and clinicians will be an important factor in reducing inappropriate management, variant reclassification is another contributing factor. Reclassification from variant of uncertain significance to benign in a timely manner may help minimize the risk of inappropriate management and unnecessary anxiety.

Because gene-based medical management recommendations pertain only to pathogenic variants in cancer-risk genes,^1,16 variant reclassification to or from pathogenic is of particular clinical importance. In this study, 8.7% of unique and 3.0% of total reclassified variants of uncertain significance were upgraded to pathogenic or likely pathogenic. This is higher than a recent study on variant reclassifications as part of genetic testing at a single clinic over a 3.5-year period.¹⁰ In that study, 3.3% (1 of 30) of reclassified variants of uncertain significance were upgraded to likely pathogenic.¹⁰ Even when only panel testing is considered for comparison, about 8.0% of unique reclassified variants in the present study were upgraded from variants of uncertain significance to pathogenic or likely pathogenic. This suggests that the rate of variant reclassification may be laboratory dependent.

Quiz Ref IDThe overall upgrade rate among all reported variants of uncertain significance was low at 3%; however, in our opinion, it is not possible to predict with clinical accuracy which variants will be upgraded to pathogenic. In addition, the number of individuals with variants of uncertain significance will likely continue to rise nationally as (1) genetic awareness increases leading to more individuals being tested, (2) disease gene panel adoption rises, (3) the number of genes included in testing increases, and (4) the cost of genetic testing decreases. As such, the absolute number of individuals with variants of uncertain significance that are later upgraded to pathogenic will continue to rise. In addition, the median time from initial to amended report was less than 2 years for variants of uncertain significance that were upgraded to pathogenic or likely pathogenic in this study. Although this is longer than for downgrades to benign or likely benign, it is consistent with the need for more time to obtain information to validate pathogenicity of variants.

The apparent decrease in median time to the amended report reflects the decreased amount of time available for variant reclassification with the increasing year. For example, the upper limit of the time to reclassification decreased as the year increased, reflecting this inherent bias. There are other complex interactions that could influence these data including, but not limited to, differences in test volume in each year, new data that affect classes of variants, and how many genes were available for testing each year. Collectively, this highlights the continued importance of an efficient and accurate reclassification program to ensure up-to-date clinical management to reduce hereditary cancer risk.

This study has several limitations. First, it was observational and is specific to the testing population for a single commercial testing laboratory. Furthermore, statistical comparisons between the UTSW subset and the full testing cohort could only be made for the exclusive cohorts (full cohort excluding UTSW). However, this analysis showed that there was a higher proportion of individuals who reported a Latin American or Caribbean ancestry or African ancestry at UTSW, which is likely related to geographic location. These ancestry-based differences likely accounted for the higher rate of variants of uncertain significance observed in the UTSW subset, because previous studies have shown higher variant of uncertain significance rates among individuals of Latin American or Caribbean or African ancestry.^5,17-19 Second, a higher fraction of the UTSW patient subset had a personal history of cancer at the time of testing relative to the overall cohort, likely due to the academic setting. The higher positive rate in the UTSW subset was, in turn, likely related to this higher incidence of cancer (ie, higher risk population).

Quiz Ref IDThird, the study population is not representative of the general population, because patients underwent hereditary cancer testing based on clinical suspicion of hereditary cancer syndromes. Fourth, testing indications were weighted for breast and ovarian cancer syndrome risks. Fifth, the time to the amended report is inherently biased toward shorter times, especially for variants that were initially identified toward the end of the analysis period. This is demonstrated by the expected decrease in median time to amended report to less than 1 year for variants initially reported in 2015 and 2016.

Sixth, the laboratory methods as well as the accumulation of knowledge helpful to accurate variant interpretation for testing and variant classification evolved over the 10-year period of this analysis. As such, the data presented represent the aggregate effects of the laboratory methods. Seventh, details for the reasons for variant reclassification and clinical follow-up for patients within the full cohort were not available. Eighth, a higher proportion of individuals in the full cohort did not specify any ancestry compared with the UTSW subset, which may be reflective of variation in the amount of detail provided by clinicians on the test request forms. This highlights the important role of clinicians who provide complete and accurate information on the test request forms.

Following hereditary cancer genetic testing at a single commercial laboratory, 24.9% of variants of uncertain significance were reclassified, which included both downgrades and upgrades. Further research is needed to assess generalizability of the findings for other laboratories, as well as the clinical consequences of the reclassification as a component of a genetic testing program.

Corresponding Author: Theodora Ross, MD, PhD, University of Texas Southwestern Medical Center, ND3.214, 5323 Harry Hines Blvd, Dallas, TX 75390 (Theo.Ross@utsw.edu).

Accepted for Publication: August 17, 2018.

Author Contributions: Dr Ross and Ms Mundt had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Mersch, N. Brown, Mundt, Esterling, Manley, Ross.

Acquisition, analysis, or interpretation of data: Mersch, N. Brown, Pirzadeh-Miller, Mundt, Cox, K. Brown, Aston, Ross.

Drafting of the manuscript: Mersch, Pirzadeh-Miller, Mundt, K. Brown, Ross.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Mersch, Cox, Aston.

Obtained funding: Ross.

Administrative, technical, or material support: Mersch, N. Brown, Pirzadeh-Miller, K. Brown.

Supervision: Mersch, Pirzadeh-Miller, Esterling, Manley, Ross.

Conflict of Interest Disclosures: All the authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Drs Cox, K. Brown, and Esterling and Mss Mundt, Aston, and Manley were employed by Myriad Genetic Laboratories at the time of this study. Ms N. Brown was employed by UT Southwestern at the time of the study, but is now employed by Myriad Genetic Laboratories. No other disclosures were reported.

Funding/Support: This work was supported by grant R01 HL132251-02 from the National Institutes of Health, the Lucy and Henry Billingsley Fund, and a Burroughs Wellcome Fund Clinical Scientist Award in Translational Research (Dr Ross). to Dr Ross holds the Jeanne Ann Plitt Professorship in Breast Cancer Research and the H. Ben and Isabelle T. Decherd Chair in Internal Medicine at UTSW Medical Center.

Role of the Funder/Sponsor: The National Institutes of Health, the Lucy and Henry Billingsley Fund, and the Burroughs Wellcome Fund had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Myriad Genetics provided no funding for this study and did not contribute to the decision to publish these data or to control its content.

Data Sharing Statement: See Supplement 2.

Additional Contributions: We thank the entire UTSW cancer genetics team of genetic counselors and assistants for their patient care and intellectual contributions to this work. We also thank Aaron Parker, (BA, SAS certified data analyst) from Moncrief Cancer Institute/UTSW Medical Center for his assistance in data comparison efforts. Mr Parker was not compensated for this work.