Multiple Primary Cancers in Patients Undergoing Tumor-Normal Sequencing Define Novel Associations

This article is featured in Highlights of This Issue, p. 303

A substantial proportion of cancers (18%) occur in patients who have already had prior malignancies, and 10% to 25% of patients with cancer subsequently develop multiple primary cancers (MPC; refs. 1, 2). Occurrence of MPC is increasing as treatments improve and cancer becomes more common in elderly patients (1, 3). Rates of MPC vary across tumor types and reflect environmental exposures, including smoking, alcohol, obesity, and prior cancer treatments, and underlying genetic factors, including known inherited, cancer syndromes (4).

Although emphasis has been placed on public health efforts to reduce risk of subsequent cancers (5), treatment related or exogenous exposures do not fully explain the risk for MPCs (6), and both known and unknown underlying genetic factors contribute to cancer risk. As concurrent sequencing of tumor and germline DNA (parallel sequencing; ref. 7) is becoming more common in cancer care, there are opportunities to study the inherited and somatic genetics underlying MPCs to define novel susceptibilities associated with specific tumor combinations (2, 8).

We sought to comprehensively characterize the clinical and genetic features of patients with MPCs who underwent massively parallel sequencing. We utilized epidemiologic methodologies to define significant co-incidence patterns of MPC and analyzed these findings with germline and somatic DNA sequencing for known cancer predisposition genes. In addition to informing clinical and preventive management, this approach sets the stage for deeper sequencing and computational approaches to facilitate gene discovery.

Participants were identified from an institutional database (MSK-IMPACT) of patients who gave written informed consent to undergo parallel sequencing via custom, targeted panel as part of an Institutional Review Board approved study (NCT01775072; refs. 9, 10). Of the 24,241 patients consented prospectively from January 2013 to February 2019, 4,696 patients with ≥2 cancers and at least one tumor (any order) sequenced were identified through database query. Subsequently, International Agency for Research on Cancer (IARC) rules (11) for MPC were applied to define separate primaries. Briefly, IARC rules state that a primary tumor is one that originates in a primary site or tissue and is not an extension, recurrence, or metastasis. The existence of MPC does not depend on time. Only one tumor shall be recognized as arising in an organ or a pair of organs with the following exceptions: (i) systemic cancers involving multiple different organs and (ii) neoplasms of different morphology.

Subsequently, 356 patients were excluded, resulting in 4,340 patients with multiple (≥2) primary cancers and tumor sequencing of at least one tumor, with 764 (17.6%) having undergone tumor sequencing of >1 sample. These 4,340 patients with MPC were included in the epidemiologic analysis, where clinical characteristics were represented by summary statistics, and tumor types were tabulated for each individual patient and counted separately (pooled analysis). For the subset of 1,580 patients consenting to receipt of germline analysis of ≥76 cancer predisposition genes, it was permitted to link clinicopathologic characteristics and germline pathogenic/likely pathogenic variants (PV) on individual identified samples (Fig. 1). All analyses were conducted in accordance with recognized ethical guidelines under MSK IRB 12-245 and 18-420.

Patient sex, age at first cancer diagnosis, self-reported race/ethnicity, and Ashkenazi Jewish (AJ) ancestry (self-reported or documented in the medical record) were abstracted from the medical record and clinical databases. Smoking status was defined as ever versus never smoker. Body mass index (BMI) was obtained from the time of tumor sequencing. Data regarding cancer diagnoses, both before and after enrollment onto MSK-IMPACT, were collected from the institution's Cancer Database (CDB) which comprehensively collects data via review of ICD codes, pathology reports, and medical records to encompass self-reported cancer diagnoses, regardless of where treatment occurred. IARC rules (11) were used to evaluate tumor type from ICD codes, histology codes, and laterality to define MPC as summarized above. Tumor types were then grouped per Oncotree (12) categories to facilitate analysis. Cases with conflicting or ambiguous diagnoses were manually reviewed by medical oncologists (YLL and KC).

To assess distribution and relationship of different types of cancers, tumor pairs were constructed and plotted on the basis of frequency, regardless of time of occurrence (Supplementary Fig. S1). Figures and heatmaps of tumor pairs were created and stratified by sex, age at diagnosis (<50 and 50+; ref. 13), number of MPCs (2 vs. 3+), consent for receipt of germline findings (yes vs. no), and germline findings (positive vs. negative for PV). Tumor pairs were also depicted for those with synchronous primaries, defined as 1st and 2nd cancers with same age of diagnosis.

Age- and sex-specific incidence rates for 35 cancer types were obtained from the SEER-21 database as reference. For this analysis, cancer types in our cohort were grouped according to SEER groupings, and those without corresponding SEER data were excluded from this analysis. For a tumor pair, the probability that the patient develops the second tumor, for a given first primary tumor, was calculated using the age-adjusted SEER rates for the time from onset of first cancer to either the second cancer for MPC or last follow-up for single primary cancers (SPC) from our institutional database (MSK-IMPACT; Supplementary Materials and Methods). The expected count for the tumor pair is the sum of all these probabilities, and the SIR is estimated as the ratio of observed to expected count. The confidence bounds for the SIR were calculated using percentile bootstrap intervals by resampling the subjects and repeating the SIR calculations (14). Given observed differences in distribution of date of diagnosis for first primary cancer compared with second primary cancer or patients with SPC, a weight function was used to downweigh the MPC cases whose first primaries were diagnosed farther in the past. Different weighting levels were explored in a sensitivity analysis. This analysis was performed overall and then repeated excluding cancers diagnosed within 1 year to control for synchronous primaries. Analyses were restricted to SIRs with weighted expected frequency ≥1 and weighted number of pairs ≥2 to exclude outliers, and sensitivity analysis was performed for additional pairs with expected frequency ≥0.5. SIR analysis was not performed for likelihood of developing third or later cancers due to limited sample size. Groupings associated with known exposures or hereditary cancer syndrome were annotated (6, 15, 16), and scoring of known inherited cancer syndromes was performed as described in Supplementary Table S1.

Blood and tumor samples were analyzed using a custom panel (MSK-IMPACT), and germline variant calling was performed as described previously (10, 17). Identified variants were independently assessed and manually curated, applying current standards for variant classification by the American College of Medical Genetics and Genomics (18), to define likely pathogenic/pathogenic variants, here referred to as PV. Germline variants from the .bam files of the nontumor DNA sequence with mapping and base quality scores of >20 were called using MuTect (19) and GATK Haplotypecaller (20) using 25% variant frequency and 20× coverage thresholds. All variants with <1% population frequency in the Exome Aggregation Consortium (ExAC) database were interpreted. Copy-number variations were assessed using an in-house developed pipeline and were confirmed with an additional laboratory tests (21). Variants of uncertain significance were reviewed but not reported in this analysis. PV (mutations) were classified as high [relative risk (RR), >4], intermediate (RR, 2–4) or low (RR, <2) penetrance, recessive, or of uncertain clinical actionability based on disease risks and prior modeling (17, 22, 23). Genes of uncertain clinical actionability included MRE11A, RAD50, TERT, RECQL, ERCC3, RAD51B, PAX5, and the CHEK2 I157T. Genotype–phenotype associations were assessed, and concordance was defined as having at least one tumor type known to be associated with the PV per our genetics clinical guidelines.

LOH was assessed using the FACETS algorithm (24), which was performed on tumor samples from patients in whom a pathogenic germline variant was identified (n = 324). As some patients had multiple tumor samples tested (n = 359 tumor samples), and 38 samples from 34 unique patients had two germline PV identified, there were a total of n = 397 instances where a germline PV was assessed for LOH in each tumor sample. Of these, output data from FACETS was available for 381, with the algorithm having failed for 16 samples where LOH was unable to be determined due to lack of coverage at the specific loci (n = 11), low tumor burden (n = 2), genome fragmentation at the loci (n = 2), and unknown reason (n = 1). The proportion of tumors exhibiting LOH was quantified for the entire cohort and then stratified by gene penetrance (high vs. not high) and concordant/discordant tumor type. Given variability in tumor purity, sensitivity analysis was performed at different purity levels.

A subset of the human sequencing data (MSK-IMPACT) utilized in this study are available via cbioportal.org. The full SIR dataset generated in this study is not publicly available due to ongoing research projects but are available upon request from the corresponding author. Other data generated in this study are available within the article and its supplementary data files.

Of the 24,241 patients undergoing tumor sequencing, 4,340 (18%) patients had multiple primary tumors. Of these, 2,424 (56%) were female. Median age of first cancer diagnosis was 58 (range 0–89), and 100 patients (2%) had their initial cancer diagnosis before adulthood (<18 years of age). The majority (n = 3,697, 85%) were Non-Hispanic White, and 919 (21%) were of AJ ancestry. Among those with MPC, 2,230 (51%) patients described themselves as ever smokers. Median BMI at time of tumor sequencing was 27 (range 16–47) in men and 25 (range 15–54) in women. Most patients (n = 3,465, 80%) had two primary tumors; however, 683 (16%) had three, 141 (3%) had four, and 51 (1%) had ≥5 primary tumors, Table 1 and Fig. 1. Synchronous primaries were found in 877 (20%) of patients with MPCs. The most common synchronous cancers were breast–breast (n = 170), lung–lung (N = 101), bladder–prostate (N = 48), uterus–ovary (N = 31), and melanoma–melanoma (N = 26; Supplementary Fig. S2).

Of the 9,812 total primary tumors observed among patients with MPC, the most common tumor types were breast (N = 1960, 20%), lung (N = 1387, 14.1%), prostate (N = 856, 8.7%), melanoma (N = 773, 7.9%), and bladder cancer (N = 454, 4.6%). In men, the most common tumor types were prostate (N = 856, 19.8%), lung (N = 552, 12.8%), and melanoma (N = 451, 10.5%), whereas, in women, the most common tumor types were breast (N = 1,926, 35.1%), lung (N = 827, 15.1%), and uterus/endometrial (N = 357, 6.5%; Supplementary Table S2). In those with 5+ primary tumors, the most common tumor types were melanoma (N = 79), lung (N = 38), breast (N = 30), colon (N = 16), and prostate (N = 13) cancers (Supplementary Table S3).

Of 3,921 tumor pairs in women, the most common were breast–breast (n = 593, 15%), breast–lung (n = 330, 8%), and lung–lung (n = 230, 6%; Fig. 2A). Of 3,021 tumor pairs in men, the most common were bladder–prostate (n = 199, 7%), melanoma–melanoma (n = 172, 6%), and prostate–lung (n = 168, 6%; Fig. 2B). In those <50 years old, breast–cancer pairs were observed frequently, whereas prostate and lung cancer pairs were observed frequently in those age 50+ (Fig. 3A). Ovarian and pancreatic cancers were observed frequently in those with two primaries. Lung and melanoma pairs were observed frequently in those with 3+ primaries. Breast cancers were common in both groups (Fig. 3B). To assess for bias in those consented to receipt of germline findings compared with those not consented, we compared tumor pairs in each group and found that lung cancers and melanoma were overrepresented in the group not consented to germline results (Fig. 3C). In those with positive germline findings, breast–breast, breast—ovary, and breast–pancreas pairs were common. Melanoma, thyroid, prostate, and lung pairs were common in those without germline findings (Fig. 3D).

Weighted (w = 0.9), sequential 1st–2nd metachronous tumor pair combinations (synchronous removed) seen in excess of expected SIR included lymphoma–lung, lymphoma–uterus, breast–sarcoma, and breast–brain in women and melanoma–melanoma, melanoma–pancreas, prostate–mesothelioma, prostate–sarcoma, and kidney–kidney in men (Fig. 4; Supplementary Table S4). Synchronous primaries (yellow) included uterus–ovary, lung–lung, rectum–lung, and colon–colon in women and kidney–kidney, colon–rectum, colon–colon, and bladder–prostate in men (Supplementary Table S4). In this analysis, metachronous combinations were scored as associated with known hereditary cancer syndromes (blue) or possible exposures such as smoking (green) and obesity (black). Subsequently, combinations not thought to be related to a known hereditary cancer syndrome or environmental exposure were identified and included lymphoma–lung, lymphoma–uterus, breast–brain, and melanoma–lung in women and prostate–mesothelioma, prostate–sarcoma, melanoma–stomach, prostate–brain, and lymphoma–lung in men (Supplementary Table S4). Many of these associations remained significant using different weighting factors (Supplementary Table S5). Relaxing thresholds for expected frequency (≥0.5) revealed additional associations including melanoma–melanoma, breast–gallbladder, and lymphoma–thyroid in women and prostate–gallbladder, melanoma–brain, and lymphoma–pancreas in men (Supplementary Table S5).

Of the 4,340 patients, germline genetic testing data were available in 1,580 patients, of which 324 (21%) patients had at least one germline PV identified. Rates of germline PV were similar in those with synchronous versus metachronous cancers (19% vs. 21%, P = 0.54). Those with PV were younger at first cancer diagnosis (P < 0.001) and more likely to have >2 primary cancers (P = 0.004). Those consenting to return of germline results (n = 1,580) were less likely to have ever smoked compared with those not consenting to receipt of germline results (n = 2,760; P < 0.001). Although there was a higher proportion of AJ patients consented to receipt of germline results compared with those not consented (23% vs. 20%, P = 0.019), the proportion of those with positive results on germline testing was similar between AJ and non-AJ patients (26% vs. 22%, P = 0.133; Table 1).

We identified 361 PVs, with 33 patients having two, and 2 patients having three variants. Of these, 158 (44%) were in high penetrance genes (e.g., BRCA1/2, mismatch repair genes, and TP53), 73 (20%) in moderate penetrance genes (e.g., CHEK2 and ATM), and the remainder 130 (36%) were in low penetrance genes or genes of uncertain clinical utility (Fig. 5). Among all PV, 104 (29%) were founder mutations, of which 78 (75%) were AJ founder mutations. There was a higher prevalence of high (60% vs. 44%) and moderate (24% vs. 19%) penetrance genes and a lower prevalence of low/recessive and uncertain (16% vs. 37%) penetrance genes in those with three or more primary tumors compared with those with two primary tumors, P = 0.002.

The distribution of tumor types in those with an identified germline PV were consistent with known associations, and there was high genotype–phenotype concordance overall (68%). We observed a higher genotype–phenotype concordance in high (89%) compared with moderate (58%) and low (15%) penetrance genes (58%), P < 0.001 (Fig. 5; Supplementary Table S6). The most frequently observed high penetrance genes were BRCA1/2, MLH1/MSH2, and TP53, which all exhibited high canonical genotype–phenotype concordance. The most frequently observed moderate penetrance genes were CHEK2 and ATM, which exhibited lower levels of genotype–phenotype concordance (Fig. 5; Supplementary Table S6).

Of the 381 tumors profiled in 324 patients with at least one pathogenic germline variant, 174 (46%) exhibited LOH. LOH was significantly higher for high (62%) compared with moderate (44%) and low (29%) penetrance genes, P < 0.01 overall and for comparisons of high versus moderate and high versus low penetrance genes (Fig. 5). LOH rates were similar when analyzing PV groups with varying tumor purity cutoffs (Supplementary Table S7). LOH was observed in 64% of concordant and 55% of discordant tumors in high penetrance cases and 45% of concordant and 37% of discordant tumors in moderate penetrance cases (Fig. 5).

In this study of 24,241 patients with cancer undergoing parallel tumor-normal sequencing, 18% had MPCs. Novel tumor pairs not attributed to known inherited cancer syndromes or exposures were identified including lymphoma–lung, lymphoma–uterus, breast–brain, and melanoma–lung in women and prostate–mesothelioma, prostate–sarcoma, melanoma–stomach, prostate–brain, and lymphoma–lung in men. Among patients with MPC who consented to germline genetic testing, 21% had at least one PV identified, of which 44% were in high penetrance genes. LOH in tumors was observed for 46% of tumor variants, and this rate was higher (62%) for high penetrance genes as compared with moderate/low penetrance genes (33%).

The MPC rate of 18% reported here is consistent with prior studies and SEER data showing rates ranging from 2% to 25% (1, 2, 25). Older (1992–2008) SEER database studies derived the incidence of second primary cancers to be 8.1% (25). However, recent reports show that MPC rates are increasing over time, particularly in the elderly population (age ≥ 65; refs. 2, 3). An updated analysis of the SEER database (2009–2013) found an MPC rate of 18.4% and observed that this rate was higher for those age 65 and older (1). The median age of diagnosis for initial cancer for our cohort was younger, and only 1046 (24%) of patients were ≥65 years at initial cancer diagnosis. For this analysis, we used the IARC classification for MPC as it is more stringent and leads to lower MPC rates than the SEER definition (26).

Prior population-based studies utilizing cancer registry data have characterized the epidemiology of MPCs and primarily identified an excess of cancers associated with exposures such as smoking and alcohol, for example, lung, bladder, esophageal, and head/neck (1, 25–27). A SEER analysis (1992–2011) of >1.5 million cancer survivors found that 156,444 subsequently developed MPCs, with most cancers associated with environmental exposures (6). Those reports highlighted the need for preventive strategies; however, those studies were limited by lower-than expected rates of common cancers such as prostate and breast cancer and lack of germline analysis. In contrast, this study did not observe an excess of exposure-related 1st–2nd cancer pairs compared with prior population-based studies (6, 25), potentially reflecting the different tumor composition of this population referred to a comprehensive cancer center compared to registry-based analyses.

Even after removing pairs presumably associated with known inherited cancer syndromes or exposures, significant associations of tumor pairs involving melanoma, lymphoma, uterine, breast, prostate, lung, sarcoma, and thyroid cancers were observed. Of the novel tumor pairs, the statistical association of lung cancer following lymphoma was shown in a prior meta-analysis, and treatments such as radiotherapy (RT), particularly in Hodgkin's lymphoma, were hypothesized to influence this association (28). Notably, 7 of 20 (35%) women and 7 of 19 (37%) men with lung cancer after lymphoma received RT in our cohort, and other cancers after lymphoma (e.g., breast cancer) may also be related to radiation, a finding that should be explored in future studies with more comprehensive assessments of prior therapies. Some cancers may be incidentally detected during workup of another cancer (e.g., kidney–kidney and bladder–prostate) with unclear genetic connection, although rates of positive germline findings were similar in synchronous versus metachronous cancers. Other pairs may represent novel tumor associations that require more comprehensive germline analysis and assessment of exposures (5). Recognition of these novel clusters may have implications for targeted screening and prevention in cancer survivors. For example, many of these cancers have validated screening modalities (e.g., mammogram, prostate specific antigen, colonoscopy, low-dose CT for pulmonary lesions, and skin evaluations), and enhanced surveillance may be considered in survivors of certain cancers if these associations were to be verified.

Few studies have evaluated germline findings in patients with MPC, and none have assessed associations with somatic sequencing. Chan and colleagues examined 1,191 Asian patients with cancer, >80% female, of which 19.4% had MPCs with common tumor pairs including breast–breast, breast–ovary, colon–colon, and colon–endometrial (29). They found that patients with MPC were more likely to have germline findings compared with those with single cancers (34.5% vs. 25.8%) with most common variants in high penetrance genes (>90%; ref. 29). Another study of 212 patients with MPC, 84% female, referred from genetics clinics observed that 20.7% had positive germline findings on clinical testing (30). We previously reported a rate of positive germline findings of 19.7% in an unselected, pan-cancer population (17), with a rate of 17.9% in single primary patients with cancer from that same dataset compared with our rate of 21% in this MPC cohort, P = 0.12. Many prior cohorts were predominantly female, whereas, our cohort is more balanced. In those MPC cases with PVs in highly penetrant genes, genotype–phenotype concordance was high, and the prevalence of LOH was substantial in tumors, particularly in concordant tumors, supporting their etiologic role as “drivers” of these phenotypes. Further studies are needed to evaluate the contribution of moderate/low penetrance genes (31, 32) in the broad spectrum of tumor types observed, as well as gene–gene and gene–environment interactions (4), and the role of genetic modifiers (33, 34), and further work building on this study is ongoing.

Notably, those with no identified PV in known cancer susceptibility genes comprised much of the cohort, suggesting the potential for undiscovered, monogenic or polygenic drivers, or modifiers of tumorigenesis. Whitworth and colleagues performed whole-genome sequencing (WGS) in 460 individuals with MPCs who had undergone genetic assessments and found previously undetected germline PV in moderate/high penetrant genes in 67 (15.2%) of cases, of which 29 had tumor phenotypes discordant with those typically observed (8). In that series, unlike the current and previous series (17), tumor genotyping was not performed to assess LOH to infer a causative role; previously performed WGS in that series (8) also did not detect any clinically relevant noncoding variants, although a small number of structural variants of unclear clinical significance were found. An additional approach to discover shared genomic drivers of MPC would be a comprehensive genomic interrogation and functional assessment of not only the germline but also of multiple tumor genomes from the same patient to identify shared variants or structural alterations, and this work is ongoing.

This study has several limitations. It is comprised of mostly white patients with advanced cancer and a large AJ population (17). There was a potential ascertainment bias in patients offered tumor sequencing, types of tumors represented, and those consenting to receipt of germline results, potentially biasing our analysis. In a previous series (17), adjusting the MSK IMPACT case mix to a population derived case mix resulted in a modest decrease in actionable germline mutation rates observed, suggesting that the rate of germline PV in patients with MPC in the general population is likely to be even lower than the 21% reported here. As an additional limitation, certain tumor histologies (e.g., neuroendocrine tumors) span multiple anatomical sites, complicating analysis. To provide a first approximation of associations, SIR analysis was limited to the subset of the first two cancers occurring, yet, sample sizes for individual tumor pairs were small, resulting in high SIRs with wide confidence intervals. To adjust for this, we sought to normalize the distribution of age at diagnosis for the first primary cancer to reduce bias and performed sensitivity analyses with different weighting factors as well as thresholds of expected frequency to allow consideration of rare cancer associations. As shown by this analysis, strong associations with rare tumors, for example, gall bladder cancers, are evident if thresholds for expected frequency are set to ≥0.5, whereas significance for many other associations diminish as the degree of normalization of the MPC incidence rates is increased. Similarly, all association effects observed here, were impacted by use of single-cancer incidence rates from SEER to calculate expected rates of tumor pairs, which does not account for synchronous cancers and thus may underestimate pairs with hereditary predisposition or common environmental exposure. To address this potential bias, the analysis was adjusted by removing synchronous cancers and highlighting cases with suspected inherited cancer syndromes and/or environmental exposures. Because of these limitations and assumptions, future, prospective studies of cancer patients with a distribution representing the SEER population with long-term follow-up are needed to verify the putative associations, including those of synchronous tumors, reported here. Despite these limitations, this study contains the largest cohort to date of patients with MPC with germline and tumor analysis.

MPC represented a substantial burden of disease in an oncologic pan-cancer cohort of patients undergoing parallel genomic sequencing. Only a small proportion of those patients had germline findings in the limited genes tested, with those PVs observed mostly in high penetrant genes corresponding to clinical phenotype. Standardized analysis of rates of co-incident MPCs revealed novel tumor pairs, not explained by known inherited cancer syndromes or exposures, which could potentially be used clinically to target prevention and screening. More comprehensive evaluation, including sequencing of genomes and/or transcriptomes of primary tumors from individual patients may uncover novel genetic drivers of multiple tumor phenotypes.

Y.L. Liu reports grants from NIH NCI Core grant P30 CA008748 and Robert and Kate Niehaus Center for Inherited Cancer Genomics during the conduct of the study as well as grants from AstraZeneca and GlaxoSmithKline outside the submitted work. K.A. Cadoo reports grants from NCI Core grant P30 CA00874 during the conduct of the study as well as personal fees from Astra Zeneca, MJH Life Sciences, MSD Ireland, and GSK Ireland and other support from MSD Ireland, Immunogen, Pfizer, and Roche Ireland outside the submitted work. Y. Wang reports grants from NIH during the conduct of the study as well as personal fees from New York University outside the submitted work. B. Devolder reports grants from NIH NCI Core Grant during the conduct of the study. M.F. Berger reports personal fees from Eli Lilly and PetDx outside the submitted work. D.B. Solit reports personal fees from Pfizer, Loxo/Lilly Oncology, BridgeBio, Scorpion Therapeutics, Fore Therapeutics, and Vividion Therapeutics outside the submitted work. D.F. Bajorin reports personal fees from Bristol Myers Squibb, Merck, Dragonfly, and Fidia Farmaceutici outside the submitted work. Z.K. Stadler reports other support from Adverum, Alcon, Neurogene Inc., Gyroscope, and RegenexBio outside the submitted work. J. Vijai reports a patent for Diagnosis & Treatment of ERCC3-Mutant Cancer pending. K. Offit is a founder of AnaNeo Pharmaceuticals outside the submitted work; in addition, K. Offit reports a patent for Diagnosis & Treatment of ERCC3-Mutant Cancer pending. No disclosures were reported by the other authors.

Y.L. Liu: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. K.A. Cadoo: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. S. Mukherjee: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. A. Khurram: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration. K. Tkachuk: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. Y. Kemel: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–review and editing. A. Maio: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–review and editing. S. Belhadj: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. M.I. Carlo: Conceptualization, writing–review and editing. A. Latham: Conceptualization, writing–review and editing. M.F. Walsh: Conceptualization, writing–review and editing. M.E. Dubard-Gault: Conceptualization, writing–review and editing. Y. Wang: Data curation, software, formal analysis, investigation, visualization, methodology, writing–review and editing. A.R. Brannon: Data curation, software, formal analysis, visualization, methodology, writing–review and editing. E. Salo-Mullen: Conceptualization, data curation, formal analysis, investigation, methodology, writing–review and editing. M. Sheehan: Conceptualization, data curation, formal analysis, project administration, writing–review and editing. E. Fiala: Data curation, validation, investigation, methodology, writing-review and editing. B. Devolder: Data curation, investigation, project administration, writing–review and editing. S. Dandiker: Data curation, investigation, project administration, writing–review and editing. D. Mandelker: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. A. Zehir: Conceptualization, data curation, writing–review and editing. M. Ladanyi: Conceptualization, data curation, writing–review and editing. M.F. Berger: Conceptualization, data curation, formal analysis, writing–review and editing. D.B. Solit: Conceptualization, data curation, writing–review and editing. C. Bandlamudi: Conceptualization, data curation, investigation, methodology, writing–review and editing. V. Ravichandran: Conceptualization, data curation, formal analysis, writing–review and editing. D.F. Bajorin: Conceptualization, writing–review and editing. Z.K. Stadler: Conceptualization, investigation, writing–review and editing. M.E. Robson: Conceptualization, writing–review and editing. J. Vijai: Conceptualization, data curation, formal analysis, writing–review and editing. V. Seshan: Conceptualization, data curation, formal analysis, supervision, investigation, methodology, writing–original draft, writing–review and editing. K. Offit: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, writing–original draft, writing–review and editing.

We would like to thank the patients and families that participated in this research study. This work was supported by the Robert and Kate Niehaus Center for Inherited Cancer Genomics and the Sharon Corzine Foundation. MSK and its researchers are supported by the NCI Core grant P30 CA008748. D.F. Bajorin, J. Vijai, and K. Offit are funded by the Bladder SPORE P50CA221745. J. Vijai, K. Offit, A. Khurram, S. Dandiker, and Y. Kemel are funded by the Sharon Corzine Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.