Clinical Factors Associated with Urinary Tract Cancer in Individuals with Lynch Syndrome

With an estimated 1:279 prevalence in the general population, Lynch syndrome is one of the most common inherited cancer predisposition syndromes, conferring markedly increased lifetime risks of gastrointestinal, gynecologic, urinary tract, and other cancers (1–4). Lynch syndrome is caused by inheritance of pathogenic germline variants in the DNA mismatch repair (MMR) genes (MLH1, MSH2, MSH6, and PMS2) or EPCAM, and identifying Lynch syndrome carriers can greatly facilitate genetically driven cancer prevention strategies. Specifically, there are proven risk-reducing interventions for the three most common Lynch syndrome–associated cancers: early and frequent colonoscopic surveillance (5) and aspirin chemoprevention (6) can substantially reduce colorectal cancer incidence for male and female Lynch syndrome carriers, and risk-reducing hysterectomy and salpingo-oophorectomy (7) virtually eliminates female Lynch syndrome carriers' risk of Lynch syndrome–associated endometrial and ovarian cancer, respectively.

In aggregate, urinary tract cancers (UTC), including renal pelvis, ureter, bladder, and possibly kidney cancers, are arguably the second and fourth most common Lynch syndrome–associated cancers in male and female Lynch syndrome carriers, respectively, and are the most common Lynch syndrome–associated neoplasms for which there is no proven effective screening or prevention strategies (1–3). The very limited data examining UTC surveillance in unselected Lynch syndrome carriers have found unacceptably low sensitivity and high false positive rates (8), and consequently most professional society guidelines (9–14) do not recommend UTC screening in Lynch syndrome, even though it is a significant contributor to the overall burden of Lynch syndrome–associated malignancy. Furthermore, recent prospective analyses have found inferior survival outcomes for Lynch syndrome–associated UTC compared with the more classic Lynch syndrome–associated malignancies, with prospective data demonstrating 71% [95% confidence interval (CI), 34%–89%] and 81% (95% CI, 42%–95%) 10-year survival rates for Lynch syndrome–associated kidney/ureter and bladder cancers, respectively, versus 88% (95% CI, 77%–94%) and 93% (95% CI, 85%–97%) for Lynch syndrome–associated colon and endometrial cancers, respectively (1).

Because the performance characteristics of any cancer screening modality are inherently improved when used in individuals at particularly increased risk, identifying risk factors for Lynch syndrome–associated UTC is critical for the development of effective patient-specific UTC surveillance and prevention strategies. Multiple studies have consistently found that Lynch syndrome carriers with pathogenic germline MSH2 variants have particularly high likelihood for developing UTC (1, 3, 15–21), but other risk factors for Lynch syndrome–associated UTCs remain unknown. The aim of this study was to examine patient-specific factors that may help identify which individuals with Lynch syndrome are at particular risk for UTC.

As described previously (22), we analyzed clinical data from a cohort of 52,758 consecutively ascertained individuals unknown to be related to one another who underwent germline testing of two or more Lynch syndrome genes (MLH1, MSH2, MSH6, PMS2, and EPCAM) at a commercial laboratory (Myriad Genetic Laboratories, Inc.) from June 2006 to July 2013 for evaluation of suspected Lynch syndrome. Individuals undergoing site-specific germline analysis for a specific Lynch syndrome variant were not included, and all germline testing was ordered as syndrome-specific testing for Lynch syndrome genes only, rather than multisyndromic multigene panel testing. Germline testing methodologies and variant classification were performed by Myriad Genetic Laboratories, Inc., as described previously (22, 23).

Clinical data were obtained from the test order form completed by patients' healthcare provider ordering germline Lynch syndrome testing, as described previously (23). Data collected included sex, age at genetic testing, race, personal history of cancer (including specific cancer types and ages at diagnosis), and family history of cancer (including relationship to the proband, specific cancer types, and ages at diagnosis) in first- and second-degree relatives (FDR and SDR, respectively). The results of germline testing along with the aforementioned clinical data were provided to investigators as a deidentified dataset by Myriad Genetic Laboratories, Inc. Individuals with pathogenic or likely pathogenic alterations detected in MLH1, MSH2, MSH6, PMS2, or EPCAM were collectively defined as Lynch syndrome carriers and those lacking any such germline alterations (including those with germline variants of uncertain significance) were defined as noncarriers (23). Family history of UTC was examined as both a categorical (yes/no) and continuous variable (aggregate number of FDRs and SDRs with UTC).

The χ² test was used to analyze differences between categorical variables. Continuous variables were compared using the Kruskal–Wallis test. Clinical factors associated with UTC in Lynch syndrome carriers were analyzed using univariate and multivariate logistic regression models. Selected variables found to have a significant association with UTC on univariate analysis were included in the multivariate analysis. Clinical variables with possible association to UTC were assessed via their ORs with 95% CIs. P values were two-tailed and considered statistically significant at alpha < 0.05. SAS statistical software version 9.4 (SAS Institute, Inc.) was used for data management and univariate analysis, and R version 3.3.3 (R Foundation for Statistical Computing) was used for multivariate logistic regression. A waiver of consent for study participants was obtained because analyses were performed on deidentified data and did not require patient contact. The study was approved by the Dana-Farber/Harvard Cancer Center Institutional Review Board.

After excluding 1,672 individuals from analysis due to missing clinical data (n = 1,664) or presence of multiple pathogenic germline Lynch syndrome variants (n = 8), the final study cohort consisted of 51,086 individuals (Table 1), 3,828 (7.5%) of whom carried of a pathogenic germline Lynch syndrome variant: 1,346 MLH1, 1,639 MSH2, 670 MSH6, 145 PMS2, and 28 with EPCAM variants. Compared with noncarriers, Lynch syndrome carriers were significantly more likely to have a personal history of any UTC (4.1% vs. 1.2%; P < 0.0001; OR 3.22; 95% CI, 2.66–3.90; P < 0.0001) and were significantly more likely to have ureter/renal pelvis cancer (1.6% vs. 0.1%; P < 0.0001), bladder cancer (1.8% vs. 0.4%; P < 0.0001), and kidney cancer (1.3% vs. 0.7%; P = 0.0003). In addition, Lynch syndrome carriers were significantly more likely to have a family history of any UTC in FDR/SDRs than noncarriers (9.6% vs. 6.5%; P < 0.0001). Compared with noncarriers, Lynch syndrome carriers with pathogenic germline variants in MLH1 (OR 1.64; 95% CI, 1.11–2.27; P = 0.01), EPCAM/MSH2 (OR 6.07; 95% CI, 4.93–7.48; P < 0.0001), and MSH6 (OR 2.31; 95% CI, 1.43–3.71; P = 0.0006) were significantly more likely to have a personal history of UTC, but not those with PMS2 (OR 0.58; 95% CI, 0.08–4.15; P = 0.59) variants.

Of the 158 Lynch syndrome carriers with a history of UTC, 62 (39.2%) had ureter/renal pelvis cancer, 67 (42.4%) had bladder cancer, 49 (31.0%) had kidney cancer, and 21 (13.3%) had multiple urinary tract cancers. Seventy-three of 158 (46%) Lynch syndrome carriers with UTC were male, 48 (30.4%) had a family history of UTC in at least one FDR and/or SDR, and 113 (71.5%) harbored pathogenic germline EPCAM/MSH2 variants (Table 2; Supplementary Table S1). Of the 369 Lynch syndrome carriers with a family history of UTC in ≥1 FDR/SDR, 202 (54.7%) carried pathogenic germline EPCAM/MSH2 variants, 35 of whom had both a personal and family history of UTC (Supplementary Table S2). Compared with Lynch syndrome carriers without UTC, Lynch syndrome carriers with UTC were more likely to be Caucasian (69.6% vs. 58.9%; P = 0.05), harbor pathogenic EPCAM/MSH2 variants (71.5% vs. 42.3%; P < 0.0001), and have a personal history of other Lynch syndrome cancers besides UTC (88.6% vs. 82.5%; P = 0.047), including colorectal, endometrial, gastric, and small bowel cancers.

By multivariable logistic regression analysis, personal history of UTC among Lynch syndrome carriers was independently associated with male sex (OR 1.95; 95% CI, 1.38–2.76), increasing age (OR 2.44 per 10 years; 95% CI, 2.11–2.82), familial burden of UTC (OR 2.69 per FDR/SDR with UTC; 95% CI, 1.99–3.63), and germline EPCAM/MSH2 variants (OR 4.01; 95% CI, 2.39–6.72; ref: MSH6/PMS2) but not MLH1 variants, race, or personal history of Lynch syndrome–associated cancer other than UTC (Table 3).

Of the 158 Lynch syndrome carriers with UTC, 143 (90.5%) had at least one of the following characteristics: male sex, pathogenic germline EPCAM/MSH2 variants, or family history of UTC in one or more FDR/SDR. Conversely, only 15 of 1,251 (1.2%) Lynch syndrome carriers lacking all three of these characteristics had a personal history of UTC. Of the 49 Lynch syndrome carriers with particularly strong family histories of UTC (≥2 FDR/SDRs with UTC), only 7 (14.3%) had a personal history of UTC.

This study of over 51,000 individuals with suspected Lynch syndrome and more than 3,800 confirmed Lynch syndrome carriers found UTCs to be the second and fourth most common malignancy (behind colorectal, endometrial, and ovarian cancers) in male and female Lynch syndrome carriers, respectively. Furthermore, the vast majority of Lynch syndrome carriers with UTC in this cohort also had a personal history of other Lynch syndrome–associated malignancies, indicating potential missed opportunities for Lynch syndrome–associated cancer prevention and screening. These findings are consistent with other large cohorts (1–3), and emphasize the critical need for effective UTC surveillance for Lynch syndrome carriers. Compared with gastrointestinal and gynecologic malignancies, however, UTCs remain strikingly understudied in Lynch syndrome, and there are minimal data about patient-specific factors that contribute to UTC risk in Lynch syndrome carriers. Our findings demonstrate that male sex, increasing age, pathogenic EPCAM/MSH2 variants, and familial burden of UTC are each independently associated with UTC in Lynch syndrome carriers, suggesting that a predictable fraction of patients with Lynch syndrome is most likely to benefit from UTC risk-reduction strategies.

Numerous prior studies (1, 3, 15–21) have clearly established that Lynch syndrome carriers with pathogenic germline variants in MSH2 are at disproportionately increased risk of UTC, compared with Lynch syndrome carriers with pathogenic variants in other DNA MMR genes. In particular, recent data from 3,119 Lynch syndrome carriers in the multinational Prospective Lynch Syndrome Database described a cumulative incidence to age 75 years of 17.0% and 8.1% for ureter/kidney cancer and bladder cancer, respectively, among MSH2 carriers, versus 4.6% and 4.1% for MLH1 carriers, 3.0% and 8.2% for MSH6 carriers, and 0% for PMS2 carriers (1). Another study (3) of 2,118 Lynch syndrome carriers from the German and Dutch national Lynch syndrome registries found pathogenic MSH2 variants to be significantly associated with history of bladder cancer (HR 5.42; 95% CI, 1.89–15.56) and other urothelial cancer (HR 8.27; 95% CI, 2.95–23.19), compared with MLH1 and MSH6 variants, although no significant association was found with male or female sex for either bladder or other urothelial cancers. Our data confirm this important association with MSH2 variants and significantly add to the existing body of literature by identifying other potential independent risk factors, including familial burden of UTC.

To date, the role of family history as a risk factor for UTC in Lynch syndrome carriers has not been thoroughly studied. One study examining 136 UTCs from 288 Lynch syndrome families in the Danish national registry found that only 30 (22.1%) of UTCs developed in Lynch syndrome carriers with a family history of UTC, and the authors thus concluded that UTC screening should not be restricted only to those with such family histories (19). Although our data likewise demonstrate that a minority (30.4%) of patients with Lynch syndrome with UTC had a family history, our multivariate analysis found familial burden of UTC to have a significant association with personal history of UTC, such that the likelihood of UTC in Lynch syndrome carriers was almost tripled for each affected FDR/SDR.

Our data can inform UTC surveillance and prevention strategies by identifying a subset of Lynch syndrome carriers with particularly increased risk of UTC (beyond just those with MSH2 variants) while also providing reassurance to the large fraction of patients with Lynch syndrome for whom such surveillance may be particularly low yield. In fact, >90% of all Lynch syndrome–associated UTCs in our large cohort developed in patients with Lynch syndrome who were male, EPCAM/MSH2 variant carriers, and/or had a FDR/SDR with UTC, whereas UTCs were very rarely reported in Lynch syndrome carriers lacking all three of these features.

Unfortunately, there are currently no evidence-based approaches to UTC screening and prevention for individuals with Lynch syndrome. The largest study to date of UTC screening in Lynch syndrome examined 1,868 urine cytology screening specimens from 977 individuals with known or suspected Lynch syndrome enrolled in the Danish HNPCC register (8). Investigators found that only 2 of 1,868 (0.11%) of urine cytology specimens led to the diagnosis of an asymptomatic UTC and that there were at least 11 false positive urine cytology results for every 1 true positive result (8). Of the seven UTCs diagnosed in individuals undergoing regular urine cytology screening, five (71.4%) were in individuals who developed UTC symptoms in spite of normal urine cytology in the preceding 36 months, suggesting a poor (28.6%) sensitivity to such screening in unselected Lynch syndrome carriers (8). Other forms of UTC screening (e.g., urinalysis for microscopic hematuria, ultrasound, and ureterocystoscopy) have not been robustly studied in patients with asymptomatic Lynch syndrome, and the question of how to best identify early-stage asymptomatic Lynch syndrome–associated UTCs remains a critical unanswered question. Our findings, however, demonstrate that a predictable subset of patients with Lynch syndrome are at particularly elevated risk for UTC, meaning that UTC screening should be more sensitive and specific in these higher risk carriers than in all-comers with Lynch syndrome.

Novel forms of molecular-based UTC screening (24, 25) attempting to detect neoplastic genomic alterations in urine specimens have shown early promise in case–control studies. Given that Lynch syndrome–associated UTCs may have different molecular phenotypes than non-Lynch syndrome UTCs, it is unclear whether such approaches would be appropriate for Lynch syndrome–associated UTC screening. In light of our findings, we would propose that molecular-based UTC screening be prospectively studied in Lynch syndrome carriers to evaluate the efficacy of such approaches, particularly focusing on those who may be at highest risk (e.g., males, carriers of pathogenic EPCAM/MSH2 variants, and those with family histories of UTC). Prospective evaluation will also allow for critical examination of other clinical, behavioral, and lifestyle factors that could not be addressed in this study, including tobacco use and exposure.

Currently, no professional society guidelines overtly recommend routine UTC surveillance in all asymptomatic individuals with Lynch syndrome; however, these guidelines vary significantly in their conditional recommendations related to Lynch syndrome–associated UTC surveillance (Table 4). The question of how (and whom) to screen for Lynch syndrome–associated cancers beyond colorectal cancer is a critical real-world problem for all extracolonic cancers in Lynch syndrome, not just UTC. Risk-reducing hysterectomy and salpingo-oophorectomy effectively minimize the risk of Lynch syndrome–associated gynecologic cancer, but there are currently no evidence-based methods by which to prevent or even screen for cancers of the stomach, small intestine, pancreaticobiliary tract, brain, or urinary tract in Lynch syndrome carriers. As such, a common practice (endorsed by several of the aforementioned guidelines; refs. 11–13, 26) has been to selectively screen Lynch syndrome carriers for non-colorectal non-gynecologic cancers based on family history of these specific malignancies. Our finding that familial burden of UTC is independently associated with Lynch syndrome patients' likelihood of UTC lends support to this approach of family history–based risk assessment while also identifying other clinical factors that may identify high-risk Lynch syndrome carriers. It will be of particular interest to examine whether family history of other Lynch syndrome–associated malignancies (gastric, pancreatic, biliary, and small bowel cancers) is a similarly effective risk stratification technique in future studies.

Several important strengths of our study are worth highlighting. Our cohort was consecutively ascertained through a large commercial testing laboratory and allowed us to evaluate a geographically diverse population that is representative of the greater U. S. population of patients with known or suspected Lynch syndrome. The size of our cohort allowed us to simultaneously assess the association of multiple clinical factors with Lynch syndrome–associated UTC in parallel (e.g., controlling for the impact of MSH2 variants to quantify the relative effect of family history of UTC). In addition, evaluating family history of UTC as both a continuous and categorical variable allowed us to uncover the incrementally strengthened relationship with each additional FDR/SDR. Finally, our study included PMS2 and EPCAM carriers, whereas most previous studies assess UTC in Lynch syndrome have focused on only those with germline variants in MLH1, MSH2, and MSH6.

We acknowledge that there are several important limitations to our study. First, because this was a cross-sectional study we were unable to prospectively study the impact of these clinical factors on UTC risk, screening, or survival. We did not have any data on behavioral, lifestyle, or otherwise modifiable UTC risk factors such as smoking or chemical exposures, precluding our ability to assess whether such factors (often shared within families) may account for the observed link between personal and family history of UTC among Lynch syndrome carriers. The predominance of Caucasian and female patients is likely due to the ascertainment from a commercial genetic testing laboratory, and the cohort may likewise have been biased toward individuals with higher socioeconomic status, those receiving care at tertiary care centers, possible lower likelihood of smoking, and other potential confounding factors. All germline testing was performed specifically to assess for Lynch syndrome, rather than multigene panel testing, which may have biased the cohort to more penetrant clinical histories, particularly those with pathogenic germline MLH1 and MSH2 variants. As such, there were comparatively few PMS2 carriers in this study (even though such pathogenic variants appear to be the most prevalent in the general U.S. population; ref. 4), although recent data (27) have suggested that PMS2 carriers may have no increased risk of UTC compared with the general population. Although this study's primary comparison was between Lynch syndrome carriers with a personal history of UTC and Lynch syndrome carriers without such personal history, it is important to acknowledge that the noncarriers in this study were all ascertained to the cohort because of a personal/family history of Lynch syndrome–associated cancer and are thus not representative of the general population. We recognize the possibility of survival bias in that some patients with Lynch syndrome with a cancer diagnosis may not have lived long enough to undergo Lynch syndrome testing and thus would not have been included in this cohort. We also acknowledge that these data cannot effectively quantify the risk of UTC in Lynch syndrome carriers versus the general population, because the noncarriers in this study were all referred for germline Lynch syndrome testing and likely had a higher likelihood of UTC in their personal and family histories than the general population.

Finally, we were unable to verify the accuracy or completeness of each individual's reported personal and family history of cancer, which were collected from test request forms. In particular, this raises the possibility of recall bias, considering that a patient with UTC may be more likely to report a family history of UTC. To this end, however, a recent publication (26) studied the accuracy and completeness of data collected on commercial laboratory test requisition forms for 824 probands undergoing genetic testing and 3,954 relatives. By medical record review, this study (26) found >99% accuracy for probands', FDRs', and SDRs' cancer diagnoses provided on test requisition forms, although varying completeness of family history of cancer, with a higher likelihood of incomplete family history reporting in more distant family members, particularly third- and fourth-degree relatives.

However, one important consideration for our study is the postulated link between Lynch syndrome and “kidney cancer.” While our data echo those from numerous prior registry studies (1, 2, 28) which have described an increased risk of kidney cancer among individuals with Lynch syndrome, these studies and ours cannot rule out the very real possibility that diagnoses labeled as “kidney cancer” in many of these individuals actually represent upper tract urothelial (i.e., renal pelvis) carcinomas rather than renal cell carcinomas (often colloquially referred to as “kidney cancers”), which are genetically and histologically distinct neoplasms from one another (29–31). Adding further skepticism about the questionable link between Lynch syndrome and renal cell carcinomas, a recent large study (32) of paired germline and microsatellite instability testing in over 15,000 patients with cancer demonstrated that only 11 of 458 (2.4%) biopsy-proven renal cell carcinomas demonstrated microsatellite instability, none of whom had an identified germline Lynch syndrome variants; 2 patients with renal cell carcinoma were found to carry germline MSH6 variants, but their tumors lacked microsatellite instability, suggesting that they were not etiologically linked to the germline variant. In contrast, the same study (32) found microsatellite instability in 32 of 551 (5.8%) bladder/urothelial carcinomas, 12 of 32 (37.5%) whom were confirmed Lynch syndrome carriers.

In conclusion, we have identified several important clinical factors that are independently associated with UTC in individuals with Lynch syndrome. In addition to the previously well-described link to MSH2 variants, we found that male sex, increasing age, and familial burden of UTC were all independently associated with UTC in Lynch syndrome carriers. We are hopeful that, by identifying a subset of Lynch syndrome carriers with particular risk for UTC, this will allow for more rational design of personalized UTC surveillance and prevention strategies. Furthermore, the significant link with family history of UTC raises intriguing questions about risk stratification for other uncommon but potentially deadly Lynch syndrome cancers, although prospective studies will certainly be required to account for potential confounders, including recall bias and shared environmental effects. Future endeavors should focus on prospectively examining these potential UTC risk factors in prospective cohorts of patients with Lynch syndrome, improving UTC surveillance and prevention, and investigating the role of family history and other personalized clinical factors in the risk for other extracolonic Lynch syndrome–associated malignancies.

S. Syngal is a consultant for Myriad Genetics and Digital China Health Technologies, and has rights to an inventor portion of the licensing revenue from PREMM5. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.W. Wischhusen, M.B. Yurgelun

Development of methodology: J.W. Wischhusen, M.B. Yurgelun

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Ukaegbu, F. Kastrinos, S. Syngal

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.W. Wischhusen, C. Ukaegbu, H. Uno, F. Kastrinos, M.B. Yurgelun

Writing, review, and/or revision of the manuscript: J.W. Wischhusen, C. Ukaegbu, H. Uno, F. Kastrinos, S. Syngal, M.B. Yurgelun

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Ukaegbu, T.G. Dhingra

Study supervision: C. Ukaegbu, M.B. Yurgelun

This work was supported by the NIH (NCI) K24CA113433 (to S. Syngal), R01CA132829 (to S. Syngal), K07CA151769 (to F. Kastrinos), and The Pussycat Foundation Helen Gurley Brown Presidential Initiative (to C. Ukaegbu).

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.