Risk of Prostate Cancer in Lynch Syndrome: A Systematic Review and Meta-analysis

This article is featured in Highlights of This Issue, p. 367

Lynch syndrome, formerly known as hereditary nonpolyposis colorectal cancer (HNPCC), is a disorder characterized by cancer predisposition in families. The risk of disease follows an autosomal dominant pattern of inheritance, and is due to a germline mutation in one of the DNA mismatch repair (MMR) genes: MLH1 (1, 2), MSH2 (3), MSH6 (4, 5), and PMS2 (6, 7); or constitutional 3′ end deletions of EPCAM (8, 9). The frequency of pathogenic mutations in MMR genes in the general population has been estimated to be approximately 1:3,100 to 1:370 (10–12).

MMR proteins detect and excise DNA mismatches occurring during cell division. Tumors are more likely to develop from cells lacking any of the four MMR proteins, as a consequence of germline mutation and/or somatic loss of the MMR gene function. Excess of early-onset colorectal cancer and endometrial cancer are the most commonly observed phenotypes in Lynch syndrome, along with increased risks, albeit to a lesser degree, of extracolonic cancers—ovary, stomach, small bowel, ureter, and renal pelvis (13–15).

Some studies have suggested that risk of prostate cancer may be increased for men with Lynch syndrome. The majority of this evidence is from studies of prostate cancers in men known or suspected to be carriers of an MMR gene mutation, in which they have shown absence of MMR protein expression in prostate tumors by immunohistochemical (IHC) analysis, and the presence of a substantial change (contractions or expansions) in the length of short repetitive DNA sequences in tumor compared with normal tissue (microsatellite instability; MSI) in prostate tumors by PCR-based analysis (16–21). However, because the proportion of prostate tumors exhibiting MMR deficiency is inconsistent across studies, and MMR deficiency alone does not constitute proof that the tumor was caused by the germline mutation, MMR germline mutation testing in men diagnosed with prostate cancer is not currently recommended.

Attempts have been made to directly estimate the risk of prostate cancer in Lynch syndrome using studies of cancer history of men from Lynch syndrome families (14, 15, 19, 22–36). These studies have generally been limited in terms of size and analytic methods used, so as a consequence, estimates are imprecise and inconsistent. Because of the inconsistent evidence of prostate cancer being part of Lynch syndrome, screening for prostate cancer is not currently recommended for Lynch syndrome families (37). Without being tailored to people at high risk, prostate cancer screening by detection of prostate-specific antigen (PSA) is considerably inaccurate and represents significant harms for an individual (38). If men with Lynch syndrome are shown to be at increased risk for prostate cancer, they may as an identifiable group benefit from such screening. Our aim was to synthesize all published molecular and epidemiologic studies to determine the evidence for prostate cancer risk being increased by a germline mutation in an MMR gene.

We used PubMed (39) to search for all relevant studies of prostate cancer associated with Lynch syndrome that were published by May 9, 2013 using the following combinations of key terms: “Lynch syndrome or HNPCC or MMR” and “prostate cancer or extracolonic cancer or cancer risk.” No language restrictions were imposed. References from relevant articles, letters, reviews, and editorials were searched to identify any additional relevant studies. Studies were reviewed initially on the basis of title and abstracts, and then all full manuscripts for those that appeared relevant were obtained and checked for eligibility. Studies were categorized as either “molecular studies” that reported MMR-deficiency status of prostate cancer tumors of MMR gene mutation carriers or “risk studies” that investigated risk of prostate cancer for confirmed or suspected MMR gene mutation carriers.

Molecular studies were eligible if they analyzed and reported MMR-deficiency status either by MSI and/or immunohistochemistry of prostate cancer tumors in confirmed MMR gene mutation carriers. Molecular studies that did not conduct germline testing and confirm MMR gene mutation were excluded, as were studies that observed these measured molecules in external cell lines.

Risk studies were eligible if they were case–control studies or retrospective or prospective cohort studies that estimated the risk of prostate cancer for men confirmed or suspected to have an MMR gene mutation relative to that for noncarriers or to the general population. Risk studies were excluded if they did not genetically or clinically or pathologically define the sample, if they did not report prostate cancer diagnoses, or limited investigation of risk to specific allelic variants, or were editorials or literature reviews.

For molecular studies, we defined a tumor to be MMR deficient if it had a high level of MSI (>30% of markers tested) or absence of expression of any of the four MMR genes by immunohistochemistry. We defined an MMR gene mutation carrier as confirmed if a pathogenic MMR gene mutation was identified.

For risk studies, subjects were defined as either confirmed MMR gene mutation carriers, or suspected as they were untested and a relative of a confirmed MMR gene mutation carrier, or were a member of a family meeting either the Amsterdam (40, 41) or Bethesda (42) criterions.

For molecular studies, we extracted the following data: year of publication, country in which the study was performed, specific MMR gene(s) mutated, ages at diagnoses of prostate cancer, and the MSI and/or IHC status of the prostate tumors.

For risk studies, we extracted the following data: year of publication, country in which the study was performed, recruitment and ascertainment method of mutation carriers and relatives, mutation status, numbers of confirmed or suspected MMR gene mutation carriers, specific MMR gene(s) mutated, observed and expected numbers of prostate cancers, estimates of prostate cancer risk for mutation carriers compared with noncarriers or the general population [expressed as standardized incidence ratio, standardized morbidity ratio, OR, relative risk (RR), or HR], and absolute or cumulative risk of prostate cancer for mutation carriers (expressed as lifetime risk to a specified age or 10-year risk).

For molecular studies, we divided the total number of MMR-deficient prostate cancers by the total number of prostate cancers in men confirmed to be MMR gene mutation carriers to estimate the proportion of prostate cancers in MMR gene mutation carriers that were MMR deficient. Under the assumption that having a tumor with MMR deficiency meant that the prostate cancer had been caused by the MMR gene mutation, the RR of prostate cancer for men with MMR gene mutations is estimated by RR = N/(N − n), where N is the total number of prostate cancer-affected mutation carriers and n is the number of these for which their tumor exhibited MMR deficiency. The 95% confidence interval (CI) was estimated by assuming that n has a binomial (N;p) distribution with P = n/N.

For risk studies, we used Stata 12.0 (43) to conduct a meta-analysis generating summary estimates by different study designs (clinic-based retrospective cohorts, population-based retrospective cohorts, retrospective cohorts from both population and clinic sources, prospective cohorts, cohorts of mutation carriers with a prior diagnosis of colorectal cancer, and cohorts of all men who potentially carry a mutation from mutation carrying families). Random and fixed effects models were fitted, and heterogeneity was tested between and within groups using I² statistic. A funnel plot was generated to test the evidence of publication bias.

We identified 381 studies through the literature search, and excluded 306 because they were unrelated to prostate cancer in Lynch syndrome (based on the title and abstract). Of the 75 studies that were retrieved for full assessment for eligibility, 18 molecular studies were excluded because 12 studied MMR-deficiency status of prostate cancer in men whose mutation status was unknown, and six studied cell lines; and 34 risk studies were excluded because eight were literature reviews, three were editorials, 19 did not report prostate cancer (they only estimated risk of cancers already known to be associated with Lynch syndrome), and four studied MMR polymorphisms in prostate cancers. A total of 23 studies were included in the systematic review including six molecular studies and 18 risk studies (one contained data suitable for both molecular and risk; Fig. 1).

Of the six molecular studies, two were single case reports (16, 17) and four were case series (18–21) in which prostate cancer tumors of MMR gene mutation carriers were analyzed for MMR deficiency by MSI and/or immunohistochemistry (Table 1). When all case series were combined, 32 of 44 (73%, 95% CI, 57%–85%) prostate cancers in mutation carriers were MMR deficient. Under the assumption that having a tumor with MMR deficiency meant that the prostate cancer had been caused by the MMR gene mutation, the RR of prostate cancer for men with an MMR gene mutation was estimated to be 3.67 (95% CI, 2.32–6.67).

Of the 18 risk studies, six noted the presence of prostate cancers in a Lynch syndrome cohort but did not quantify the risk (Table 2; refs. 25, 26, 28–31). Of the 12 studies that did quantify the risk, six of these reported a statistically significant increase in prostate cancer for Lynch syndrome (19, 27, 32, 33, 35, 36). The remaining six studies found no evidence for an increased risk of prostate cancer (14, 15, 22–24, 34). Three studies estimated prostate cancer risk using retrospective cohort of family members including relatives of confirmed carriers who were not tested themselves for a germline mutation (24, 27, 35). Eight retrospective cohort studies estimated prostate cancer risk only for mutation carriers using population-based resources (22) or clinic-based resources (19, 24, 32, 34–36) or both population- and clinic-based resources (15). Only one study estimated risk using a prospective cohort design; estimating prostate cancer risk for carriers without a prior diagnosis of any cancer (14). Three studies estimated prostate cancer risk for carriers with a prior diagnosis of colorectal cancer (Table 2; refs. 23, 27, 33).

To account for differences in study designs and the populations to which they apply, we conducted a meta-analysis generating summary estimates by study design. There was no evidence of heterogeneity between studies and we therefore present a fixed-effect model (Fig. 2). Retrospective cohorts were also categorized by ascertainment source for analysis. Compared with men from the general population, these studies together showed a 2.13-fold increased risk of prostate cancer (95% CI, 1.45–2.80) for male mutation carriers in clinic-based retrospective cohorts, 2.11 (95% CI, 1.27–2.95) for male mutation carriers with a prior diagnosis of colorectal cancer, and 2.28 (95% CI, 1.37–3.19) for all men who potentially carry a mutation from mutation carrying families. There was no evidence of an increased risk of prostate cancer for male mutation carriers in a prospective study [standard incidence ratio (SIR) 2.49; 95% CI, 0.89–5.87; ref. 14] or a retrospective study using a population-based resource (SIR 2.9; 95% CI, 0.8–7.4; 22) or a retrospective study based from both population and clinic resources (RR 0.92; 95% CI, 0.21–1.63). Funnel plot analysis did not reveal any evidence of publication bias of the studies included in the meta-analysis (Fig. 3).

The mean or median age at diagnosis of prostate cancer for men with Lynch syndrome was reported to be approximately 59 to 60 years in the four studies (14, 19, 26, 32), 64 to 65 years in the four studies (31, 33, 34, 36) and 67 to 69 years in the two studies (15, 35), compared with the median age of 69 years for the U.S. general population (44).

We have attempted to elucidate whether prostate cancer should be considered a Lynch syndrome cancer. Through systematic review, we identified 23 studies that reported data on prostate cancer in Lynch syndrome. Molecular evidence, though sparse, indicates that the phenotypic expression of MMR gene mutation operates at the cellular level in these tumors, and if we assume the only cause for MMR-deficient prostate cancers is the genetic mutation in these carriers, then this evidence is consistent with men with Lynch syndrome being two to three times more likely to develop prostate cancer. Risk studies showed an overall increase in risk of prostate cancer in men with Lynch syndrome with or without a prior diagnosis of colorectal cancer, concurrently supporting the molecular literature.

The two single case studies (16, 17), both documented MSH2 mutation carriers with prostate cancer that displayed high-MSI and absence of MSH2 and MSH6 protein expression through IHC analysis. Both cases were male and had prior cancers. Four studies did provide both MSI/IHC phenotype and MMR mutation genotype (18–21). Overall, 73% of prostate cancers observed in MMR gene mutation carriers were MMR deficient, and absence of MMR protein expression was completely concordant with that of the underlying germline mutation in all cases. Furthermore, the largest case series (20) showed that tumors from MSH2 mutation carriers had a higher proportion of MMR deficiency (83%, 95% CI, 61%–95%) compared with MLH1 (40%, 95% CI, 5%–85%) and MSH6 (25%, 95% CI, 1%–81%) mutation carriers (P = 0.01). They also observed that MMR-deficient prostate cancer tumors had increased levels of tumor-infiltrating lymphocytes compared with MMR-proficient tumors, which is commonly observed in other Lynch syndrome tumors (45, 46). These studies strongly support an association of prostate cancer with MMR gene mutations, especially MSH2. However, they are limited, though in that we are unable through these methods to understand whether MMR gene mutation is causing cancer, modifying its severity, or having its phenotype activated on account of tumorigenesis via another cause.

A strength of this review is that it incorporates epidemiologic risk studies to augment the association showed by molecular studies. We have attempted to quantify risk of prostate cancer associated with MMR gene mutations with consideration of study designs.

Three of the studies included in the meta-analysis (24, 27, 35) estimated prostate cancer risk using retrospective cohort of all family members, that is, including carriers as well as relatives who were not tested for germline mutation status. As a proportion of these untested relatives will on balance of probability be noncarriers, these studies will underestimate the penetrance if it is increased by germline mutation. Of note, one of these studies (35) did not observe a significant increase in risk of prostate cancer.

Eight retrospective cohort studies estimated prostate cancer risk only for carriers using population-based resources (22), clinic-based resources (19, 24, 32, 34–36), or both population- and clinic-based resources (15). Two studies (19, 22) included obligate carriers as well as confirmed. This should be of no consequence, as by definition their carrier status is locked in by the status of other family members relative to their pedigree. However, Grindedal and colleagues (19) also classified as “obligate carriers” those that had MMR-deficient tumors, thus potentially inflating the estimate of penetrance if MMR-deficient prostate cancers can have an alternate etiology.

Given the potential effect of ascertainment on the estimates of risk, we conducted meta-analysis by the stratifying of cohort studies on the source of their samples. The motivation of attending a family cancer clinic may include having a family history of cancer or a recent diagnosis of cancer, and therefore, mutation carriers identified through clinic attendance may have higher risk of cancer. Of the six clinic-based retrospective studies included in the review, four studies (19, 24, 32, 35) did not attempt to correct for the family history ascertainment in their statistical analyses. This implies that the estimates of prostate cancer risk from these studies are likely to be upwardly biased (47) if any of the family members attended the clinics, or genetic testing for MMR gene mutations was instigated because of a family history of prostate cancer. Pande and colleagues (34) attempted to analyze excluding the probands on the basis that they were ascertained “largely because of being cancer affected,” yet neither of the estimates including or excluding probands was statistically significant. In this case, even conditioning on the proband alone may not be sufficient to ensure these populations are representative of all mutation carriers in terms of prostate cancer risk (48).

Two retrospective studies used a statistical method called modified segregation analysis that is capable of producing unbiased estimates of risk by adjusting for ascertainment (15, 36). Raymond and colleagues (36) conditioned on the genotype and cancer status of the proband but did not condition on the cancer status of the relatives, therefore penetrance estimates could be biased if family members attended clinics or were tested for MMR gene mutations because of a family history of prostate cancer. In contrast, Dowty and colleagues (15) conditioned on the genotype and cancer status of the proband and the cancer status of the relatives for clinic-based families, and therefore produced unbiased estimates of penetrance even if attendance to a family cancer clinic or testing for MMR gene mutations was due to prostate cancer. If families' attendance or MMR gene testing was not due to prostate cancer, the estimates from these studies will still be unbiased but the conditioning may incur a loss of power to detect an increased risk of prostate cancer (49).

There is only one population-based retrospective study (22) that estimated risk by comparing the observed number of prostate cancer in confirmed and obligate carriers with the number expected on the basis of their age, and they found no evidence of an increased risk. They ascertained carriers from a population cancer registry and therefore their estimate is less likely to be upwardly biased because of family history. However, due to the wide CIs on the estimate of RR, conclusions from this study alone are limited.

Referral of families to clinics for investigation of Lynch syndrome may also be hampering the investigation of prostate cancer as part of Lynch syndrome. Where entry into family cancer clinics or genetics clinics is contingent upon meeting Amsterdam-I or -II criteria (40, 41), it is likely that there will be a focus on families with higher levels of colorectal and endometrial cancers and some additional cancers, but not prostate cancer. It is possible that research on existing Lynch syndrome families will therefore miss prostate cancer, as it will be underrepresented.

Although difficult to conduct, prospective studies are immune from these recall and ascertainment biases. The prospective study by Win and colleagues (14) did not provide evidence of an increase but was probably underpowered to detect a moderate increase in prostate cancer risk. Future meta-analyses or pooled analyses are required to incorporate additional prospective studies with the advantage of increased power, particularly if the follow-up time is lengthened.

The final study design for which we estimated prostate cancer risk was as a second primary cancer for carriers with a prior diagnosis of colorectal cancer (23, 27, 33). This is a clinically relevant finding given colorectal cancer cases are subject to cancer surveillance. The strength of the association implicates elevated prostate cancer risk in men who are more susceptible to cancer (given their prior diagnosis of colorectal cancer). This type of study overcomes the bias for primarily observing cancers associated with Lynch syndrome classification.

In addition to the studies reviewed in this analysis, there have been four studies examining the role of single-nucleotide polymorphisms (SNP) that are not considered high-risk mutations on prostate cancer. Langeberg and colleagues (50) observed a SNP in MLH1 (rs9852810) was associated with an increased risk of prostate cancer, more aggressive cancer, and prostate cancer recurrence. MLH1 polymorphisms (51) and MSH3 polymorphisms (52, 53) have also been reported to be associated with prostate cancer. These studies do provide some additional evidence to the mutation studies for an etiologic role of MMR genes on prostate cancer risk.

It seems from the evidence provided by molecular studies that testing for MMR protein expression in prostate cancer tumors can be informative for Lynch syndrome in men from a family with a suspected MMR gene mutation where no colorectal tumors are available for testing. However, a formal comparison of prostate tumors from men known not to be MMR gene mutation carrier is needed to confirm the utility of such a test.

The 2- to 3-fold increased risk provided by risk studies may be an important factor in consideration for prostate cancer screening in men with Lynch syndrome. The positive predictive value of PSA screening for prostate cancer in these men may be high enough to warrant recommended screening. Studies on the aggressiveness or nuclear grade of prostate tumor in men with an MMR gene mutation would be relevant to studies on harm-benefit ratio of age-specific screening for such men.

There is some evidence that prostate cancer is more commonly diagnosed in men with an MSH2 mutation compared with men with a mutation in one of the other MMR genes. Except for the study of Rosty and colleagues (20), most studies have been underpowered to observe any differences in prostate cancer risk by specific MMR gene mutations. Large cohorts will be required to measure separate prostate cancer risks for specific MMR gene mutation carriers. Clarification of the associated risk for specific gene mutations will increasingly inform optimal screening strategies of prostate cancer.

Given almost all previously published studies were of Caucasian men, and that prostate cancer incidence varies substantially by country (54), the current findings may not be applicable to non-Caucasian populations.

The molecular and epidemiologic risk evidences in combination present a convincing argument for the inclusion of prostate cancer into the Lynch syndrome cancer spectrum. Certainly the available molecular evidence is consistent with, although does not prove, prostate cancer being caused by an MMR gene mutation. Overall, the published risk studies support a moderately increased risk of prostate cancer for men with an MMR gene mutation, although the evidence is not consistent and many of the studies have limitations in design and analysis. Further research that overcomes these limitations is suggested and warranted to finalize the question of prostate cancer is a Lynch syndrome cancer.

No potential conflicts of interest were disclosed.

A.K. Win was supported by the Picchi Brothers Foundation Cancer Council Victoria Cancer Research Scholarship, Australia. M.A. Jenkins is a National Health and Medical Research Council Senior Research Fellow, Australia.

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.