The risk of endometrial cancer (EC) subsequent to a diagnosis of colorectal cancer in women with a germline mutation in a mismatch repair gene [Lynch syndrome or hereditary non-polyposis colon cancer (HNPCC)] is unknown. We estimated the risk of EC following a diagnosis of colorectal carcinoma (CRC) for women with Lynch syndrome. A retrospective cohort study was performed on women diagnosed with CRC with a germline mutation in a mismatch repair (MMR) gene (Lynch syndrome cases), and women with microsatellite stable (MSS) CRC who were not known to carry a germline mutation (non-Lynch cases), identified from the Colon Cancer Family Registry. The incidence of EC following CRC was estimated and compared for women with and without Lynch syndrome, using adjusted hazards ratios calculated for time at risk among each group. A total of 112 women with Lynch syndrome and a previous diagnosis of CRC were compared with 908 women without Lynch and with a MSS CRC diagnosis. The estimated 10-year cumulative risk of EC subsequent to CRC was 23.4% [95% confidence interval (CI): 15–36%] for Lynch syndrome women compared with 1.6% (95% CI: 0.7–3.8%) for non-Lynch women. After adjusting for ascertainment, age at diagnosis and diagnosis of other cancers, risk of subsequent diagnosis with EC was elevated sixfold in women with Lynch syndrome compared with non-Lynch women (HR 6.2; 95% CI 2.2–17.3; p = 0.001). Approximately one quarter of women diagnosed with Lynch syndrome-associated CRC developed EC within 10 years. This supports the sentinel cancer concept and suggests that active and early management is important for these women.
Lynch syndrome [also known as hereditary non-polyposis colon cancer (HNPCC)] is caused by a germline mutation in one of several DNA mismatch repair (MMR) genes, MLH1, MSH2, MSH6 or PMS2. Carriers of MMR gene mutations are at increased risk of cancers of the colon, endometrium, ovary, upper urologic tract, stomach, small bowel, biliary/pancreas, skin and brain.1 In individuals who carry such mutations, inactivation of the remaining normal allele in a cell results in dysfunctional or absent DNA MMR. These unrepaired mismatches typically occur in regions of repetitive nucleotide sequences, commonly referred to as microsatellites. The consequent phenotype, referred to as microsatellite instability (MSI), is the signature change in Lynch syndrome-associated cancers.2, 3 Detection of a deleterious germline mutation in a particular MMR gene in a case of cancer with MSI unequivocally establishes the diagnosis of Lynch syndrome, with MLH1 and MSH2 sequence variants accounting for ∼80% of all MMR gene mutations.4
Early age of cancer diagnosis and multiplicity of cancers are considered hallmarks of Lynch syndrome. A meta-analysis5 of data from three population-based studies,6–8 and previous clinic-based work,9 estimated that the risk of colorectal cancer for MLH1 and MSH2 carriers to age 70 years was 53% for males, 33% for females and for endometrial cancer (EC) was 44%. Although the increased risk of EC in women with Lynch syndrome is well established, the risk of EC after Lynch-associated colorectal cancer is unknown. Possible reasons for a change in EC risk after colorectal cancer diagnosis include, but are not limited to: effects of radio- or chemo-therapy for the colorectal cancer, increased surveillance, weight change and changes in exposures to environmental risk factors. A number of attempts have been made to address this question, but the relationship could not be estimated with sufficient precision due to the limited number of cases analyzed or the design of the original study.10–12 If women with Lynch syndrome-related colorectal cancer have an increased risk for subsequent primary gynecological cancer, they could be advised to undertake interventions such as screening13, 14 and/or risk-reducing surgery to prevent subsequent primary cancers.
Hence, the aim of this study was to estimate the risk of EC subsequent to a diagnosis of colorectal cancer in women with Lynch syndrome. To achieve this aim, we analyzed cancer risk using a large cohort of MMR mutation carriers from the international Colon Cancer Family Registry (Colon CFR).15
Material and Methods
Subjects for this retrospective cohort study were participants from the Colon CFR. The Colon CFR consists of data and biospecimens from clinic- and population-based colorectal cancer cases and controls as well as their families. Details of recruitment methods have been described previously.15 Briefly, clinic-based ascertainment was as follows: individuals with colorectal cancer were selected from multiple-case colorectal or Lynch syndrome cancer families attending cancer family clinics in the USA (Mayo Clinic, Rochester, MN; and Cleveland), Australia (Melbourne, Adelaide, Perth, Brisbane and Sydney) and New Zealand (Auckland). For population-based ascertainment, probands were defined as incident colorectal cancer cases identified by population-based cancer registries in the USA (Puget Sound, WA; the State of Minnesota; Los Angeles County, CA; AZ; CO; NH; parts of NC; and HI), Australia (Victoria) and Canada (Ontario). Probands were asked for permission for the study to contact selected relatives for participation.
For probands and participating relatives, information on demography, personal and family history of cancer, cancer screening, lifestyle factors, cancer surgery and gynecological surgery were obtained by interview, questionnaire or extraction from clinical records. Efforts were made to verify reported cancer diagnoses using family reporting, pathology reports, medical records and death certificates. All probands and participating relatives were asked to provide a blood sample for DNA analysis and for consent to retrieve archived colorectal cancer tissue. Women with both colorectal carcinoma (CRC) and EC were not more likely to be eligible for the study and were not more likely to be tested for MMR mutations than women with CRC only.
Participants were followed-up and reinterviewed ∼4 to 5 years from the date of the baseline interview. They were also asked to report any new cancers or deaths in relatives including those who had participated at baseline and were unable to participate in the follow-up interview or who had died.
Colorectal cancer tumor testing for MSI and germline testing for MMR gene mutations
Colorectal cancers were tested for MSI using a panel of ten markers as previously described.16 If one or more markers tested displayed a difference between normal and tumor tissues, the tumor was defined as MSI; otherwise, the tumor was classified as microsatellite stable (MSS). Germline mutation testing was conducted for three MMR genes (MLH1, MSH2 and MSH6) by DNA sequence analysis, and large insertion deletions were screened by using multiplex ligation-dependent probe amplification.17 A deleterious mutation was defined as a variant predicted to result in a stop codon, a frameshift mutation, a large insertion or deletion or a missense mutation judged to be deleterious. All relatives of identified mutation carriers who provided a DNA sample were subsequently tested for the same variant.
For this study, Lynch syndrome cases were defined as females with a previous diagnosis of colorectal cancer who had a documented deleterious germline mutation in MLH1, MSH2 or MSH6. Non-Lynch cases were defined as females with a previous diagnosis of colorectal cancer that was MSS and not known to have a deleterious mutation in MLH1, MSH2 or MSH6. Cases were excluded if: the date of diagnosis of colorectal cancer or EC, age at diagnosis, age at death or age at hysterectomy were missing or incomplete; data had only been provided by proxy; or if EC was diagnosed, or a hysterectomy occurred, before the first diagnosis of colorectal cancer. We also excluded any MSI cases for whom testing for MMR gene mutations was not available. Details of the selection process for this analysis are given in Table 1.
Table 1. Selection process for Lynch1 and non-Lynch groups2
Chi-square tests (or Fisher's exact test when cell sizes were small) were used to assess the differences between women with Lynch syndrome and those without for a number of key demographic and clinical characteristics (as given in Table 2). Among women with Lynch syndrome, chi-square tests were also conducted to determine whether differences in the mutated MMR genes involved were statistically significant.
Table 2. Cohort characteristics by study group
The principal outcome of interest was EC. For each participant, follow-up time was calculated by subtracting the date of first colorectal cancer diagnosis from the date of death or last known date alive. Time at risk was derived from follow-up time by truncating at date of diagnosis of EC or censoring at date of hysterectomy if either of these events occurred. If only year and month were known for any relevant date, then the 15th of the month was used as an approximation. If only year was known, then June 30th was used as an approximation.
Median values were calculated for age at diagnosis of first colorectal cancer, follow-up time and time at risk for the Lynch and non-Lynch groups and differences in the distributions were evaluated using the non-parametric Wilcoxon–Mann–Whitney test.
Directly age-standardized incidence rates for EC following colorectal cancer were calculated for both the Lynch and non-Lynch groups, based on the American Standard Population (year 2000) and using broad age categories (<50 years, 50–59 years and 60 years and over).
Kaplan–Meier curves were used to assess the differences in the cumulative risk of being diagnosed with EC following colorectal cancer between the Lynch and non-Lynch groups, with a log-rank test used to test for equality of the curves. The analysis was restricted to 10 years time at risk because of insufficient data to calculate reliable estimates beyond that time.
A Cox proportional hazards model was fitted to time at risk to calculate hazard ratios for the development of EC among cases with Lynch syndrome compared with the non-Lynch group. To account for familial clustering, a robust variance estimation approach was used. The model was adjusted for age group at time of first diagnosis with colorectal cancer (under 50 years vs. 50 years and older) as well as presence of cancers other than colorectal cancer or EC and ascertainment type (population-based vs. clinic-based). Explanatory variables were tested individually and collectively for proportionality before being included in the model.
All analyses were performed using the SAS statistical package, version 9.1 for Windows (SAS Institute, Cary, NC).
Characteristics of the cohort are summarized in Table 2. There were 112 women with a previous diagnosis of colorectal cancer who had Lynch syndrome (Lynch group) and 908 women with a previous diagnosis of colorectal cancer who did not have Lynch syndrome (non-Lynch group). Approximately one-third (32.1%, n = 36) of the women with Lynch syndrome and 1.7% (n = 15) of the non-Lynch women had at least one other relative included in this analysis. The Lynch group were generally younger when first diagnosed with colorectal cancer, with a median age of 39 years compared with 54 years for the non-Lynch cases (p < 0.001). Few participants reported using tamoxifen–only one of the Lynch group (0.9%) and 15 (1.7%) from the non-Lynch group.
After a diagnosis of colorectal cancer, almost half of the cases (44.6%) with Lynch syndrome had a hysterectomy, compared with only 4.1% of the non-Lynch group (p < 0.001). The reasons for those hysterectomies were not stated. None of the 50 cases in the Lynch group who had a hysterectomy died within 5 years of follow-up. Among the 62 Lynch group cases that did not have a hysterectomy, six (9.7%) had died within 5 years.
About 1 in 5 of the women in the Lynch group were subsequently diagnosed with EC (n = 22, 19.6%) compared with 0.9% (n = 8) in the non-Lynch group (p < 0.001). Of the 30 reported EC diagnoses in both groups, we were able to verify 23 (76.7%) by either pathology reports or medical records.
Two of 112 (1.8%) cases of the Lynch group developed ovarian cancer subsequent to colorectal cancer compared with three of 908 (0.3%) in the non-Lynch group. The proportion of cases diagnosed with cancers other than endometrial or ovarian cancer subsequent to colorectal cancer was significantly higher in the Lynch group (17.0 vs. 5.8%; p < 0.001).
There were a total of 1,059 person years at risk among the Lynch group and 5,383 person years at risk among the non-Lynch group. Although differences in the distribution of time at risk for EC between Lynch and non-Lynch CRC cases were statistically significant (p = 0.011), the median time at risk was similar for the two groups (6.3 years and 6.0 years, respectively).
For the Lynch group, there was no significant difference (p = 0.284) in the MMR gene that was mutated for those who were subsequently diagnosed with EC and those who were not (Table 3). The majority of the deleterious mutations occurred in either the MLH1 gene (50.0% for both those with and without EC) or the MSH2 gene (40.9% for those with EC and 47.8% for those without subsequent EC). In the group of mutation carriers who developed EC, two women (9.1%) had a MSH6 mutation compared with two women in the group who did not (2.2%).
Table 3. Subsequent diagnosis of endometrial cancer by type of deleterious mutation for colorectal cancer cases with Lynch syndrome
Crude incidence rates of EC following colorectal cancer were 20.8 cases per 1,000 years at risk for women in the Lynch group and 1.5 cases per 1,000 years at risk for women in the non-Lynch group. Corresponding age standardized rates were significantly higher in the Lynch group with 18.6 cases per 1,000 years at risk (95% confidence interval (CI) = 11.7–28.2) in contrast to 1.1 cases per 1,000 years at risk (95% CI = 0.5–2.1), respectively (p < 0.001).
The 10-year cumulative risk of EC following colorectal cancer in the Lynch group was 23.4% (95% CI = 14.5–36.3%) compared with the corresponding risk of 1.6% (95% CI = 0.7–3.8%; p < 0.001) for the non-Lynch group (see Fig. 1).
After adjusting for age at colorectal cancer diagnosis, presence of cancers other than colorectal or EC and ascertainment type (Table 4), it was estimated that women in the Lynch group were about six times more likely to be diagnosed with EC following colorectal cancer (hazard ratio = 6.2; 95% CI = 2.2–17.7, p = 0.001).
Table 4. Adjusted hazard ratios for endometrial cancer following colorectal cancer
We found that for every 100 women diagnosed with Lynch syndrome-associated colorectal cancer, around 23 (95% CI = 15–36) will be diagnosed with EC within 10 years if they do not have a hysterectomy. This compares with less than two women being diagnosed with EC for every 100 women who are diagnosed with colorectal cancer who have not inherited a mutation in a MMR gene, although the non-Lynch group available for our study were somewhat younger on average than is generally the case.
The concept of a “sentinel” cancer suggests that Lynch carriers are at increased risk of developing subsequent primary cancers following an initial (sentinel) cancer diagnosis compared with non-Lynch carriers. Two other research groups suggested that the risk of a second primary cancer in Lynch carriers is 25 and 50% by 10 and 15 years, respectively, after diagnosis of a first cancer.18, 19 In a series of 117 women with dual primary (endometrial and colon) cancers from 223 Amsterdam families, 16 women had synchronous cancers.20 Of the remaining 101 women, 52 developed an endometrial or ovarian cancer first, whereas 49 cases were diagnosed with colorectal cancer first and had a second primary cancer subsequently. Hence, colorectal cancer can be seen as a sentinel event in approximately half of women with Lynch syndrome affected with multiple malignancies, and EC represents the sentinel cancer in the remainder, a finding consistent with the present report. Our data confirm the sentinel cancer concept; the risk of subsequent EC in cases with Lynch syndrome on the occasion of a colorectal cancer event is a significant increase over what would be expected in unselected cases.
Clinically, the concept of sentinel cancer is potentially very helpful not only for affected individuals but also for their relatives. The proportion of women with EC diagnosed at or before the age of 50 years who carry germline mutations in a MMR gene varies between 9 and 18%.21, 22 Current evidence supports colorectal screening in those cases. In one observational analysis, screened cases had more than a 60% reduction of colorectal cancer incidence compared with unscreened cases.23 Data from the United Kingdom suggested a 72% reduction in mortality attributed to screening colonoscopy for individuals with a strong family history of colorectal cancer,24 and women diagnosed with Lynch syndrome-related EC are advised to have regular colorectal and other screening postoperatively.
Until recently, no population-based data were available regarding the risk of EC following a Lynch syndrome-associated colorectal cancer diagnosis. Current recommendations are based on the extrapolation of evidence from various contexts. This is the first study of sufficient size to quantify the risk of EC following Lynch syndrome-related colorectal cancer, thus supporting early and active management in those young women.
Recent reports have addressed the efficacy of EC screening in Lynch carriers. In a series of 35 premenopausal and six postmenopausal women who were MMR gene mutation carriers or who fulfilled the Amsterdam criteria and mismatch gene mutation status was unknown, three premalignant lesions were detected within a total of 197 patient years at risk.25 In addition, one patient was diagnosed with interval EC based on her symptoms. Generally, it is believed that endometrial sampling/biopsy is superior to transvaginal ultrasound alone for EC surveillance in premenopausal women.13, 14 In postmenopausal women, the threshold for endometrial thickness on transvaginal ultrasound has been analyzed several times and ultrasound might be equivalent to endometrial sampling for that group of patients.26, 27 In a multi-institutional retrospective cohort analysis, Schmeler et al. investigated 315 women with documented germline MMR mutations.28 A total of 61 women had undergone prophylactic risk-reducing hysterectomy and 47 women had undergone prophylactic bilateral salpingo-oophorectomy. These women were matched with mutation-positive women who had not undergone prophylactic surgery. The prophylactic surgery group experienced no occurrences of endometrial or ovarian cancer, comparing favorably with 33% EC and 5% ovarian cancer in the control group. The results from this study suggest that prophylactic hysterectomy with bilateral salpingo-oophorectomy is effective in preventing gynecological cancer in women with Lynch syndrome.
Although 100% of EC could be prevented by prophylactic surgery, this may or may not translate into survival benefits. Given that the incidence of ovarian cancer is consistently low in Lynch syndrome series,29, 30 and given the normally excellent prognosis of clinically diagnosed EC,31 the actual survival benefits may be modest at best. To date, only case reports32–34 but no data from cohort series or clinical trials are available on the use of a levonorgestrel intrauterine device and its potential ability to prevent EC.
Clinically, one of the major research questions is: “Should women diagnosed with Lynch-associated colorectal cancer require specific management addressing this risk?” Several outcome parameters are relevant to this important question. Our study focused on the incidence rates of EC and we demonstrate a high risk of developing EC after Lynch-associated colorectal cancer but only five cases of ovarian cancer were diagnosed. Two of those five women had Lynch syndrome and three were non-Lynch. Unfortunately, we are unable to answer whether screening or prophylactic, risk-reducing hysterectomy will improve survival.
Because the recruitment to the Colon CFR was weighted toward early-onset colorectal cancer,15 the majority of women in this series (96% for Lynch patients and 62% for non-Lynch) were diagnosed with colorectal cancer before the age of 60 years. Inference of these findings to cases diagnosed at a relatively young age is therefore straightforward, but a potential limitation may be the inability to infer our findings to cases diagnosed at an older age.
Another potential limitation of this study relates to the Colon CFR recruitment process. Although no clinic-based families were ascertained because of a case of EC following a colorectal cancer, some families may have been recruited due to EC occurring in a relative; hence, for the estimate of increased risk over non-Lynch women, we adjusted for ascertainment type. However, this recruitment strategy is unlikely to have resulted in biased estimates as women with both CRC and EC were not more likely to be eligible for the study and were not more likely to be tested for MMR mutations than women with CRC only. Less than a third of all cases completed their baseline interview within 1 year of being diagnosed with colorectal cancer, meaning that those who died within the 1st year were less likely to be recruited to the study than those who survived longer. It is unclear what effect excluding these women would have on the calculation of cumulative risk of EC.
In summary, our article suggests that about one quarter of cases who were diagnosed with Lynch syndrome-associated colorectal cancer who do not have a hysterectomy will develop EC within 10 years. Given the extent of that risk, identification of these patients and proactive management is justified to reduce the burden of this cancer. Management options include risk-reducing, prophylactic total hysterectomy and bilateral salpingo-oophorectomy28 or regular screening with endometrial sampling on a yearly basis for premenopausal women or transvaginal ultrasound examinations for postmenopausal women.
The Colon CFR work was supported by the National Cancer Institute, National Institutes of Health under RFA #CA-95-011 and through cooperative agreements with the members of the Colon Cancer Family Registry (CFR) and principal investigators.