Material and methods

This study includes women who had been diagnosed with ovarian cancer, belonged to a family with a demonstrated pathogenic mutation in one of the MMR genes MLH1, MSH2 or MSH6, and who fulfilled at least one of the following three criteria:

1. Had tested positive for the mutation and/or

2. Demonstrated loss of the relevant gene product by immunohistochemistry (IHC) of her ovarian cancer or other tumour if mutation testing was not possible and/or

3. Was an obligate carrier in between two relatives fulfilling one of the above two criteria.

Eleven European centres for hereditary cancer contributed all their patients known to fulfil these criteria. In total, 144 women were included in the study. Twenty-one women were included from Oslo, Norway; 6 from Bergen, Norway; 27 from Helsinki, Finland; 16 from Manchester, UK; 6 from Southampton, UK; 4 from Cardiff, UK; 5 from Pamplona, Spain; 10 from Barcelona, Spain; 14 from Milan, Italy, 26 from Leiden, The Netherlands, and 9 from Szczecin, Poland. Some of the cases included from the Netherlands had previously been included in a study by Crijnen et al5 The aim of that study was to compare survival in stage matched ovarian cancer caused by Lynch syndrome and sporadic ovarian cancer. Most of their cases were stage I–II at diagnosis, and they found no difference in survival compared to controls with sporadic ovarian cancer with the same stage at diagnosis.

For each woman information was gathered on date of birth, mutated gene and mutation, date of cancer diagnosis, histological description of the tumour, FIGO (International Federation of Obstetricians and Gynaecologists) stage, last date she was known to be alive or date of death, cause of death and other cancers. Seven out of the 27 women included from Finland had attended a screening programme for gynaecological cancer. Most women included from other centres had not attended a screening programme, but the exact numbers were not available.

All observations were drawn from the medical files in the collaborating centres. All genetic testing was done after informed consent from the patient if alive or by her relevant family member if dead. All testing was subjected to national legislation. No names or personal identifiers were exported from the medical files. No research registry including patient names outside the medical files was established.

Statistics

The Kaplan–Meier algorithm was used to estimate survival functions, starting at the time of ovarian cancer diagnosis.

Observation time was censored at last follow-up. Events were scored as death caused by ovarian cancer, death caused by other HNPCC/Lynch syndrome associated cancer, death caused by other cancers, or as other cause. When calculating ovarian cancer specific survival, deaths from all other causes were scored as no event. When calculating survival for HNPCC/Lynch syndrome associated cancer, ovarian cancer, cancers included in the Amsterdam II criteria, duodenal (n=1), gastric (n=1) and pancreatic cancer (n=1) were scored as event. No deaths from brain tumours were observed. The observed cancer deaths not scored as associated with HNPCC were: two cases of breast cancer, one cancer of the bladder, one sarcoma, and one adrenal gland carcinoma.

Two-by-two tables were assessed by Fisher's exact p. Differences between observed and expected values were subjected to χ² analyses. One-sided t test was used to compare observed versus expected means.

Results

One hundred and forty-four women who had been diagnosed with ovarian cancer and who met the criteria were identified. Fifty-one (51/144=35.4%) had a mutation in MLH1, 78 (78/144=54.2%) had a mutation in MSH2, and 15 (15/144=10.4%) had a mutation in MSH6. The women had been diagnosed to have ovarian cancer in 1946 or later. Mean age of onset was 44.7 (range 20.1–81.9, 95% confidence interval (CI) 43.2 to 46.2), compared to 51.2 years in carriers of BRCA1-mutations with ovarian cancer (p=0.000) and 57.5 in carriers of BRCA2 mutations with ovarian cancer (p=0.000).9

In 9/144 (6.3%) FIGO staging was not available. One hundred and ten out of the remaining 135 (81.5%) had FIGO stage 1 or 2 disease compared to 30.6% of BRCA mutation carriers as previously published (p=0.0000).11

Crude 5, 10, 20 and 30 year survival was 78.6%, 73.3%, 64.3% and 40.3%, respectively.

Five, 10, 20 and 30 year survival specific for deaths due to HNPCC /Lynch syndrome associated cancer were 79.2%, 75.7%, 68.4% and 47.3%, respectively (table 1, figure 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Disease specific survival for Lynch syndrome related cancers including ovarian cancer.

Five, 10, 20 and 30 year survival specific for deaths due to ovarian cancers were 82.7%, 80.6%, 78.0% and 71.5%, respectively (figure 2). Five and 10 year ovarian cancer survival specific for FIGO stage 1 and 2 was 89% and 87%, and for FIGO stage 3 and 4 it was 59% and 53% (p=0.000 for stage 1+2 vs stage 3+4) (table 2, figure 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Ovarian cancer specific survival.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Ovarian cancer specific survival according to FIGO stage.

A total of 29 out of 144 women (20.1%) died of their ovarian cancer. One of the women scored as dead from her ovarian cancer was diagnosed with colorectal cancer some months after her ovarian cancer. It is not known whether this was a new primary tumour or whether the ovarian cancer was a metastasis from the colorectal cancer. Eleven of the 29 women (37.9%) who died of their ovarian cancer died within 1 year, and 21 (72.4%) within the first 3 years. Of the 21 who died within 3 years, 11 (11/21=52.4%) had FIGO stage 1 or 2 disease and 8 (8/21=38.1%) had FIGO stage 3 or 4 disease. Two had unknown FIGO staging. They were diagnosed between 1962 and 2006.

The cancers included in the study were diagnosed over a period of 60 years in seven different countries. Numbers having died from ovarian cancer were considered insufficient to analyse effect of year of birth. We do list some details on histopathology, while being aware of non-homogenous scorings. Seven cases had no description of histology. Of the remaining 137, 129 (129/137=94.2%) were infiltrating. One hundred and thirty-six of the cancers were described as epithelial (including one epidermoid carcinoma, one squamous cell carcinoma, one carcinosarcoma and one pseudosarcomatous carcinoma, one cancer of the fallopian tubes and one Brenner tumour) and one as non-epithelial (granulosa cell tumour). Thirty (30/137=22.9%) of the cancers with histological description were reported to be serous. Of these, 20 (20/30=66.7%) were FIGO stage 1 or 2 and 8 (8/30=26.7%) were FIGO stage 3 or 4. For the last two, FIGO staging was unknown. Five and 10 year survival for serous ovarian cancers was 71.7% and 65.2%, compared to 83.6% and 82.3% for the cancers not described as serous (p=0.15) (figure 4). As mentioned, the non-homogeneous scoring methodology made us consider these findings with caution.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Ovarian cancer specific survival for serous versus non-serous cases.

Seventy-two out of 144 (50.0%) of the women developed at least one concurrent or later cancer included in the HNPCC/Lynch syndrome tumour spectrum. Thirty-three (33/144=22.9%) had a concurrent gynaecological tumour diagnosed with their ovarian cancer. For 128 women (128/144=88.9%), ovarian cancer was their first cancer (one of these was the woman who was diagnosed with colorectal cancer some months after her ovarian cancer diagnosis, as mentioned above).

Discussion

Eighty percent of ovarian cancers in MMR mutation carriers were diagnosed at stage 1 or 2. Ten year ovarian cancer specific survival was as high as 80%. This contrasts with both BRCA associated and sporadic ovarian cancer where 2/3 of ovarian cancers are diagnosed as stage 3–4 and 10 year survival is as low as 36%–47%) (table 2).11–13

About 50% of the women developed other cancers mentioned in the Amsterdam II criteria following their ovarian cancer. Thirty year disease specific survival for HNPCC/Lynch syndrome associated cancers was 47.3% (table 1)—that is, among the ovarian cancer patients about half were long time survivors, and half of those who died actually died from another concurrent or later occurring cancer caused by the MMR mutation.

In the present series, 18.5% of the cancers were diagnosed at stage 3 or 4. As expected, they had a significantly worse prognosis than ovarian cancer diagnosed at stage 1 or 2. Five year survival for advanced stage cancers was 59% compared to 28% in the general population, and 10 year survival was 59%, compared to 19% for BRCA mutation carriers.11 13 Survival for advanced stage ovarian cancer may be better in MMR mutation carriers than in BRCA mutation carriers or in the general population. Crijnen et al reported that survival rates for HNPCC associated cancers were not significantly different from sporadic ovarian cancers matched for age, stage and year of diagnosis.5 Their series, like ours, had mainly early stage cancers and they may have failed to detect a more favourable prognosis in advanced cancers.

Interestingly, 50% of the women who died within the first 3 years of their ovarian cancer diagnosis had stage 1 or 2 disease. Our data did not allow for a close analysis of why these early stage cancers had such a poor prognosis.

In our series only 22% of the cancers were described as serous. We have recently reported that 55.1% in a series of prospectively detected ovarian cancers in BRCA1/BRCA2 mutation carriers were serous.14 According to Hogg and Friedlander,15 high grade serous and undifferentiated cancers dominate among ovarian cancers diagnosed at advanced stage. Serous cancers are therefore likely to be associated with a poor prognosis. However, as many as 66.7% of the serous cancers observed were stage 1 or 2, and survival was not different from the non-serous cases. The observed difference in survival between MMR mutation carriers and BRCA1/BRCA2 carriers may therefore not be explained by a lower prevalence of serous cancers in the present series.

Our study was undertaken to validate the assumed poor prognosis in MMR associated ovarian cancer.

Most of the women included had not attended screening on a regular basis, and the diagnoses had been made over a period of 60 years. Thus, we expected a survival similar or worse than BRCA associated ovarian cancer as reported by Evans et al.11 Surprisingly, we observed that prognosis was better than in BRCA associated cancer. The histopathological details did not indicate any excess of borderline cancers. According to Hogg and Friedlander,15 stage 1 ovarian cancers are biologically different from ovarian cancers of advanced stage, and they question whether ovarian cancer diagnosed at an early stage would inevitably progress to more advanced stages. Our results may support their notion and suggest that germline mutations in the MMR genes and the BRCA genes may predispose to two different types of infiltrating ovarian cancer.

Carriers of BRCA mutations have a lifetime risk of developing ovarian cancer between 25–60%.9 10 Survival is poor, and there are no reports indicating that it is improved through early diagnosis and modern treatment.11 PBSO is therefore recommended after childbearing. There are a few retrospective reports on ovarian cancer risk in Lynch syndrome families. The highest estimated lifetime risk is 12%.1 Other reports present an estimate between 4–8%.2 3 According to Watson et al2 lifetime risk is highest in carriers of MSH2 mutations. Using a lifetime risk of about 10%, our observations of a 20% risk of dying from that cancer if diagnosed provided a lifetime risk of dying of ovarian cancer of about 2%. Whether or not this is a risk high enough to justify PBSO should be discussed.

Findings in a limited retrospective series like ours should be considered with caution and our results need to be validated by others. One question to address in further studies is whether or not Lynch syndrome patients have a favourable prognosis because they are diagnosed at an early stage, or if they as a group have cancers with a more favourable prognosis and which is reflected in an early stage at diagnosis and a good prognosis. Modern treatments including chemotherapy may influence both survival of ovarian cancer, and of other cancers that are diagnosed later on as well. Our study was not designed to address these questions and ideally they should be tackled in prospective series.

In conclusion, the patients with ovarian cancer caused by germline mutations in the MMR genes in the series examined had a better prognosis than ovarian cancer caused by mutations in the BRCA genes or in the general population. We find our results compatible with the notion that the MMR genes may predispose to a biologically different type of ovarian cancer than BRCA1, BRCA2 or in the general population, characterised by early stage and favourable prognosis.