Tel.: +46-73-684-1243, Fax: +46-8-31-49-75
Endometriosis as a prognostic factor for cancer survival†
Article first published online: 2 DEC 2010
Copyright © 2010 UICC
International Journal of Cancer
Volume 129, Issue 4, pages 948–955, 15 August 2011
How to Cite
Melin, A., Lundholm, C., Malki, N., Swahn, M.-L., Sparen, P. and Bergqvist, A. (2011), Endometriosis as a prognostic factor for cancer survival. Int. J. Cancer, 129: 948–955. doi: 10.1002/ijc.25718
The article is presented at World congress of Endometriosis, Sydney Australia 2009, Nordic congress of Endometriosis, Trondheim, Norway 2009 and ESHRE congress, Rome, Italy 2010.
- Issue published online: 24 JUN 2011
- Article first published online: 2 DEC 2010
- Accepted manuscript online: 14 OCT 2010 12:30PM EST
- Manuscript Accepted: 21 SEP 2010
- Manuscript Received: 19 MAY 2010
- Nordic Cancer Union, NCU
- cancer survival;
- prognostic factor
Studies have shown an increased risk of malignancies in women with endometriosis. Little is known about the impact of endometriosis on cancer survival. We investigated whether the survival after a diagnosis of a malignancy differs in women with a previously diagnosed endometriosis compared to other women. Women with a first time diagnosis of a malignancy in 1969–2005, were identified using the National Swedish Cancer Register (NSCR). By use of the National Swedish Patient Register (NSPR) we identified all women with a diagnosis of endometriosis during the same period and linked these patients with the data from the NSCR. The cohort comprised 4,278 women with endometriosis and a malignancy, and 41,831 randomly selected matched women without endometriosis. Cox regression was used for all calculations to obtain crude and adjusted cause specific mortality rates, measured as hazard ratios (HR) with 95% confidence intervals (CI). A total of 46,109 women entered the study. There was a statistically significant better survival for women with endometriosis for all malignancies combined (HR = 0.92) and for breast cancer (HR = 0.86) and ovarian cancer (HR = 0.81) specifically. For breast cancer the survival enhancing effect in women with endometriosis decreased with increasing parity. There was poorer survival in malignant melanoma for women with endometriosis (HR = 1.52). The survival in a malignancy is better in women with a previously diagnosed endometriosis compared to women without endometriosis especially for breast and ovarian cancers. The prognosis of malignant melanoma is poorer in women with endometriosis.
Women with endometriosis have an increased risk for several types of malignancies.1–6 Ever since the 1920s when Sampson first described tumor growth at the same location as endometriosis, several clinical, histological and epidemiological studies have shown an association between different types of malignancies and endometriosis.7–12 Endometriosis is defined as a chronic, inflammatory, estrogen dependent disease, affecting 5–15% of all women of fertile ages, and a common cause of infertility.9, 10
Between two and four percent of women with endometriosis are postmenopausal. The main symptoms are dysmenorrhea, dyspareunia and also non-cyclic lower abdominal pain.
Endometriosis is a benign disease which in many ways behaves as a malignancy with genetic polymorphism, loss of control of cell proliferation, infiltrative growth and even multiple dissemination both locally and distantly.11 The coexistence of endometriosis and cancer has been estimated to occur in 0.7–5.0% of all cases with ovarian endometriosis.7, 12–14 Ovarian cancer is the most common type of malignancy reported in association with endometriosis.
The knowledge about endometriosis as a prognostic factor for survival in a malignancy is poor. However there are indications that endometriosis might have an impact on the prognosis of malignancies.15–17 The biological mechanism behind this is not clear.
The aim of this study was to investigate, in a population based sample, whether the survival after a diagnosis of a malignancy differ in women diagnosed with endometriosis compared to women without this diagnosis.
Material and Methods
We identified all Swedish women with a first diagnosis of one of 18 different types of malignancies between 1969 and 2005 from the National Swedish Cancer Register (NSCR), using the Swedish National Registration Number (NRN), which is unique to every Swedish resident. The NSCR is based on the International Classification of Diseases (ICD)-7 codes for diagnoses of malignancies. We limited our study to include 18 of the most common and for this study most relevant types of malignancies.
To create our study cohort, cancer cases from the NSCR were linked with data from the National Swedish Patient Register (NSPR), by use of the NRN, identifying those women who also had been discharged from a hospital with a diagnosis of endometriosis between 1969 and 2005. The diagnoses for endometriosis were coded according to ICD 8, 9 and 10. For ICD 8 the codes 625.30–625.33, 625.38 and 625.39 were used, for ICD 9 the codes 617A-617G and 617X, and for ICD 10 the codes N80.0-N80.9.
Patients diagnosed within an open ward system, in private practice or as a day surgery procedure, were not included in this study.
A total of 4,309 women were identified as having both a hospital discharge diagnosis of endometriosis for the first time and a first diagnosis of one of the 18 different types of malignancies and were included in this study. Only women who had their malignancy diagnosed 30 days or more after their endometriosis diagnosis were included to avoid the risk of including accidentally diagnosed cases of cancer that were discovered at time of endometriosis surgery, because these cancers might have been diagnosed in an earlier stage than they would have been without the endometriosis surgery and therefore influence the survival. These women constituted the exposed individuals in our cohort. For each of these 4,309 women, we randomly selected up to a maximum of 10 women from our study population that did not have a hospital discharge diagnosis of endometriosis from the NSPR. These women constituted the unexposed individuals of our cohort. The unexposed women were matched for year of birth (±2 years) and living in the same county as the corresponding exposed woman at the time of her hospital discharge diagnosis of endometriosis. Moreover, they had to be diagnosed with the same type of malignancy and the date of the diagnosis of the malignancy had to be at least 30 days after the corresponding exposed woman's date of endometriosis diagnosis.
We excluded 20 exposed women because no matching unexposed women could be found. The main reasons for this were rare types of malignancies in combination with living in scarcely populated areas of Sweden. Another 105 women, 94 unexposed and 11 exposed, were excluded because date of the diagnosis of a malignancy and date of death were the same.
In the end, 4,278 exposed women and 41,831 unexposed women constituted our study cohort, with 1–10 unexposed women matched to each exposed woman.
Data on parity were collected from the multi generation register (MGR). This register comprises all individuals registered in Sweden from 1961, and born in 1932 or later. By use of the Swedish NRN, familial relationships (father, mother, children and siblings) between individuals in the MGR can be established and data on parity was calculated.
Data regarding time of death and cause of death were collected from the Causes of Death Register (CDR) (Fig. 1).
Information on stage and histological subtype at the diagnosis of the malignancy were not included in the NSCR before 2005. However, we were able to retrieve this information regarding ovarian cancer from three of six Regional Cancer Registers in Sweden (Stockholm region, the Northern region and the Western region). In these regional registers, stage and histological subtypes were classified according to the International federation of gynecology and obstetrics (FIGO)-classification of gynaecological cancers.18 Information on stage of disease and histological subtype was identified in 64 exposed and 154 unexposed women with ovarian cancer.
Other types of malignancies
Only the most common types of malignancies and for this study the most relevant types were included. We used the statistics from the NSCR and the results from our previously published studies to decide on which type of malignancies that were the most common ones and/or could be relevant for the connection to endometriosis.3, 19, 20 In Sweden the five most common types of malignancies in women are breast cancer, colon cancer, lung cancer, endometrial cancer and skin cancer (not including malignant melanoma). Malignant melanoma is the sixth most common malignancy, rectal cancer the seventh and ovarian cancer the eight most common malignancies in Swedish women. About 30% of all cases of malignancy in Swedish women are breast cancer and this fact is shared by many countries in the industrialized part of the world.19 Skin cancer (not including malignant melanoma) is a cancer that affects mainly older women, with a low mortality and has not previously been shown to be associated with endometriosis and/or reproductive factors and was therefore not included in this study. Rectal cancer has not been shown to be connected to endometriosis and was not included in this study. In our previous studies we have shown an increased risk for kidney cancer and endocrine cancers in women with endometriosis and these types were therefore included in this study.3, 20
Each woman was followed from the date of the diagnosis of the malignancy until she died, emigrated or until the end of year 2005, whichever came first.
We identified the following potential confounders; age and calendar year at the malignancy diagnosis, parity, stage and histological subtypes for ovarian cancer and location for malignant melanoma. Each of these variables were categorised in the following way: age was categorised as 0–54 years, ≥55 years; calendar year as 1969–1989, 1990–1999, 2000–2005; parity as 0, 1–2, and 3 or more children; stage as FIGO 1A-C, FIGO 2A-C, FIGO 3A-C and FIGO 4. Histological subtypes were divided into five categories according to the FIGO classification; serous, mucinous, endometrioid, clear-cell and other. Location of malignant melanoma were categorised as head, trunk, arms or legs, multiple location or unspecified location. Differences between exposed and unexposed women regarding age at the diagnosis of the malignancy, parity, calendar time period at malignancy diagnosis and follow up time were tested using the Students t test for continuous variables and χ2-test for categorised variables.
Cox regression models, which adjust for follow-up time, were used for all analyses to obtain crude and adjusted hazard ratios (HR) and 95% confidence intervals (CI). Cause specific mortality rates were used in the analyses only counting events where the person died from the same type of malignancy that was diagnosed at the inclusion of the study. All analyses were adjusted for age at diagnosis of the malignancy, parity and calendar time at malignancy diagnosis and stratified on matching strata to account for the study design. For malignant melanoma we performed the analyses adjusted for location and for ovarian cancer we made separate analyses for stage at cancer diagnosis and for histological subtype for the subgroup of women where we had obtained this information. Analyses were performed both for all malignancies together and for each type of malignancy separately.
To study a potential modification of the association between endometriosis and cancer survival, we tested interaction terms between the exposure variable and age at malignancy diagnosis, calendar time at malignancy diagnosis and parity.
In a subgroup analysis of the exposed women only, we also investigated the impact of age at endometriosis diagnosis on survival after a diagnosis of a malignancy as well as the impact of time lapse between endometriosis diagnosis and the malignancy diagnosis on survival, utilizing Cox regression models. These analyses were adjusted for parity and calendar time for malignancy diagnosis.
All statistical computations were done using STATA 10.1.
The Regional Ethics Board at the Karolinska Institutet, Stockholm approved the study.
The average age at first time diagnosis of a malignancy was 56.4 years (SD = 10.4 years) for women with endometriosis and 56.7 years (SD = 10.4 years) for the other women (p = 0.06). There were no statistically significant differences at age at malignancy diagnosis for any of the specific malignancies studied (Table 1). There was neither any statistically significant difference in the distribution of the year of malignancy diagnosis between the two groups. However, there were fewer births among women with endometriosis, compared to the other women (p < 0.001). The mean age at endometriosis diagnosis was 43.8 years (range 16–85 years) and it constantly increased over the study period, from 41.4 years in 1968–1979 to 51.4 years in 2000–2005. The mean follow up time was 6.4 years for women with endometriosis and 6.2 years among the other women (p = 0.11).
Women with endometriosis showed overall a better prognosis after a malignancy diagnosis, compared to women without endometriosis, HR = 0.92 (95% CI = 0.86–0.98) (Table 2). Specifically they showed a better prognosis for breast (HR = 0.86, 95% CI = 0.75–0.97) and ovarian cancer (HR = 0.81, 95% CI = 0.65–1.01). Contrary to this, women with endometriosis showed a worse prognosis after a diagnosis with malignant melanoma (HR = 1.52, 95% CI = 1.02–2.27) (Table 2). Figure 2 displays smoothed hazard estimates for all malignancies, breast cancer, ovarian cancer and malignant melanoma, illustrating the risk of cause-specific mortality for women with and without endometriosis. The hazards for the two groups follow each other fairly well over time and the tests for non-proportional hazard was statistically significant only for endometrial and cervical cancer. These two malignancies had very few fatal cases and these estimates should be interpreted with caution.
Regarding effect modification, there was a statistically significant interaction between endometriosis diagnosis and age at malignancy for all malignancies combined, p = 0.04, such that the better prognosis for women with endometriosis was more pronounced for those with a diagnosis of a malignancy at 55 years of age or later, than if diagnosed at younger age. A similar association was seen for pancreatic cancer (p = 0.04, data not shown) and ovarian cancer (p = 0.03), and also for breast cancer, although not statistically significant (p = 0.06) (Table 3). There was no statistically significant difference in prognosis between women with and without endometriosis and ovarian cancer when stratifying for stage and histological subtype in the available subgroup (data not shown).
For breast cancer, there was an interaction between endometriosis and parity (p = 0.03) with lower HR for exposed nulliparous women than for women who had given birth. Thus the favourable prognosis in breast cancer for women with endometriosis was more pronounced among nulliparous than among parous women (Table 3).
There were no significant interactions between diagnosis of endometriosis and calendar time at malignancy diagnosis, for any type of malignancy.
However, although not statistically significant (p = 0.11), it is worth noting that, for malignant melanoma the highest HR between women with and without endometriosis was found among those diagnosed with malignancy during the 1990s (HR = 2.40 95% CI = 1.33–4.34) (Table 3).
Neither was there any significant associations found between age at endometriosis diagnosis and survival, nor for the time lapse between endometriosis diagnosis and the malignancy diagnosis for any of the malignancies studied (data not shown).
To our knowledge, this is the largest population based cohort study published concerning survival in women with endometriosis after a diagnosis of a malignancy and also the only study including other types of malignancies than ovarian cancer. The survival was higher after a malignancy diagnosis in women with endometriosis as compared to women without endometriosis, above all for breast cancer and ovarian cancer. However, for malignant melanoma there was a lower survival for women with endometriosis. We found that age at malignancy diagnosis was an important modifier of the effect of endometriosis on survival in all malignancies combined, as well as for pancreatic and ovarian cancer, with a better prognosis in the postmenopausal age group. There was also a similar tendency for breast cancer, however not statistically significant.
Strengths of this study are the large number of women included, the long follow-up time and the ascertainment of cancer diagnosis, parity and time of death. Another strength is the fact that our analyses are cause specific as only death in the same malignancy as diagnosed initially is counted as an event.
A general problem in studies like this is related to the endometriosis diagnosis, i.e., only women who have been omitted to hospital, stayed overnight and then discharged with a diagnosis of endometriosis were included. Because diagnostic procedures have changed over time, especially toward more day surgery laparoscopies in the last 15 years, presumably the moderate to severe cases with endometriosis, rather than the more mild cases are included in this study. The mean age at endometriosis diagnosis in this material increased with calendar time, which can probably be explained by changes in the diagnostic routines and the fact that we only include women who have been diagnosed with a malignancy after the endometriosis diagnosis. The younger women who have been diagnosed with endometriosis during our study period may still be too young to have developed a malignancy. Consequently we also had a rather high mean age at endometriosis diagnosis. Another reason might be that the awareness of endometriosis has increased in the medical society and that the diagnosis is made also in older postmenopausal women. One could argue that the better survival rates in women with endometriosis and ovarian cancer is because of the fact that these women more often visit a gynaecologist for examinations. If this is the case, it might have an impact on the date for diagnosing the malignancy and thus possibly the tumor stage and prognosis. For this reason we performed a stratified analysis on stage and histological subtype. However, we were only able to retrieve this information on 7.8% of all women with ovarian cancer included in this study and therefore no real conclusions could be drawn from this analysis.
Prognostic factors for malignancy survival
Very little is known about the impact of endometriosis on cancer survival. Most published studies have focused on ovarian cancer survival, where endometriosis has been indicated to be a positive prognostic factor,15–17, 21 In a nested case control study women with endometriosis and ovarian cancer were shown to have more favorable prognosis. This might have been attributed to the fact that the women with endometriosis had a lower stage of cancer, lower tumor grade and a different distribution of histological subtypes. The women with endometriosis were also on average younger at time of ovarian cancer diagnosis which could have contributed to a better prognosis.15 This finding was confirmed in one of our previous studies where we showed that women with endometriosis are diagnosed with ovarian cancer earlier in life than other women.3 However, our study shows a better prognosis in women with endometriosis and ovarian cancer in the age group 55 years and older indicating that even though women with endometriosis are diagnosed with ovarian cancer earlier in life than other women, the effect on prognosis is higher in the older age groups.
Reproductive and hormonal factors like parity, use of oral contraceptives, and a history of tubal ligation or hysterectomy have not been shown as significant factors for better survival of ovarian cancer. Breastfeeding is an exception and has been shown to be a significant protective factor.22 Number of lifetime ovulations and age at menarche, factors also associated with the risk of developing endometriosis, have been shown to be associated with ovarian cancer survival, i.e., the more lifetime ovulations, the poorer the prognosis.23
Epidemiological studies have shown women to have an advantage in survival in malignant melanoma over men. A relationship between endogenous or exogenous hormones and risk for malignant melanoma has not been clearly shown.25 Our study indicated (showed) a worse prognosis in malignant melanoma if the women had endometriosis, especially during the calendar period 1990–1999, compared to women without endometriosis. This raises the question about the impact of hormonal treatment for endometriosis. The clinical treatment preferences for endometriosis have changed over the decades. During the 1980's and in the beginning of the 1990's, danocrine, a testosterone derivative (17α-ethinyl testosterone), was the “gold standard” for endometriosis treatment. Estrogens inhibit invasion of malignant melanoma but, interestingly, dehydroepiandrosterone (DHEA) seems to enhance invasion.26 After the early 1990s, danocrine was not in use anymore and our study shows a lower HR for women with endometriosis and malignant melanoma during the last time period, 2000–2005. However, whether or not danocrine enhance invasion of malignant melanoma and thereby influences the prognosis of survival is not known. Our findings are new and have to be confirmed in additional studies.
Estrogens have been linked to several types of malignancies for example ovarian cancer, breast cancer and malignant melanoma.27 The effects of estrogens are mediated by two types of estrogen receptors, ERα and ERβ. Studies on ovarian endometriosis have shown an increased level of ERβ and a decreased level of ERα in endometriotic tissue as compared to normal uterine endometrium.28, 29 Studies have shown decreased expression of ERβ in tumor versus normal tissue in for instance breast cancer, ovarian cancer and malignant melanoma indicating that ERβ has a protective effect against a hyper proliferation induced by ERα.27, 30–38 These studies indicate that loss of ERβ is a marker for more invasive tumor growth and poorer prognosis. Whether endometriotic tissue in ovaries that develop cancer has reduced expression of ERβ is not known. We have previously shown an increased risk of breast cancer in women who got their endometriosis diagnosis after menopause, indicating the complex involvement of reproductive hormones in cancer development. However it could also be an indirect marker of more ERβ positive disease in endometriotic women and therefore result in a better prognosis.3
In the search for confounders there are of course other factors of potential interest like smoking habits, body mass index (BMI), physical activity, hormonal endometriosis treatment and cancer treatment. Data has indicated that for instance higher BMI, smoking and physical activity lowers the risk of endometriosis.39 High BMI has been shown to increase cancer mortality, both in breast and ovarian cancer.40 Another study showed both high BMI and smoking to be a factor for poorer prognosis in breast cancer, but the results regarding physical activity were inconclusive.41 However, this type of information, as well as information on hormonal endometriosis treatment, is not available in the registers used in our study and were therefore impossible to control for.
The registers also do not provide any information about cancer treatment given that might impact cancer survival. However, there are no indications that an endometriosis history would have any impact on the treatment alternative offered to cancer patients. There has been suggested that endometriosis associated ovarian cancer (EAOC) should be treated differently than other epithelial ovarian cancers because of its specific characteristics but to our knowledge no such results from treatment trials have so far been published.15, 42, 43 Further studies are needed to explain why women with endometriosis may have a better prognosis after a diagnosis of a malignancy.
This study shows that endometriosis is both a positive prognostic factor for some malignancies and a negative for others. The improved survival after a diagnosis of a malignancy in women with previously diagnosed endometriosis remained after adjustment for well-known confounders. The positive effect of endometriosis on survival was larger in the postmenopausal age group. This might reflect a common genetic predisposition in women with endometriosis allowing the endometriosis disease to establish but at the same time being an advantage in cancer survival. How this knowledge should be used in the everyday clinical work must be further evaluated.
- 17Prognostic implication of endometriosis in clear cell carcinoma of the ovary. 39th Annual Meeting of the Society-of-Gynecologic-Oncologists 2008: 336–44., , , , , .
- 19Statistics from Centre of Epidemiology, Cancer statistics. Swedish National Board of Health and Welfare, 2005.
- 43Endometriosis-associated ovarian cancer: a clinicopathologic review. J Obstet Gynaecol Can 2004; 26: 709–15., , .