Postmenopausal cancer risk after self-reported endometriosis diagnosis in the Iowa Women's Health Study
Endometriosis, the presence of endometrial tissue outside the uterus, has an estimated prevalence between 4% and 10%. A recent study reported that women with a hospital discharge diagnosis of endometriosis were at increased risk of cancer at any site, breast cancer, ovarian cancer, and hematopoietic malignancies, especially non-Hodgkin lymphoma (NHL).
The authors examined whether self-reported diagnosis of endometriosis was associated with increased risk of various cancers in the Iowa Women's Health Study. Incident cancer cases were identified from 1986 through 1998. Cox proportional hazards regression models were used to calculate relative risks and 95% confidence intervals for the particular cancers of interest.
Of 37,434 participants in this analysis, 3.8% reported a history of endometriosis at baseline in 1986. After 13 years of follow-up, 1795 breast, 188 ovarian, and 243 NHL cases were detected in the cohort. Endometriosis was not associated with risk of all cancers combined (age-adjusted relative risk [RR], 0.9; 95% confidence interval [CI], 0.7–1.2), breast carcinoma (RR, 1.0; 95% CI, 0.8–1.3), or ovarian carcinoma (RR, 0.8; 95% CI, 0.2–2.4). However, endometriosis was significantly associated with risk of NHL (age-adjusted RR, 1.8; 95% CI, 1.0–3.0), especially diffuse NHL (RR, 3.2; 95% CI, 1.8–5.6). Multivariate adjustment had minimal effect on the point estimates of risk (NHL RR, 1.7; 95% CI, 0.97–2.7; diffuse NHL RR, 3.3; 95% CI, 1.9–5.9). Endometriosis was not associated with elevated risk of lung, urinary tract, endometrial, melanoma, or colorectal carcinomas (RRs, 1.2, 0.8, 1.2, 0.7, and 0.7, respectively).
These results corroborate a previously reported association between endometriosis and increased risk of NHL but not cancer at other sites. Cancer 2002;94:1612–18. © 2002 American Cancer Society.
Endometriosis, the presence of viable endometrial tissue outside the uterus, is one of the most common gynecologic disorders in women of child-bearing years.1, 2 Estimates of the prevalence of endometriosis have varied widely depending on the specific population studied, but most estimates in the general population range from 4% to 10%.3 Women with this condition may be asymptomatic or may experience a variety of symptoms including chronic and severe pelvic pain, heavy menstrual bleeding, cramping, or even infertility. Although its etiology is unknown, the process is thought to begin when endometrial tissue shed during menstruation is refluxed via the fallopian tubes into the peritoneal cavity where cellular growth is stimulated by both immunologic and peritoneal factors.2 Risk factors for endometriosis include increasing age, reproductive health factors related to increasing exposure to menstruation (shorter cycle length, longer flow, and reduced parity), low levels of exercise, and higher levels of peripheral body fat.3
An association between malignancy and endometriosis has been reported in the literature for many years. Primarily, this association has consisted of case reports of endometriosis tissue undergoing malignant transformation.4 This type of transformation is thought to be rare, occurring in only 0.7–1.0% of patients with endometriosis4 and typically occurs in endometrial tissue located on the ovarian surface.5–7 Recently, however, Brinton et al.8 reported that women who were hospitalized with endometriosis were at increased risk of subsequent cancer in nonendometriosis tissue. These authors utilized the Swedish Inpatient Register for the time period of 1969 to 1983 to find all women with a hospital diagnosis of endometriosis (n = 20,686) and linked them to the National Swedish Cancer Registry to identify women who developed cancer through 1989. In that study, women with endometriosis were at slightly, but statistically significant, increased risk of cancer overall (standardized incidence ratio [SIR], 1.2; 95% confidence interval [CI], 1.1–1.3), breast carcinoma (SIR, 1.3; 95% CI, 1.1–1.4), ovarian carcinoma (SIR, 1.9; 95% CI, 1.3–2.8), and non-Hodgkin lymphoma (NHL; SIR, 1.8; 95% CI, 1.2–2.6).8 The study was limited by lack of information on possible confounding factors that may have influenced the results.
In addition to the link between endometriosis and ovarian cancer noted previously, increased risk of breast cancer has been noted in at least two earlier reports. Moseson et al.9 reported a nonsignificant elevated odds ratio (OR, 1.8) between endometriosis and breast cancer in a case–control study of 354 cases and 747 controls. A similarly elevated risk (relative risk [RR], 1.7) was reported from a Swedish cohort study among 15,844 female surgical patients.9, 10 A more recent case–control study found no association between history of endometriosis and breast cancer.11 We know of no published reports that show an association between lymphoma and endometriosis, other than that of Brinton et al.8
The mechanism through which endometriosis may be associated with cancer is unclear. One possibility is that both share similar risk factors.8 For example, both breast cancer and endometriosis have been linked to nulliparity and late age at birth of a first child. Other possibilities include endometriosis treatment, which often involves hormonal therapies such as danazol or oral contraceptives that may increase cancer risk, or there may be long-term effects of the immune reaction to endometriotic tissue implantation.
Self-reported history of endometriosis diagnosis was recorded on the 1986 baseline questionnaire in the Iowa Women's Health Study (IWHS). This cohort of more than 40,000 postmenopausal women is linked annually to the Iowa Cancer Registry to determine if cohort members have developed a malignancy. Using this cohort, we attempted to replicate the results of Brinton et al.8 and evaluated the hypothesis that self-reported endometriosis diagnosis was positively associated with breast cancer, ovarian cancer, and NHL. As a check on the specificity of the association, other cancer endpoints also were evaluated.
The IWHS is a prospective cohort study designed to identify risk factors for cancer and other chronic diseases in postmenopausal women.12 In January 1986, a questionnaire was mailed to 99,826 women who had been selected at random from a list of women ages 55–69 years holding a valid Iowa driver's license in 1985. The 41,836 respondents (42.7% response rate) form the cohort under study. Responders were approximately 3 months younger and 0.38 kg/m2 lighter on average than survey nonresponders.13 Cancer rates among responders and nonresponders were quite similar for most cancer types, with the exception of lung carcinoma, other smoking-related cancers, and all-site cancer, which were somewhat higher in nonresponders.13 Self-reported items included reproductive factors, education, medication use, alcohol use, smoking habits, and family history of cancer. Endometriosis was ascertained by asking, “Have you ever been told by a doctor that you have endometriosis?” If women responded yes, they were asked to provide an age at first diagnosis. Physical activity was ascertained through three questions about participation in leisure exercise and, if any, the frequency of moderate-intensity and heavy-intensity activities. These were combined to create a three-level activity score (low, medium, and high). Body fat distribution was estimated as the ratio of circumference measurements of the waist and hips (WHR) using a paper tape measure that was mailed with the questionnaire. Measurements by this method have been shown to be both accurate and reliable.14 Body mass index was calculated as the weight in kilograms divided by the height in meters, squared. Follow-up questionnaires were mailed in 1987, 1989, 1992, and 1997 to update vital status and current address. Deaths were ascertained by annual linkage to the Iowa death certificate database, supplemented by linkage to the National Death Index.
Exclusions and Cancer Incidence
Women reporting at baseline previous cancers other than nonmelanoma skin carcinoma (n = 2293) or a menstrual period within the last year (n = 569) were excluded from all analyses. The total cohort at risk for incident cancer was 37,434 women. Further exclusions were applied to women who reported bilateral oophorectomy (n = 6592), hysterectomy (n = 12,603), total or partial mastectomy (n = 354), or previous chemotherapy (n = 70), when examining risk of ovarian cancer, endometrial cancer, breast cancer, or NHL, respectively. When examining the “any cancer” endpoint, analyses were run both among the cohort after baseline exclusions (excluding only women with previous cancer and premenopausal) and after the more restrictive exclusions were applied (baseline exclusions plus any women reporting bilateral oophorectomy, hysterectomy, mastectomy, or previous chemotherapy). Estimates were essentially identical; thus, only the results applying the less restrictive exclusions (baseline exclusions only) are presented.
Incident cancer cases occurring from 1986 through 1998 were identified through the State Health Registry of Iowa, part of the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program.15 A match was performed annually between the list of cohort members and the records of Iowans with incident cancer in the registry. The match involves multiple stages of computer matching using social security numbers; driver's license numbers; combinations of first, last, and maiden name; zip code; and birth date. Each possible match is confirmed manually. Data regarding the diagnoses were abstracted by registry personnel from medical records and pathology reports according to SEER protocol.16
For each woman, age time scales were constructed for 1) age at baseline and 2) age at cancer diagnosis, death, or relocation outside of Iowa. If none of these events applied, the woman was assumed to be cancer free and living in Iowa, and her age on December 31, 1998 was used for the time scale. Women continued to be followed up and contributed to the time scale if there was a diagnosis of a cancer other than the particular cancer endpoint under study.
Rate ratios (RR) and 95% CIs were calculated using Cox proportional hazards regression models. For all Cox models, survival was modeled as a function of age,17 because age is a better predictor of baseline cancer risk than is length of follow-up time in this study. All statistical tests were two sided, and all analyses were conducted using the SAS (SAS Institute, Inc., Cary, NC) and Splus (Mathsoft, Inc., Seattle, WA) software systems.
Of the 37,434 women in the IWHS cohort at risk, 1392 (3.8%) reported at baseline that they had ever been diagnosed with endometriosis. The median age at diagnosis was 38 years (interquartile range, 31–45).
Table 1 provides information on women with and without self-reported endometriosis diagnosis on the 1986 baseline questionnaire. Women with endometriosis were more likely to be younger, better educated, nulliparous, and to report use of hormone replacement therapy than women without endometriosis. They were also more likely to have a lower body mass index, a lower waist-to-hip ratio and have a history of blood transfusion, hysterectomy, and dilatation and curettage.
Table 1. Demographic, Lifestyle, and Anthropometric Factors in Women with and Without Endometriosis at Baseline in 1986 in the Iowa Women's Health Study
|Age at cohort entry (yrs)|
| < High school||3192||9||42||3|
| High school||18,228||52||595||43|
| > High school||13,307||38||753||54|
|Alcohol intake (g/day)|
| < 4 g/day||8149||23||374||27|
| 4+ g/day||6913||20||350||25|
|Pack-years of smoking|
| < 20||4682||14||232||17|
|Age at first live birth (yrs)|
| < 20||6865||20||233||17|
| > 29||2006||6||71||5|
| 5 or more||6805||20||113||8|
|Oral contraceptive use||6246||18||404||29|
|Hormone replacement therapy|
| Formerly using||3504||10||386||28|
| Currently using||9231||27||556||40|
|History of blood transfusion||8676||26||540||41|
|History of hysterectomy||10,646||30||1005||72|
|History of dilatation and curettage||15,312||45||864||63|
|Body mass index (kg/m2)|
| ≤ 23.45||8728||25||441||32|
| > 29.68||8690||25||260||19|
|Waist to hip ratio|
| ≤ 77.70||8649||25||441||32|
| > 89.24||8706||25||255||18|
Table 2 presents the number of cases of each specific cancer type among women with endometriosis, as well as the age-adjusted associations of endometriosis for the various cancer types. The average time interval between endometriosis diagnosis and cancer diagnosis was quite long, usually approximately 30 years. The cancer type with the shortest interval between endometriosis diagnosis and cancer diagnosis was ovarian cancer, with an average interval of 22.7 years. The other intervals ranged from 28.2 for endometrial cancer to 32.0 years for lung cancer. If the number of exposed cases permitted, other previously reported important covariates were added to regression models. Those cancer types with few exposed cases were only adjusted for age. Endometriosis was not associated with risk of all cancers combined, nor for any specific cancer type with the exception of NHL. Women who reported having received diagnosis of endometriosis were 1.8 times more likely to develop NHL than women without endometriosis (95% CI, 1.0–3.0). The association remained elevated after adjustment for transfusion history, marital status, and alcohol intake (RR, 1.7; 95% CI, 1.0–2.9), known risk factors for NHL in this cohort.
Table 2. Self-Reported Endometriosis Diagnosis and Risk of Various Cancer Endpoints in the Iowa Women's Health Study, 1986–1998
|Cancer at any site||176 (3.23)||279,255||0.87||0.65–1.16||0.90||0.77–1.05|
|Breast cancer||67 (3.7)||418,004||1.01||0.79–1.29||0.96||0.75–1.23|
|Colorectal cancer||23 (2.7)||421,915||0.73||0.48–1.10||—b|
|Endometrial cancer||7 (1.7)||279,255||1.20||0.57–2.53||—b|
|Lung cancer||25 (4.4)||421,915||1.23||0.83–1.83||1.11||0.74–1.66|
| NHL–diffuse type||13 (5.3)||420,382||3.16||1.78–5.61||3.28||1.85–5.85|
|Ovarian cancer||3 (1.6)||347,275||0.78||0.25–2.44||—b|
|Urinary tractc||7 (3.0)||421,915||0.81||0.38–1.73||—b|
The associations of endometriosis with NHL were examined by common subtypes according to the classification scheme of Groves et al.18 Of the 15 women with endometriosis who developed NHL, 13 had diffuse NHL, 1 had follicular, and 1 had small lymphocytic NHL. Women who reported having endometriosis were 3.2 times more likely to develop diffuse NHL than women without endometriosis (95% CI, 1.8–5.6). Relative risks were further elevated (RR, 3.3; 95% CI, 1.8–5.9) after adjustment for marital status and alcohol intake. Further adjustment factors (diabetes, transfusion history, estrogen use, farm residence) were examined and did not influence the risk estimate (data not shown). We also categorized NHL into lymph node versus extranodal sites. Eight women with endometriosis developed lymph node NHL, and seven women developed extranodal NHL. An association was observed for extranodal NHL (RR, 3.1) but not lymph node NHL (RR, 1.3).
Endometriosis is one of the most common gynecologic disorders among women of child-bearing age. It is the third leading cause of gynecologic hospitalizations and a common reason for hysterectomy.3 Thus, if endometriosis were to confer increased risk for various cancer types, the number of women who could be affected would be quite high. Our goal was to evaluate, within the IWHS, the associations between endometriosis and various cancer endpoints as reported by Brinton et al.8 We were unable to confirm associations between endometriosis and cancer at any site, breast cancer, or ovarian cancer. However, like the Brinton8 report, women in our cohort who reported a past diagnosis of endometriosis were at elevated risk of NHL. Risks were especially elevated for diffuse NHL and extranodal disease.
The Brinton report,8 included 28 cases of NHL from a population of 20,686 women; 15.7 cases were expected in the group, yielding a RR of 1.8 (95% CI, 1.2–2.6). The current analysis from the IWHS included 243 cases from 37,434 women. Unlike the current report, Brinton et al. were unable to examine NHL subtypes nor were they able to adjust for other known NHL risk factors.
The association between endometriosis and NHL found in this and the previous study8 may be linked to abnormal cellular and humoral immune function reported in endometriosis patients.2, 19 Patients with endometriosis have been reported to have increased numbers of peritoneal macrophages and lymphocytes19 and higher secretion of cytokines and growth factors.19 Although not consistently found,20 some investigators have reported elevated levels of interleukin-1 in patients with endometriosis.21 Other studies have suggested abnormalities in levels of interleukin-6, tumor necrosis factors, and vascular endothelial growth factor in patients with endometriosis.19
Humoral immunity abnormalities in endometriosis patients center primarily around increased autoimmunity. Gleicher et al.22 reported that between 40% and 60% of women with endometriosis had elevated autoantibody titers. Furthermore, some cases of endometriosis have been shown to be associated with a generalized polyclonal B-cell autoimmune activation.19, 23 Although the specific lymphoma cell type (i.e., B cell or T cell) of women who develop lymphoma in the IWHS was not collected, most NHL in the IWHS is expected to be B cell.24 Thus, there may be a link between B-cell activation in endometriosis and development of B-cell lymphomas. This autoimmune response may be a predisposing factor25 that increases the risk of developing a genetic change by increasing the relative proportion of lymphoid cells involved in the autoimmune response.
A second possible explanation of the endometriosis/NHL link may be an increased risk caused by medications prescribed to treat endometriosis. Danazol, a medication commonly used to treat endometriosis for more than 20 years, has been shown to have immunosuppressive effects, including reduction of in vitro lymphocyte proliferation and suppression of autoantibody production.26 Although it is not entirely clear how this or other medications used for endometriosis treatment might trigger lymphoma, other factors associated with immune suppression have been linked to risk of lymphoma (i.e., human immunodeficiency virus, antirejection transplant drugs, steroid use).27 It is intriguing that the types of NHL found in this analysis to be most strongly associated with endometriosis, diffuse and extranodal NHL, are types of NHL commonly associated with immunosuppression.
A third possible underlying cause of the link between endometriosis and NHL may be environmental toxins, such as 2,3,7,8-tetrachlorodibenzo-p-dioxin (TTD or dioxin). Rhesus monkeys chronically exposed to dioxin over a period of 4 years were much more likely than control animals to later develop endometriosis (P > 0.001).28 An increased risk of malignancy in humans, including NHL, also has been linked to dioxin. For example, 15 years after an accidental release of dioxin to the environment in Italy in 1976, the risk of NHL had increased almost 3-fold (RR, 2.8; 95% CI, 1.1–7.0).29
In this analysis, we found no association between endometriosis and breast cancer. This may indicate that the previous association reported by Brinton et al.8 was a random occurrence. Alternatively, it may reflect differences in the populations studied. The earlier report by Brinton et al.8 was among women with a median age of 39 years at entry into the study, with a mean age at cancer diagnosis of 52 years. Our population in the IWHS was strictly postmenopausal women older than age 55 years at study initiation, with a mean age of 69 at cancer diagnosis. Menopausal status is known to be important in endometriosis. Clinically, menopause has been observed to “cure” endometriosis.26 It may be that the latency period between development of endometriosis and breast cancer is short enough that any women with endometriosis who were going to develop breast cancer did so before age 55 and therefore would not have been included in our analyses.
In the analyses of endometriosis and ovarian cancer, we detected no elevated risk for ovarian cancer in women with endometriosis. However, the number of women who developed ovarian cancer in this cohort is quite limited. One hundred ninety-one cases of ovarian cancer have occurred in the IWHS since 1985, but only 3 of those had reported endometriosis at baseline. Therefore, we were not powered to detect a statistically significant association in this group.
A limitation to this study is the reliance on self-reports of endometriosis in this cohort. Endometriosis is notoriously underdiagnosed, usually requiring a surgical procedure such as a laparoscopy to make a definitive diagnosis.3 Many women with endometriosis have few symptoms and therefore never receive a clinical diagnosis or are unable or unwilling to seek medical care that would lead to a diagnosis of endometriosis.3 Unfortunately, confirmation of endometriosis diagnosis via medical records was infeasible in our study. However, both underdiagnosis and poor self-report would most likely attenuate our results. Although this attenuation might explain the lack of association between endometriosis and ovarian or breast carcinoma, it is unlikely to lead to the positive association noted for NHL in this report.
In conclusion, these data are consistent with a previously reported link between endometriosis and subsequent development of lymphoma, especially diffuse type lymphoma. There was no evidence of associations between endometriosis and cancer at any other site.