Oral contraceptive use, reproductive history, and risk of epithelial ovarian cancer in women with and without endometriosis

American Journal of Obstetrics and Gynecology (2004) 191, 733e40

www.elsevier.com/locate/ajog

GENERAL OBSTETRICS AND GYNECOLOGY: GYNECOLOGY

Oral contraceptive use, reproductive history, and risk of epithelial ovarian cancer in women with and without endometriosis Francesmary Modugno, PhD, MPH,a,* Roberta B. Ness, MD, MPH,a Glenn O. Allen, MPH,a Joellen M. Schildkraut, PhD,b Faith G. Davis, PhD,c Marc T. Goodman, PhDd Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, and University of Pittsburgh Cancer Institute, Pittsburgh, Pa,a Cancer Prevention, Detection and Control Research Program, Department of Community and Family Medicine, Duke University Medical Center, Durham, NC,b Division of Epidemiology and Biostatistics, School of Public Health, University of Illinois at Chicago, Chicago, Ill,c and Cancer Research Center, University of Hawaii, Honolulu, Hi d Received for publication November 24, 2003; revised February 9, 2004; accepted March 11, 2004

–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––– KEY WORDS Ovarian cancer Endometriosis Estrogen Oral contraceptive

Objective: Women with endometriosis may be at an increased risk of ovarian cancer. It is not known whether reproductive factors that reduce the risk of ovarian cancer in general also reduce risk in women with endometriosis. We investigated whether the odds ratios for ovarian cancer that were associated with oral contraceptive use, childbearing, hysterectomy, and tubal ligation differ among women with and without endometriosis. Study design: We pooled information on the self-reported history of endometriosis from 4 population-based case-controlled studies of incident epithelial ovarian cancer, comprising 2098 cases and 2953 control subjects. We obtained data on oral contraceptive use, childbearing, breastfeeding, gynecologic surgical procedures, and other reproductive factors on each woman. Multivariable unconditional logistic regression was used to calculate odds ratios and 95% CI for ovarian cancer among women with endometriosis compared with women without endometriosis. Similar methods were used to assess the frequencies of risk factors among women with and without endometriosis. Adjustments were made for age, parity, oral contraceptive use, tubal ligation, family history of ovarian cancer, and study site. Results: Women with endometriosis were at an increased risk of ovarian cancer (odds ratio, 1.32; 95% CI, 1.06-1.65). Using oral contraceptives, bearing children, and having a tubal ligation or hysterectomy were associated with a similar reduction in the odds ratios for ovarian cancer among women with and without endometriosis. In particular, the use of oral contraceptives

Supported in part by the National Cancer Institute grants K07-CA80668 (F.M.), R01CA61095 (R.B.N), R01-CA-58598 (M.T.G), N01-CN67001 (M.T.G), R01-CA61093 (F.G.D.) and RO1 CA76016 (J.M.S.). * Reprint requests: Francesmary Modugno, PhD, MPH, Department of Epidemiology, Graduate School of Public Health, 516A Parran Hall, 130 DeSoto St, Pittsburgh, PA 15261. E-mail: [email protected]

0002-9378/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ajog.2004.03.035

734

Modugno et al forO10 years was associated with a substantial reduction in risk among women with endometriosis (odds ratio, 0.21; 95% CI, 0.08-0.58). Conclusion: Women with endometriosis are at an increased risk of epithelial ovarian cancer. Long-term oral contraceptive use may provide substantial protection against the disease in this high-risk population. Ó 2004 Elsevier Inc. All rights reserved.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––– More than 5.5 million women in the United States and Canada are affected by endometriosis, a condition that is characterized by the presence of endometrial tissue implants that are outside the uterine lining.1 The disease has a complex pathologic mechanism with different forms that include peritoneal and rectovaginal septum and ovarian endometriosis.2 Emerging data suggest that women with endometriosis are at an increased risk for having an array of other diseases, including fibromyalgia, rheumatoid arthritis, and ovarian cancer.3,4 Brinton et al3 observed a 90% excess risk of ovarian cancer among a cohort of 20,686 Swedish women who were hospitalized for endometriosis between 1969 and 1983. The risk was elevated particularly among women with a long-standing history of ovarian endometriosis (standardized incidence ratio, 4.2; 95% CI, 2.0-7.7). Ovarian cancer is the most common cause of death from a gynecologic malignancy in the United States.5 In 2003,O25,000 American women will be diagnosed with ovarian cancer;5 more than one half of these women will die within 5 years.5 Survival is better with localized, early-stage disease, but most patients have metastatic disease.6 To date, no effective early detection techniques have been identified, and primary prevention represents an important opportunity for reducing ovarian cancer morbidity and mortality rates. Using oral contraceptives (OCs), bearing children, and having a tubal ligation or hysterectomy have been shown to reduce ovarian cancer risk consistently among women in general.7-9 Whether these factors reduce the risk of ovarian cancer in women with endometriosis is unknown. In this study, we aimed to evaluate the benefit that is associated with OC use, childbearing, tubal ligation and

Table I

having a hysterectomy among women at risk for ovarian cancer because of a history of endometriosis.

Methods Subjects We pooled data on the history of endometriosis from 4 population-based, ovarian cancer case-controlled studies that recruited women from 4 regions of the United States from 1993 through 2001. Cases for the analyses represent women with ovarian cancer who participated in 1 of these 4 studies. Control subjects in our analyses are women without ovarian cancer who served as control subjects in the original studies. Specific descriptions of each study’s methods are provided in the original publications and are summarized in Table I. Briefly, Ness et al10 recruited women who were 20 to 69 years old from 39 hospitals in the Delaware Valley and ascertained control subjects who were frequency-matched to cases by age and area of residence by random-digit dialing or Health Care Financing Administration (HCFA) files. A total of 767 cases (88% of eligible subjects) and 1367 control subjects (72% of eligible subjects) were enrolled. Goodman et al11 recruited women who were 18 to 87 years old from any of the major hospital centers on the island of Oahu, Hawaii, and Los Angeles. Control subjects who were frequency-matched to cases by age, race, and location were identified from within statewide population registries or HCFA files. A total of 606 cases (65% of eligible subjects) and 613 control subjects (72% of eligible subjects) were enrolled. A third (and ongoing) study12 recruited women who were 20 to 74

Study sites and their characteristics that were included in the pooled analysis of 4 case-control studies, 1993-2001

Study Goodman et al

11

Kupelian et al13 Moorman et al12 Ness et al10

Location

Period of ascertainment Case selection

Control Selection

Hawaii and Los Angeles County, Calif Cook County, Ill North Carolina Delaware Valley, Pa

1993-1999

Population register HCFA files Random digit, HCFA for R64 y Random digit, HCFA Random digit dialing, HCFA files

* Percent recruited in Hawaii. y Percent recruited in Los Angeles, CA.

1994-1998 1999-2001 1993-1998

Surveillance, Epidemiology, and End Results registries Hospitals NC Central Cancer Registry Hospitals

735

Modugno et al years old with the use of a rapid case ascertainment system in conjunction with the population-based North Carolina Central Cancer Registry to identify 374 newly diagnosed cases of epithelial ovarian cancer within a 48-county region of North Carolina. Four hundred seventy-two control subjects were identified through random digit dialing and HCFA lists and were matched to cases by age and race. The response rates were 83% and 68% for cases and control subjects, respectively. Because this third study is ongoing, only data that were available at the time of the current analyses are included in this report. The fourth study13 recruited women who were 18 to 74 years old and who had received a diagnosis of ovarian cancer throughout all the hospitals within Cooks County, Illinois. Control subjects, frequency-matched to cases by age and race, were identified through random-digit dialing or HCFA lists. A total of 351 cases (75% of eligible subjects) and 501 control subjects (53% of eligible subjects) were enrolled. In all 4 studies, women who reported a history of bilateral oophorectomy were ineligible to serve as control subjects because they were not at risk for ovarian cancer. Each study obtained written informed consent from participants and was approved by the appropriate institutional review boards. This pooled analysis is exempt from institutional review board approval because all data were provided in an anonymous fashion and could not be linked to any individual. A total of 2098 cases and 2953 control subjects were included in the combined analysis.

Exposure information, history of endometriosis and data quality All 4 studies carried out standardized in-person interviews using modified versions of a single questionnaire. This questionnaire based recall for reproduction events on a life calendar, with important events during a woman’s life used to enhance her memory for date-related information. Women in each study were asked about a personal diagnosis of endometriosis by a health care provider and about infertility because of endometriosis. Each study also obtained information on risk factors

Cases N

and potentially important confounding variables that were related to ovarian cancer risk (such as age, parity, and oral contraceptive use, family history of ovarian cancer, history of tubal ligation, history of hysterectomy, race, and level of education). All data were checked for internal consistency, and corrections or clarifications were requested from the original investigators when necessary. To test the appropriateness of pooling the data estimates from 4 studies, we calculated a chi-squared statistic for heterogeneity.14 When significant heterogeneity among the studies was found, we provide the heterogeneity statistic and the range of odds ratios (ORs).

Statistical analyses Pooled ORs, with corresponding 95% CIs, were calculated as the primary measure of effect size. Because the 4 studies used frequency, rather than individual matching on the basis of broad criteria (such as age within 5-10 year intervals), we used unconditional logistic regression models to adjust for any additional effects of potential confounding variables. Included in the models were study site; age, parity, and ovarian cancer duration as continuous variables; and hysterectomy, tubal ligation, and family history of ovarian cancer as dichotomous (yes/no) variables. These were the only variables that were found to be relevant confounders in a previous analysis that examined endometriosis as 1 of many factors that related infertility to ovarian cancer.15 Although that study did examine endometriosis and other causes of infertility as risk factors for ovarian cancer, the current analysis focuses exclusively on endometriosis and factors that may modify the risk of ovarian cancer that is associated with endometriosis. Because models with and without the hysterectomy variable were almost identical, we report the results and exclude that variable from the models. Tests for trend (probability value) were performed by the coding of each factor (ie, parity, OC duration) as a grouped linear variable. A probability value for the interaction between a history of endometriosis and parity, OC duration, hysterectomy, or tubal ligation on the

Control Subjects Recruited (%) y

606

66* 65

351 374 767

75 83 88

N

Recruited (%) 613

501 472 1,367

Age range (y)

Matching parameters

75* 72

18-87

Frequency age, race, and location

53 68 72

18-74 20-74 20-69

Frequency age, race Frequency age, race Frequency age, residence

y

736

Modugno et al

Table II Demographic description of ovarian cancer cases and control subjects from population-based case control studies in the United States, 1993-2001 Demographic Age (y) !30 30-39 40-49 50-59 60-69 R70 No. of livebirthsz None 1-2 R3 Race White Black Hispanic Asian Other Education ! High school High school OHigh school Family history of ovarian cancer No Yes Tubal ligationx No Yes HysterectomyL No Yes Diagnosed with endometriosis{ No Yes Oral contraceptive use duration# Never !10 Y R10 Y

Control subjects (n)*

Cases (n)y

135 357 828 753 615 265

89 (4.2%) 211 (10.1%) 535 (25.5%) 580 (27.6%) 499 (23.8%) 184 (8.8%)

(4.6%) (12.1%) (28.0%) (25.5%) (20.8%) (9.0%)

425 (14.4%) 1358 (46.0%) 1169 (39.6%)

597 (28.5%) 878 (41.8%) 623 (29.7%)

2158 (73.1%) 341 (11.5%) 47 (1.6%) 225 (7.6%) 182 (6.2%)

1543 (73.5%) 180 (8.6%) 44 (2.1%) 178 (8.5%) 153 (7.3%)

330 (11.2%) 850 (28.8%) 1773 (60.0%)

252 (12.0%) 641 (30.6%) 1205 (57.4%)

2907 (98.4%) 46 (1.6%)

2025 (96.5%) 73 (3.5%)

2119 (72.0%) 826 (28.0%)

1756 (83.7%) 341 (16.3%)

2509 (85.7%) 418 (14.3%)

1786 (85.5%) 302 (14.5%)

2759 (93.7%) 184 (6.3%)

1912 (91.5%) 177 (8.5%)

1091 (37.1%) 1569 (53.3%) 282 (9.6%)

1007 (48.1%) 961 (45.9%) 125 (6.0%)

* N = 2953. y N = 2098. z One subject missing. x Nine subjects missing. L Thirty six subjects missing. { Nineteen subjects missing. # Sixteen subjects missing.

ORs for ovarian cancer was calculated as the significance of the interaction of the dichotomous endometriosis variable and the dichotomous hysterectomy or tubal ligation variables or the trichotomous parity or OC duration variables.

To further understand the differences between women with and without endometriosis, we compared nulliparity, OC use, reported history of hysterectomy, and tubal ligation between these 2 groups, which were stratified by ovarian cancer case-control status. Differences in these data between case and control subjects were assessed for significance using either a Student t test or a c2 test. All analyses were performed with the SPSS statistical software package (version 11.0; SPSS Inc, Chicago, Ill). Analyses were 2-sided and considered significant at a probability of !.05.

Results Of the 2098 cases and 2953 control subjects, 177 cases (8.5%) and 184 control subjects (6.3%) reported a history of endometriosis. After adjustments were made for study site, OC duration, parity, age, tubal ligation, and family history of ovarian cancer, we found that women with endometriosis were more likely to have ovarian cancer than women without a history of the disease (adjusted OR, 1.32; 95% CI, 1.06-1.65). The association was greater for nulliparous women (adjusted OR, 1.75; 95% CI, 1.13-2.70). As expected, cases were less likely to have used oral contraceptives, to have borne children, and to have had a tubal ligation (Table II). Similar associations of ovarian cancer risk with the duration of OC use, parity, hysterectomy, and tubal ligation were found among women with and without endometriosis (Table III). In particular, oral contraceptive pill use decreased the ORs for ovarian cancer independent of a history of endometriosis, and the decrease in risk increased with the duration of OC use. Even though estimates of the association of ever OC use with ovarian cancer were heterogeneous among studies (P = .033), with ORs that range from 0.57 (95% CI, 0.47-0.68) to 0.84 (95% CI, 0.64-1.10), all studies produced ORs !1.0. Compared with never users of OCs, the adjusted ORs for !10 years andO10 years of use were 0.58 and 0.21, respectively (trend P = .003) among women with endometriosis. Among women without endometriosis, the adjusted ORs were 0.70 and 0.47, respectively (trend P ! .001). Not surprisingly, because OCs are often prescribed to treat endometriosis, women with the condition were more likely to have used OCs (79.2% vs 61.7% for healthy control subjects [P ! .001]; 65.9% vs 50.5% for ovarian cancer cases [P ! .001]). However, there was no difference in the percentage of women who reported long-term use between women with and without endometriosis. In addition, among OC users, women with endometriosis tended to report a shorter duration of use, although the differences were not significant (46.9 vs 50.4 months for ovarian cancer cases [P = .52]; 54.1 vs 59.2 months for healthy control subjects [P = .32]).

737

Modugno et al

Table III Multivariable-adjusted ORs for epithelial ovarian cancer by reproductive and gynecologic factors-stratified by history of endometriosis Variable Births (M) 0 1-2

All women

Endometriosis

No endometriosis

Control Cases OR subjects (%) (%) (95% CI)

Control Cases OR subjects (%) (%) (95% CI)

Control Cases OR subjects (%) (%) (95% CI)

14.3 46.1

28.6 41.8

39.6

29.7

P interaction# .37 OC duration Never 37 !10 Y 53.4

48.2 45.9

R3

R10 Y

9.6

6

P interaction# .51 Tubal ligation No 71.9 Yes 28.1

83.8 16.2

P interaction# .92 Hysterectomy No 85.8 Yes 14.2

85.6 14.4

18.6 0.46* 55.7 (0.39-0.54)y 0.36* 25.7 (0.30-0.43)y trend P ! .001

42.6 40.3

20.8 0.69x 68.9 (0.60-0.79)y 0.45x 10.4 (0.35-0.57)y trend P ! .001

34.1 61.4

0.63** (0.54-0.73)y

0.99zz (0.83-1.18)

17

14 0.31z 45.5 (0.18-0.54)y 0.22z 40.5 (0.11-0.45)y trend P ! .001

27.3 41.9

38.1 52.4

49.4 44.5

9.5

6.1

72.2 27.8

83.8 16.2

86.5 13.5

86.1 13.9

0.58L (0.33-1.03) 4.5 0.21L (0.08-0.58){ trend P = .003

67.8 32.2

83 17

76.5 23.5

80.7 19.3

0.7yy (0.41-1.25)

0.69xx (0.38-1.24)

30.8

0.48z (0.41-0.57)y 0.38z (0.31-0.45)y trend P ! .001

0.70L (0.60-0.80)y 0.47L (0.37-0.61)y trend P ! .001

0.63yy (0.54-0.73)y

1.02xx (0.86-1.23)

P interaction# .24 All models adjusted for study site, age (continuous), family history of ovarian cancer (dichotomous). see footnotes for additional information. * OC duration (continuous); endometriosis and tubal ligation (dichotomous). y P ! .001. z OC duration (continuous); and tubal ligation (dichotomous). x Parity (continuous); endometriosis and tubal ligation (dichotomous). L Parity (continuous) and tubal ligation (dichotomous). { P ! .002. # Interaction between risk factor and reported history of endometriosis. ** OC duration and parity (continuous); endometriosis (dichotomous). yy OC duration and parity (continuous). zz OC duration and parity (continuous); endometriosis, tubal ligation (dichotomous). xx OC duration and parity (continuous); tubal ligation (dichotomous).

Bearing children was also associated with a decreased risk of ovarian cancer independent of a history of endometriosis, and the decrease in risk increased with the number of children. Compared with the ORs for nulliparous women, the adjusted ORs for 1 to 2 and 3C children were 0.31 and 0.22, respectively (trend P ! .001); for women with endometriosis, the adjusted ORs were 0.48 and 0.38, respectively (trend P ! .001) for women without the disease. Again, although estimates of the evereparous ovarian cancer association were heterogeneous among studies (P = .009), with ORs that ranged from 0.33 (95% CI, 0.27-0.42) to 0.63 (95% CI, 0.430.90), all ORs were !1.0. As expected, women with endometriosis were less likely to be parous (57.4% vs

72.7% for ovarian cancer cases [P ! .001]; 81.4% vs 86.0% for healthy control subjects [P ! .001]). Having a tubal ligation was associated with a decrease in the risk of ovarian cancer among women without endometriosis (OR, 0.63; 95% CI, 0.54-0.73). The OR that was associated with tubal ligation was also reduced among women with endometriosis, although not significantly (OR, 0.71; 95% CI, 0.41-1.25). Whether they subsequently had ovarian cancer, women with endometriosis were no more likely to have had a tubal ligation than women without endometriosis (rate of tubal ligation, 17.0% vs 16.2% among ovarian cancer cases [P = .77]; 32.2% vs 27.8% among healthy control subjects [P = .21]).

738 As expected, women with endometriosis were more likely to report having had a hysterectomy independent of case/control status. In particular, 23.5% of control subjects with endometriosis reported a hysterectomy, compared with only 13.5% of control women without endometriosis (P ! .001). Similarly, among cases with endometriosis, 19.3% of the women reported a hysterectomy; 13.9% of the women without endometriosis reported the procedure (P = .06). Although hysterectomy was generally not related to the risk of ovarian cancer, estimates of this association were heterogeneous among studies (p = .03, range of ORs from 0.76 [95% CI, 0.581.01] to 1.36 [95% CI, 0.98-1.87]). Moreover, having a hysterectomy was associated with a nonsignificant decrease in risk for women with endometriosis (OR, 0.69). Although no such decrease in risk was observed in women without a history of endometriosis (OR, 1.02), the interaction was not statistically significant (p for interaction = .24).

Comment In this pooled analysis, the occurrence of ovarian cancer was a significant 30% higher among women who reported a history of endometriosis compared with those women who reported no such history. This association was greater in nulliparous women, which confirms that the association is likely due to the endometriosis and not the infertility or decreased parity caused by the endometriosis.16 Using oral contraceptives, bearing children, and having a tubal ligation reduced the risk of ovarian cancer, irrespective of a reported history of endometriosis. Moreover, longer term OC use and higher parity provided increased protection against the disease in women with and without endometriosis. Our findings have important clinical applications. In particular, the high case-fatality rate that is associated with ovarian cancer makes risk reduction critical in all women, but especially in women at an identifiably increased risk, such as women with a history of endometriosis. To date, only OCs have emerged as chemopreventive agents against ovarian cancer. OCs are prescribed commonly for women with endometriosis. Our data suggest that this clinical practice may have an added benefit: protection against ovarian cancer. Notably, in our data set, although women with endometriosis were more likely to use OCs in general, they were no more likely to be long-term users of OCs than women without endometriosis, which suggests that women with endometriosis are treated initially with OCs, but then move on to other medical treatments for the condition, potentially reducing the protection against ovarian cancer afforded by OC use. Moreover, it is possible that other treatments for endometriosis may increase ovarian cancer risk. In particular, we recently reported that the

Modugno et al use of danazol for the treatment of endometriosis significantly increased a woman’s risk of ovarian cancer,17 although treatment with leuprolide (Lupron) did not. Hence, clinicians must consider carefully the shortand long-term tradeoffs of OC use when considering alternate medical treatments for endometriosis. Our findings have a biologic basis. Both endometriosis18 and ovarian cancer19 are associated with unopposed estrogens. Possibly, an estrogen-rich, progesterone-poor hormonal environment may encourage the growth of endometriosis and promote its malignant transformation to ovarian cancer.20 OCs and childbearing, which have been shown to reduce the risk of both diseases,7-9,21-23 both reduce unopposed estrogens.24,25 This, then, may be the mechanism by which OCs and parity reduce ovarian cancer risk among women with endometriosis. In addition, both ovarian cancer and endometriosis have other factors in common (such as immune imbalance, inflammation, and an association with retrograde menstruation).20 Parity and OC use may have beneficial effects on these factors as well, for example by reducing the number of menstrual cycles and the amount of menstrual flow. These factors may also explain, in part, how OCs and parity reduce ovarian cancer risk among women with endometriosis. Selection and information biases were minimized in this pooled analysis by the population-based study designs, large sample size, structured interviews, and detailed data collection on OC use, childbearing, and history of gynecologic conditions and surgeries. In particular, because neither subjects nor interviewers were aware of our interest in endometriosis as it relates to ovarian cancer and because questions on endometriosis were included along with other questions on gynecologic conditions and medical history, it is unlikely that recall and interviewer biases affected our observations. The 2 major limitations of this study are the small number of women who reported a history of endometriosis and the derivation of these data from self-reports. In addition, because a diagnosis of endometriosis can be established reliably only through surgery, it is possible that some women who reported a history of the disease may not have had the condition (because we relied on a self-reported medical diagnosis, not a surgically proven one; we did not collect data on whether the diagnosis was surgically confirmed). Moreover, because the symptoms of endometriois (such as pelvic pain) are vague and may not correlate necessarily with disease severity, it is possible that some women with the disease were never diagnosed and thus were included in our analyses as never having endometriosis. It is also possible that women who were diagnosed with the disease did not report it because they had forgotten. Endometriosis is diagnosed commonly in women during their childbearing years (20 and 30 years old). In contrast, ovarian cancer is a disease more common in postmenopausal

Modugno et al women. Hence, the time lag between endometriosis diagnosis and ovarian cancer diagnosis, some 25 to 30 years, might have caused women with a medical diagnosis of endometriosis to forget. However, because none of these sources of misclassifying subjects would be expected to be different between cases and control subjects, they would bias our results towards the null. Hence, it is likely that the data that are reported here underestimate the true association between endometriosis and ovarian cancer and the protection afforded by OC use, childbearing, and tubal ligation. In addition, because of the small numbers of women with endometriosis, we were unable to analyze risk factors according to histologic subtypes. Endometriosis has been associated with endometrioid and clear-cell epithelial ovarian cancers.26-29 It is possible that different subtypes of ovarian cancers have different causes.30-33 However, these differences appear to be limited to mucinous tumors; clear-cell and endometrioid tumors have a risk factor profile similar to that of serous tumors, which are most epithelial tumors.32,34 In particular, age at onset of ovarian cancer appears similar across the nonmucinous histologic conditions. Our observations confirm previous reports that endometriosis increases the risk of ovarian cancer. We further found that OCs, childbearing, and possibly tubal ligation provide these high-risk women with protection against ovarian cancer. Although our findings must be confirmed, they suggest that, when women with endometriosis are being treated, the use of OCs, especially long-term use, should be encouraged.

Acknowledgments We thank Dr Anna Wu for data collection in Los Angeles, Randi Koenig for her help with the manuscript, and Dr Jeffrey L. Eppinger for his help with this work.

References 1. Ballweg ML. The endometriosis sourcebook. Milwaukee (WI): The Endometriosis Association; 1999. 2. Vignali M, Infantino M, Matrone R, Chiodo I, Somigliana E, Busacca M, et al. Endometriosis: novel etiopathogenetic concepts and clinical perspectives. Fertil Steril 2002;78:665-78. 3. Brinton LA, Gridley G, Persson I, Baron J, Bergqvist A. Cancer risk after a hospital discharge diagnosis of endometriosis. Am J Obstet Gynecol 1997;176:572-9. 4. Sinaii N, Cleary SD, Ballweg ML, Nieman LK, Stratton P. High rates of autoimmune and endocrine disorders, fibromyalgia, chronic fatigue syndrome and atopic diseases among women with endometriosis: a survey analysis. Hum Reprod 2002;17: 2715-24. 5. American Cancer Society. American Cancer Society facts and figures 2003. Washington (DC): American Cancer Society; 2003. p. 1-52.

739 6. Heintz AP, Odicino F, Maisonneuve P, Beller U, Benedet JL, Creasman WT, et al. Carcinoma of the ovary. J Epidemiol Biostat 2001;6:107-38. 7. Whittemore AS, Harris R, Itnyre J. Characteristics relating to ovarian cancer risk: collaborative analysis of 12 US case-control studies: II, invasive epithelial ovarian cancers in white women: Collaborative Ovarian Cancer Group. Am J Epidemiol 1992;136: 1184-203. 8. Rosenblatt KA, Thomas DB. Reduced risk of ovarian cancer in women with a tubal ligation or hysterectomy: the World Health Organization Collaborative Study of Neoplasia and Steroid Contraceptives. Cancer Epidemiol Biomarkers Prev 1996;5:933-5. 9. Gwinn ML, Lee NC, Rhodes PH, Layde PM, Rubin GL. Pregnancy, breast feeding, and oral contraceptives and the risk of epithelial ovarian cancer. J Clin Epidemiol 1990;43:559-68. 10. Ness RB, Grisso J, Klapper J, Schlesselman JJ, Silverzweig S, Vergona R, et al and the SHARE Study Group. Risk of ovarian cancer in relation to estrogen dose and use characteristics of oral contraceptives. Am J Epidemiol 2000;152:233-41. 11. Goodman MT, Wu AH, Tung KH, McDuffie K, Cramer DW, Wilkens LR, et al. Association of galactose-1-phosphate uridyltransferase activity and N314D genotype with the risk of ovarian cancer. Am J Epidemiol 2002;156:693-701. 12. Moorman PG, Schildkraut JM, Calingaert B, Halabi S, Vine MF, Berchuck A. Ovulation and ovarian cancer: a comparison of two methods for calculating lifetime ovulatory cycles (United States). Cancer Causes Control 2002;13:807-11. 13. Kupelian V, Davis F, Mallin K, Rosenblatt K. Tubal sterilization, hysterectomy and risk of ovarian cancer: a case-control study [Abstract 273]. Presented at SER Seattle, June 2000. Am J Epidemiol 2000;151:S69. 14. Breslow NE, Day NE. Statistical methods in cancer research. 32 ed. Lyon, France: IARC Scientific Publications; 1980. 15. Ness RB, Cramer DW, Goodman MT, Kjaer SK, Mallin K, Mosgaard BJ, et al. Infertility, fertility drugs, and ovarian cancer: a pooled analysis of case-control studies. Am J Epidemiol 2002; 155:217-24. 16. Elsheikh A, Milingos S, Loutradis D, Kallipolitis G, Michalas S. Endometriosis and reproductive disorders. Ann NY Acad Sci 2003;997:247-54. 17. Cottreau CM, Ness RB, Modugno F, Allen GO, Goodman MT. Endometriosis and its treatment with danazol or Lupron in relation to ovarian cancer. Clin Cancer Res 2003;9:5142-4. 18. Cramer DW, Missmer SA. The epidemiology of endometriosis. Ann NY Acad Sci 2002;955:11-22. 19. Risch HA. Hormonal etiology of epithelial ovarian cancer, with hypothesis concerning the role of androgens and progesterone. J Natl Cancer Inst 1998;90:1774-86. 20. Ness RB. Endometriosis and ovarian cancer: thoughts on shared pathophysiology. Am J Obstet Gynecol 2003;189:280-94. 21. Vessey MP, Villard-Mackintosh L, Painter R. Epidemiology of endometriosis in women attending family planning clinics. BMJ 1993;306:182-4. 22. Signorello LB, Harlow BL, Cramer DW, Spiegelman D, Hill JA. Epidemiologic determinants of endometriosis: a hospital-based case-control study. Ann Epidemiol 1997;7:267-74. 23. Parazzini F, Ferraroni M, Fedele L. Pelvic endometriosis: reproductive and menstrual risk factors at different stages in Lombardy, northern Italy. J Epidemiol 1995;49:61-4. 24. King RJ. Biology of female sex hormone action in relation to contraceptive agents and neoplasia. Contraception 1991;43:527-42. 25. Yen SS. Endocrinology of pregnancy. In: Creasy RK, Resnik R, editors. Maternal-fetal medicine: principles and practice. 3rd ed. Philadelphia: Saunders; 1994. p. 382-412. 26. de la Cuesta RS, Eichhorn JH, Rice LW, Fuller AF, Nikrui N, Goff BA. Histologic transformation of benign endometriosis to early epithelial ovarian cancer. Gynecol Oncol 1996;60:238-44.

740 27. Fukunaga M, Nomura K, Ishikawa E, Ushigome S. Ovarian atypical endometriosis: its close association with malignant epithelial tumours. Histopathology 1997;30:249-55. 28. McMeekin DS, Burger RA, Manetta A, DiSaia P, Berman ML. Endometrioid adenocarcinoma of the ovary and its relationship to endometriosis. Gynecol Oncol 1998;59:81-6. 29. Vercellini P, Parazzini F, Bolis G, Carinelli S, Dindelli M, Vendola N, et al. Endometriosis and ovarian cancer. Am J Obstet Gynecol 1993;169:181-2. 30. Modugno F, Ness RB, Cottreau CM. Cigarette smoking and the risk of mucinous and nonmucinous epithelial ovarian cancer. Epidemiology 2002;13:467-71.

Modugno et al 31. Modugno F, Ness RB, Allen GO. Alcohol consumption and the risk of mucinous and nonmucinous epithelial ovarian cancer. Obstet Gynecol 2003;102:1336-43. 32. Risch HA, Marrett LD, Jain M, Howe GR. Differences in risk factors for epithelial ovarian cancer by histologic type: results of a case-control study. Am J Epidemiol 1996;144:363-72. 33. Purdie DM, Webb PM, Siskind V, Bain CJ, Green AC. The different etiologies of mucinous and nonmucinous epithelial ovarian cancers. Gynecol Oncol 2003;88(suppl):S145-8. 34. Modugno F, Ness RB, Wheeler JE. Reproductive risk factors for epithelial ovarian cancer according to histologic type and invasiveness. Ann Epidemiol 2001;11:568-74.