Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
American Journal of Epidemiology Copyright © 2001 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 153, No. 9 Printed in U.S.A. Risk Factors for Mucinous and Nonmucinous Ovarian Cancer Purdie et al. Reproduction-related Risk Factors for Mucinous and Nonmucinous Epithelial Ovarian Cancer David M. Purdie,1 Victor Siskind,1 Christopher J. Bain,2 Penelope M. Webb,1 and Adèle C. Green1 The proposition that mucinous ovarian cancer has an etiology distinct from that of other histologic types has been evaluated using data from a population-based case-control study of epithelial ovarian cancer conducted in 1990–1993 among Australian women aged 18–79 years. The protective effects of parity and oral contraceptive use were greater in nonmucinous than in mucinous ovarian tumors. However, these differences appeared to be driven largely by the effect of ovulatory life, which was positively associated with nonmucinous tumors only. An association with family history of breast and/or ovarian cancer also appeared to be restricted to nonmucinous cancers. These results lend support to the hypothesis that mucinous and nonmucinous ovarian tumors develop via different causal mechanisms. Am J Epidemiol 2001;153:860–4. contraceptives, oral; histology; ovarian neoplasms; parity In a recent paper, Risch et al. (1) proposed that epithelial ovarian cancer is not a homogeneous entity from an etiologic point of view. In particular, they presented evidence suggesting that parity, oral contraceptive use, and family history of ovarian cancer have a different relation to carcinomas of the mucinous subtype than they do to other histologic types. A number of other studies have presented similar findings (2–4). A more recent publication (5), however, did not support this proposition, although differences in other reproductive factors, such as histories of infertility or hysterectomy, were suggested. Moreover, our analysis of the association between oral contraceptives and ovarian cancer showed that the odds ratio of mucinous carcinoma for a year of oral contraceptive use, adjusted for the number of lifetime ovulatory cycles, was very similar to that of both serous and endometrioid types (6). To clarify whether mucinous tumors do constitute a distinct etiologic subgroup, we present here a comparison of mucinous versus other types of epithelial ovarian cancer combined with respect to their reproductive and gynecologic risk factor profiles. states. Women diagnosed in 1991 and 1992 were recruited in New South Wales and Victoria; in Queensland, where the cancer registry was an additional source, women diagnosed in these 2 years and also in 1993 and the latter part of 1990 were eligible for enrollment. A specialist gynecologic pathologist in each state conducted an independent histologic review of all specimens. A detailed description of the study has been published (6, 7). Briefly, after consent from the attending doctor, research assistants invited all eligible cases (those aged between 18 and 79 years and competent to complete a questionnaire) to participate either prior to discharge or while attending the clinic for follow-up. Controls were selected from the electoral roll by a random procedure designed to yield an age and regional distribution similar to that of the cases; enrollment to vote is compulsory in Australia. Women with a history of ovarian cancer or bilateral oophorectomy, or who were incapable of completing the questionnaire, were ineligible for inclusion in the control series. Two cases and two controls whose calendars contained insufficient information on contraceptive practices were excluded, as were those cases known to be not on the electoral roll. Ethical approval for the study was obtained from all participating hospitals and institutions, and all women in the study, as well as all treating doctors, provided informed consent. Trained interviewers administered a standard questionnaire in person either in the clinic (cases) or in the woman’s home (some cases, all controls). The topics covered included demography, medical and surgical history, and family history of cancer. We collected details of each woman’s reproductive experience by means of a pregnancy and lactation record, including the time to return of menses following pregnancy, and a month-by-month calendar dealing with contraception and attempts to become pregnant during the reproductive years. The questionnaire used is available from the study’s web site: http://www.qimr.edu.au/research/labs/davidp/swh. MATERIALS AND METHODS We ascertained all histologically confirmed incident cases of primary epithelial ovarian cancer registered in all major gynecologic-oncology treatment centers in three Australian Received for publication February 11, 2000, and accepted for publication September 18, 2000. Abbreviation: CI, confidence interval. 1 Queensland Institute of Medical Research, Brisbane, Queensland, Australia. 2 Department of Social and Preventive Medicine, The University of Queensland, Brisbane, Queensland, Australia. Reprint requests to Dr. David Purdie, Queensland Institute of Medical Research, Post Office, Royal Brisbane Hospital, Brisbane, Queensland 4029, Australia (e-mail: [email protected]). 860 Risk Factors for Mucinous and Nonmucinous Ovarian Cancer 861 This procedure yielded 791 cases of epithelial ovarian cancer and 853 controls with data suitable for analysis, reflecting response rates of 90 percent and 73 percent, respectively. Based on histologic review, cases were classified into the following subtypes: 114 (14.4 percent) as mucinous, 415 (52.5 percent) as serous, 99 (12.5 percent) as endometrioid, and 57 (7.2 percent) as clear cell. The remainder (n 106, 13.4 percent) were mainly of mixed or undifferentiated histologic type, two being transitional cell (Brenner) tumors. The factors examined were duration of oral contraceptive use, parity, breastfeeding, tubal sterilization, hysterectomy, history of breast or ovarian cancer in a first-degree relative, and infertility as measured by self-reported difficulty in conceiving (7). For each woman we also calculated her ovulatory life within each decade of life between the age at menarche and the age at menopause (8), diagnosis, or interview, as appropriate. From the length of each full or partial decade we subtracted the total reported anovulatory period in that decade due to pregnancy, postpartum amenorrhea, oral contraceptive use, and/or amenorrhea due to illness, underweight, or other causes. We then weighted the results from each decade by the intensity of ovulation in that decade of life applicable to this population (9) and summed. We adjusted the total number of ovulations upward or downward depending on the extent to which each participant’s reported average menstrual interval was proportionately lower or higher, respectively, than a standard 4 weeks. The total number of ovulations was then divided by 13, the average number of menstrual cycles per year, to create the variable “ovulation years.” Age, years of oral contraceptive use (among users), and ovulation years were all approximately normally distributed, so mean differences were assessed using Student’s t test. Differences in the prevalence of exposure between cases with mucinous and nonmucinous ovarian tumors were tested using Pearson’s chi-square statistic. Multivariate analysis was by means of polytomous logistic regression in which the response categories were mucinous cancers, nonmucinous cancers, and controls. We conducted two analyses that differed only in that the factor ovulatory years was omitted in the first model but included in the second. Additional covariates were age in years and its square, body mass index (split at the 65th and 85th percentiles), smoking (current; past; never), alcohol consumption (none; 0.1–2.5 g/day; >2.5 g/day; unknown), and hormone replacement therapy, all of which have previously been found to be associated with disease status in these data (6, 7, 10). Agesquared was included to allow for residual confounding inadequately accounted for by the linear age term. This had the same effect as adjusting for age in 5-year bands with greater statistical efficiency. The net effect of ovulation years on the risk of mucinous and nonmucinous tumors was assessed in a model not including factors from which it was constructed, namely, duration of oral contraceptive use, parity, and breastfeeding. Significance levels for the comparisons between mucinous and nonmucinous forms of ovarian cancer were derived from a similar logistic model comparing mucinous with nonmucinous cancers in the absence of the control Am J Epidemiol Vol. 153, No. 9, 2001 group. The statistical software packages SAS (11) and SUDAAN (12) were used to conduct all analyses. RESULTS Table 1 lists the percentages of cases of mucinous and other histologic types, along with controls, according to their risk factor profile. Means, along with 95 percent confidence intervals, for age, duration of oral contraceptive use (among users), and years of ovulatory life are also presented for the three groups. Women with mucinous ovarian tumors were in general considerably younger than were cases with nonmucinous tumors or controls, although mucinous tumors did occur in women over the whole age range. The percentage of mucinous tumors that were of borderline malignancy was significantly higher than that for nonmucinous tumors (table 1). Crudely, significant differences between mucinous and nonmucinous tumors were seen for breastfeeding, family history of ovarian cancer, menopausal status, ever use of oral contraceptives, and mean duration of ovulatory life. Some of these differences, in particular the last three, can be attributed to the 10-year mean age difference between cases with mucinous and nonmucinous tumors. The odds ratios (and 95 percent confidence intervals) of mucinous and nonmucinous tumors versus controls for the factors included in the polytomous logistic regression models are presented in table 2, along with the significance level of the mucinous versus nonmucinous comparison (i.e., without controls). The duration of oral contraceptive use had a lesser protective effect, at a marginally significant level, among cases with mucinous cancer than among other cases. The protective effects for parity were also smaller in the mucinous group, albeit not significantly so. Differences between the mucinous and nonmucinous groups with respect to hysterectomy, tubal sterilization, and breastfeeding were not significant. A history of self-reported infertility was positively associated with the occurrence of nonmucinous tumors only. The difference between mucinous and nonmucinous tumors, however, was not significant. A history of breast or ovarian cancer in a first-degree relative showed a significant positive association with nonmucinous tumors only. The odds ratios associated with each year’s ovulation assessed without adjusting for oral contraceptive use, parity, and breastfeeding were 1.01 (95 percent confidence interval (CI): 0.98, 1.04) for mucinous and 1.05 (95 percent CI: 1.04, 1.07) for nonmucinous tumors (p value for difference 0.003). In light of this difference, we conducted a second analysis to examine the effects of the reproductive variables with the inclusion of ovulatory years in the multivariate model (table 2). Once ovulation years were taken into consideration, there was no longer a meaningful difference between the two histologic groups in their odds ratios for the duration of oral contraceptive use, and there was less difference between the coefficients for parity. The odds ratios for hysterectomy, tubal sterilization, breastfeeding, self-reported infertility, and history of breast or ovarian cancer in a first-degree rel- 862 Purdie et al. TABLE 1. Distribution of demographic and risk factor data for cases with mucinous and nonmucinous ovarian tumors and controls, Australia, 1990–1993 Factor Controls (n = 853) (%) Mucinous (n = 114) (%) Nonmucinous (n = 677) (%) p value*,† 64 57 40 11 68 <0.01 <0.01 15 8 31 46 62 70 20 23 7 23 16 25 36 65 58 14 13 7 22 13 29 35 44 64 14 13 13 0.81 <0.01 <0.01 0.97 0.96 0.06 7 4 13 <0.01 Borderline Postmenopausal Parity 0 1 2 3 or more Ever oral contraceptive use Ever breastfed Hysterectomy Tubal sterilization Self-reported infertility Family history of breast or ovarian cancer Controls (n = 853) Mean Age (years) Oral contraceptives (years of use among users) Ovulation (years) 95% CI§ Mucinous (n = 114) Nonmucinous (n = 677) Mean 95% CI Mean 95% CI p value*,‡ 55 54, 56 47 45, 50 57 56, 58 <0.01 7.6 20 7.1, 8.1 19, 21 6.5 21 5.3, 7.7 19, 23 5.6 29 5.0, 6.1 28, 30 0.15 <0.01 * Mucinous versus nonmucinous ovarian tumors. † Pearson’s chi-square test of association. ‡ Student’s t test. § CI, confidence interval. ative were not materially affected by the inclusion of ovulation years (table 2). Because ovulatory life is closely associated with age in premenopausal women, but not in postmenopausal women, and because women with mucinous tumors were considerably younger on average than were women with nonmucinous tumors, the analysis was repeated for pre- and postmenopausal women separately. The differences between mucinous and nonmucinous tumors seen overall did not appear to differ within the two menopausal groups. In addition, because of the excess of borderline tumors among those of mucinous histology, we repeated the analysis separately for borderline (n 139) and frankly malignant (n 652) tumors. Nearly identical effects were seen for mucinous and nonmucinous tumors among the borderline and malignant tumor groups. For instance, the odds ratio associated with each year’s ovulation in borderline tumors was 1.01 for mucinous (95 percent CI: 0.98, 1.05) and 1.07 for nonmucinous (95 percent CI: 1.03, 1.10) tumors, while among frankly malignant tumors, the risks were 1.01 for the mucinous subtype (95 percent CI: 0.97, 1.04) and 1.05 for nonmucinous subtypes (95 percent CI: 1.04, 1.07). DISCUSSION A number of authors have compared the effects of reproduction-related exposures on the risk of mucinous and nonmucinous ovarian tumors (1–5). The most consistent differences appear to be that the protective effects of parity and oral con- traceptives may be restricted to the nonmucinous forms. Most of these studies, however, have had small numbers of cases of mucinous histology and have relied on hospital records for classification of tumor types. These deficiencies could reduce the ability of these studies to detect any true differences in risk factors between the tumor types. Our findings (unadjusted for ovulation years) are consistent with those reported previously in that the effects of parity and oral contraceptive use were less marked in mucinous than in nonmucinous tumors. However, both increasing parity and longer duration of use of oral contraceptives were negatively associated with the occurrence of both tumor groups. We took the analysis a step further and observed that the differences between them appeared to be driven by a significant difference in the effect of ovulation. The increasing number of lifetime ovulations appeared to increase the risk of nonmucinous tumors only and, once this was taken into account, the differences between the risks of mucinous and nonmucinous tumors associated with the duration of oral contraceptive use and pregnancy diminished. Mucinous ovarian tumors are histologically and clinically similar to adenocarcinomas of the uterine cervix (13), and these tumors may also share some risk factors. The risk of cervical adenocarcinoma is not reduced by oral contraceptive usage (14) or by parity (15). Previous studies have also shown a low occurrence of mucinous tumors among women with a family history of ovarian cancer (1, 16–19). This finding was supported in our data, with a positive association being seen between a famAm J Epidemiol Vol. 153, No. 9, 2001 Risk Factors for Mucinous and Nonmucinous Ovarian Cancer 863 TABLE 2. Multivariate odds ratios and their 95 percent confidence intervals for mucinous and nonmucinous forms of epithelial ovarian cancer compared with controls according to reproductive risk factors, by control of lifetime ovulations, Australia, 1990–1993 Mucinous Factor OR†,‡ Nonmucinous 95% CI† OR‡ 95% CI p value* Without control for number of lifetime ovulations Parity 0 1 2 3 or more Ever use of oral contraceptives Duration of oral contraceptive use (per year) Ever breastfed Hysterectomy Tubal sterilization Self-reported infertility Family history of breast or ovarian cancer 1.00 2.06 0.89 0.91 0.62 0.94, 0.38, 0.39, 0.37, 4.49 2.11 2.11 1.04 1.00 1.08 0.59 0.41 0.52 0.66, 0.37, 0.26, 0.39, 1.75 0.92 0.66 0.68 0.15§ 0.14 0.29 0.04 0.41 0.95 0.86 0.89 0.41 0.79 0.91, 0.48, 0.50, 0.23, 0.34, 0.99 1.54 1.58 0.76 1.88 0.92 1.12 0.58 0.61 1.21 0.89, 0.80, 0.42, 0.45, 0.81, 0.94 1.56 0.79 0.84 1.81 0.09 0.40 0.24 0.12 0.18 0.66 0.25, 1.73 1.87 1.29, 2.71 0.03 With control for number of lifetime ovulations Parity 0 1 2 3 or more Ever use of oral contraceptives Duration of oral contraceptive use (per year) Ever breastfed Hysterectomy Tubal sterilization Self-reported infertility Family history of breast or ovarian cancer 1.00 1.95 0.84 0.80 0.61 0.90, 0.36, 0.35, 0.36, 4.21 1.96 1.85 1.04 1.00 1.12 0.61 0.45 0.64 0.69, 0.39, 0.28, 0.48, 1.82 0.96 0.72 0.85 0.35§ 0.24 0.38 0.11 0.98 0.92 0.84 0.87 0.42 0.77 0.88, 0.47, 0.49, 0.23, 0.32, 0.97 1.51 1.56 0.78 1.83 0.93 1.12 0.59 0.60 1.24 0.90, 0.80, 0.43, 0.44, 0.83, 0.96 1.56 0.80 0.83 1.85 0.97 0.38 0.31 0.14 0.17 0.67 0.26, 1.77 1.88 1.30, 2.73 0.03 * Mucinous versus nonmucinous ovarian tumors. † OR, odds ratio; CI, confidence interval. ‡ Adjusted for age in years, age squared, body mass index, smoking, and alcohol consumption as well as all other variables listed in the table (years of oral contraceptive use were not included in the model examining the effect of ever use of oral contraceptives). § Overall significance level for parity. ily history of breast or ovarian cancer and nonmucinous tumors only. There was a nonsignificant difference seen between the effect of a reported history of infertility and the two tumor classes, with a positive association seen in nonmucinous tumors only. This finding is in contrast to that reported by Wittenberg et al. (5) who found that the risk associated with infertility was restricted to mucinous tumors only. It should be noted here that women’s self reported difficulty in conceiving is subject to variable interpretation. Infertility is a heterogeneous combination of disorders, which may have very different etiologies. No significant differences were observed between the risks of mucinous and nonmucinous tumors for tubal sterilization, hysterectomy, and breastfeeding. Wittenberg et al. (5) found an inverse association with hysterectomy among Am J Epidemiol Vol. 153, No. 9, 2001 women with mucinous tumors only; however, in our data the protective effect of hysterectomy, if anything, was greater in nonmucinous tumors. Differences in the effect of these factors on the occurrence of mucinous and nonmucinous ovarian tumors have never been seen consistently, and any observed differences seem quite likely to be due to random variation. In conclusion, this study has provided evidence to support the conclusions reached by Risch et al. (1) that ovarian cancer is an etiologically heterogeneous collection of tumors, and important risk factors such as ovulation and genetic susceptibility may apply only to a subset of ovarian cancers, namely, the nonmucinous subtypes. Data we have presented elsewhere (10), with respect to the effects of exposure to hormone replacement therapy, also fit with this interpretation. 864 Purdie et al. ACKNOWLEDGMENTS Funding for this study was provided by the Australian National Health and Medical Research Council and the Queensland Cancer Fund. The contribution made by other members of the Survey of Women’s Health Study Group has been acknowledged previously (7). 8. 9. 10. 11. 12. REFERENCES 1. Risch HA, Marrett LD, Jain M, et al. Differences in risk factors for epithelial ovarian cancer by histologic type: results of a case-control study. Am J Epidemiol 1996;144:363–72. 2. Cramer DW, Hutchison GB, Welch WR, et al. Factors affecting the association of oral contraceptives and ovarian cancer. N Engl J Med 1982;307:1047–51. 3. Kvåle G, Heuch I, Nilssen S, et al. Reproductive factors and risk of ovarian cancer: a prospective study. Int J Cancer 1988;42:246–51. 4. WHO Collaborative Study of Neoplasia and Steroid Contraceptives. Epithelial ovarian cancer and combined oral contraceptives. Int J Epidemiol 1989;18:538–45. 5. Wittenberg J, Cook LS, Rossing MA, et al. Reproductive risk factors for mucinous and non-mucinous epithelial ovarian cancer. Epidemiology 1999;10:761–3. 6. Siskind V, Green A, Bain C, et al. Beyond ovulation: oral contraceptives and epithelial ovarian cancer. Epidemiology 2000; 11:106–10. 7. Purdie D, Green A, Bain C, et al. Reproductive and other fac- 13. 14. 15. 16. 17. 18. 19. tors and risk of epithelial ovarian cancer: an Australian casecontrol study. Int J Cancer 1995;62:678–84. Siskind V, Green A, Bain C, et al. Breastfeeding, menopause and epithelial ovarian cancer. Epidemiology 1997;8:188–91. Metcalf MG. Incidence of ovulation from the menarche to the menopause: observations of 622 New Zealand women. N Z Med J 1983;96:645–8. Purdie DM, Bain CJ, Siskind V, et al. Hormone replacement therapy and risk of epithelial ovarian cancer. Br J Cancer 1999;81:559–63. SAS Institute, Inc. SAS OnlineDoc, version 8 ed. Cary, NC: SAS Institute, Inc, 1999. Shah BV, Barnwell BG, Bieler GS. SUDAAN user’s manual. Release 7.0. Research Triangle Park, NC: Research Triangle Institute, 1996. Parmley TH, Woodruff JD. The ovarian mesothelioma. Am J Obstet Gynecol 1974;120:234–41. Ursin G, Peters RK, Henderson BE, et al. Oral contraceptive use and adenocarcinoma of the cervix. Lancet 1994;344: 1390–4. Parazzini F, La Vecchia C, Negri E, et al. Risk factors for adenocarcionoma of the cervix: a case-control study. Br J Cancer 1988;57:201–4. Greggi S, Genuardi M, Benedetti-Panici P, et al. Analysis of 138 consecutive ovarian cancer patients: incidence and characteristics of familial cases. Gynecol Oncol 1990;39:300–4. Bewtra C, Watson P, Conway T, et al. Hereditary ovarian cancer: a clinicopathological study. Int J Gynecol Pathol 1992; 11:180–7. Piver MS, Baker TR, Jishi MF, et al. Familial ovarian cancer: a report of 658 families from the Gilda Radner Familial Ovarian Cancer Registry 1981–1999. Cancer 1993;71:582–8. Narod S, Tonin P, Lynch H, et al. History of BRCA1-associated ovarian tumours. (Letter). Lancet 1994;343:236. Am J Epidemiol Vol. 153, No. 9, 2001