Serous ovarian, fallopian tube and primary peritoneal cancers: A comparative epidemiological analysis
Invasive serous cancers are diagnosed in the ovary, fallopian tube and peritoneum. It is widely believed that these are variants of the same malignancy but little is known about fallopian tube and primary peritoneal cancers. A comparison of risk factors for these tumor types may shed light on common or distinct aetiological pathways involved in these types of cancer. We investigated risk factors for the three cancers using data from a large Australian population-based case-control study. We included women with incident invasive serous ovarian (n = 627), primary peritoneal (n = 129) and fallopian tube (n = 45) cancer and 1,508 control women. Participants completed a comprehensive reproductive and lifestyle questionnaire. Logistic regression was used to calculate odds ratios (ORs) and 95% confidence intervals (CIs). Hormonal contraceptive use was inversely related to risk of all three cancers. Parity and breast-feeding were also inversely related to risk of serous ovarian and fallopian tube cancer. In contrast, parous women had an increased risk of peritoneal cancer (OR = 1.8, 95%CI 0.8–3.9), and increasing parity did not lower risk. There was also no association between breast-feeding and peritoneal cancer. However, obesity was associated with a doubling of risk for peritoneal cancer alone (OR = 2.1, 95%CI = 1.3–3.4). The strikingly similar patterns of risk for serous ovarian and fallopian tube cancers and the somewhat different results for primary peritoneal cancer suggest that peritoneal cancers may develop along a different pathway. These results also call into question the role of the physical effects of ovulation in the development of serous ovarian cancer. © 2007 Wiley-Liss, Inc.
Approximately 90% of ovarian cancers appear to arise from the surface epithelium of the ovary and these epithelial ovarian cancers affect around 15 per 100,000 women/year in the United States.1 Serous cancers, the most common subtype, account for 40–50% of invasive ovarian tumors.1, 2 Serous cancers also occur in the peritoneum and fallopian tube. Primary peritoneal cancer is diagnosed at approximately one tenth the frequency of epithelial ovarian cancer,3 and fallopian tube cancer even less frequently.4
At present the origins of serous ovarian cancers are poorly understood. While many believe that serous ovarian cancers originate from the ovarian surface epithelium or epithelial-lined inclusion cysts within the ovarian cortex,5 others have suggested that serous ovarian cancers, and indeed serous peritoneal cancers, may originate in fallopian tube epithelium.6 Traditionally, serous cancers of the ovary, peritoneum and fallopian tube have been classified as separate diseases, however given their close histological and clinical similarities, all three may be variants of the same malignancy.6 If this were the case it would have implications for our understanding of both the cellular origins of these cancers, and the carcinogenic processes involved in their development.
The incessant ovulation theory of ovarian carcinogenesis7 was proposed in light of the observed decrease in risk of epithelial ovarian cancers because of childbearing and oral contraceptive use. It suggests that the recurrent breakdown and repair of the ovarian epithelium from repeated ovulation increases the risk of cell damage and subsequent neoplastic transformation. However, this mechanism cannot account for cancers arising in the peritoneum or fallopian tube. A common aetiology for serous ovarian cancer, primary peritoneal and fallopian tube cancers would thus argue against the physical aspects of ovulation underlying serous ovarian carcinogenesis, and suggest that hormonal or biochemical factors may be responsible.
At present there are no comparative etiologic data with which to address this issue. Although the epidemiology of serous ovarian cancer has been extensively studied,8–10 little has been published on the epidemiology of fallopian tube and primary peritoneal cancers due to their relative rarity. Scant data from one case-control study, plus case-case studies and case series hint at similarities in reproductive risk factors among all three.3, 11–15 We have therefore investigated the risk factors for fallopian tube and primary peritoneal cancers individually and have compared them with those for invasive serous ovarian cancer using data from an Australian case-control study conducted between 2002 and 2005.
Material and methods
The Australian Ovarian Cancer Study (AOCS) was an Australia-wide population-based case-control study. Between January 2002 and June 2005 women aged 18–79 years with suspected epithelial ovarian cancer (including, fallopian tube or primary peritoneal cancers), were recruited primarily in gynecologic-oncology units by research nurses. Where possible, they were recruited prior to surgery (and therefore histopathological diagnosis) to allow collection of fresh tumor tissue. Women missed at treatment centers were identified through the state-based cancer registries and, with their treating doctor's permission, invited to participate (recruitment through the New South Wales and Victorian Cancer Registries was conducted in parallel under a separately funded study: the Australian Cancer Study).
A total of 3,534 women were identified with suspected ovarian cancer. Of these, 304 died before contact could be made, the treating doctor refused contact with 133, and 194 women could not be contacted. A further 167 were excluded on the basis of language difficulties (n = 70), mental incapacity (n = 33) and illness (n = 64). The remaining 2,736 women were invited to participate and, of these, 2,299 (84% of those approached and 65% of those originally identified) agreed to take part. All participants were asked to self-complete a detailed health and lifestyle questionnaire. The questionnaire covered demographic and physical characteristics, family history, medical/surgical history, lifestyle habits and reproductive and contraceptive histories. Missing information and/or inconsistencies were clarified during a subsequent telephone interview.
Histopathology reports were obtained for all those who agreed to take part and we abstracted information on tumor site, histological subtype, invasive behavior (invasive or borderline) and grade. Reports were categorized as complex (17% of all reports) if the above tumor characteristics were not clearly stated. Several quality assurance checks on the abstracted histopathology data were undertaken. Random samples of 109 complex reports and 65 straightforward reports were reabstracted by a gynecologic pathologist. For straightforward reports there was 97% agreement with the original abstractions. For complex reports the agreement was 85%, but many of the discrepancies resulted from conservative labeling by the original abstractors. For example, if the primary site was not clear, the original abstractors were likely to label it as unknown, while the pathologists were more likely to call it ovarian. For a further sample of 194 women, pathology reports and diagnostic slides were reviewed by a gynecologic pathologist and the agreement with the original abstracted data was more than 96%.
Finally, based on the abstracted pathology data 612 women were excluded because they did not have a primary incident epithelial ovarian, peritoneal or fallopian tube cancer. Of the final 1687 eligible participants, 1,576 (93%) returned a questionnaire. Here we sought to analyze invasive serous cancers of the ovary, fallopian tube and peritoneum, so for the current analyses women with borderline tumors (n = 318), nonserous cancers (n = 420) or where the primary site could not clearly be identified as ovary, peritoneum or fallopian tube (n = 37) were excluded. Only cancers specifically described by the term “serous” were included here and we classified tumors as fallopian tube or peritoneal only if there was no or at most minimal ovarian involvement. The final study group included 627 women with invasive serous ovarian cancer, 129 with serous primary peritoneal cancer and 45 with serous fallopian tube cancer.
We randomly selected potential control women aged 18–79 years from the Australian Electoral Roll, frequency matched to the entire case series by age (in five-year groups) and state of residence. Enrolment to vote is compulsory in Australia. Selected women were mailed study information and subsequently telephoned by research nurses. If a woman could not be contacted after five attempts, or if a telephone number could not be found, a second letter was sent.
A total of 3,600 potential control women were contacted and invited to participate. Of these 61 were excluded due to illness and 97 due to language difficulties. Of the 3,442 eligible women 1,611 (47%) agreed to participate and returned a questionnaire as for cases. Of these, six women reported previous ovarian cancer and 97 reported a previous bilateral oophorectomy and thus were excluded leaving 1,508 control women.
This study was approved by the Human Research Ethics Committees at the Queensland Institute of Medical Research, Peter MacCallum Cancer Centre, University of Melbourne and all participating hospitals and cancer registries.
Effects were estimated by odds ratios (OR) with 95% confidence intervals (CIs). Tests of statistical significance were two-sided. Unconditional multiple logistic regression was used to simultaneously adjust for potential confounding factors. Polytomous regression was used to compare estimates for the three cancer types using women with serous ovarian cancer as the reference group. For categorical variables with more than two categories, linear trends were compared across the three cancer types.
For this comparative analysis we investigated factors for which an association with ovarian cancer risk is well-supported by evidence from the literature. The contraceptive variable represented use of progestin-only, combination oral and injected (depot and implant formulations) contraceptives combined. The hormone replacement therapy (HRT) variable included use of oestrogen-only as well as combined oestrogen/progestin regimens (continuous and sequential use). For analyses of parity we included all pregnancies of greater than six months' duration. Duration of breast-feeding was summed across all pregnancies and only women who had had a live-birth were included in these analyses. Maximum body mass index (BMI) was derived using self-reported height and maximum nonpregnant lifetime weight and divided into categories according to World Health Organization criteria.16 We excluded exposure to factors of interest in the 12 months prior to diagnosis for cases, or the 12 months prior to first approach for controls because we considered the influence of recent exposures on the aetiology of ovarian cancer to be minimal given the likely developmental latency of the disease. It was also considered that for cases recent behaviors may have been affected by preclinical disease.
All models, unless otherwise stated, included a term for age (in 10 year brackets, as this provided the best adjustment for age while maintaining parsimony), parity, contraceptive use, family history of breast or ovarian cancer in a first degree relative, tubal sterilization, hysterectomy and education. Other potential confounders which did not change risk estimates by more than 10% were excluded from the final models. To investigate trends for ordinal variables, the median value for each category was used in a continuous term and the associated p-value assessed. All analyses were conducted using SAS statistical software (version 9.1, SAS Institute, Cary, NC).
Women diagnosed with fallopian tube or primary peritoneal cancer were on average significantly older than those diagnosed with serous ovarian cancer (p = 0.02 and p < 0.0001 respectively) (Table I). The mean age of controls was significantly lower than all three groups of case women because controls were frequency matched to the entire national ovarian cancer case-series which included women with borderline tumors and mucinous cancer that were, on average, younger than women with invasive serous cancers. To maximize power we used the entire control group here and adjusted all analyses for age. Controls were also more likely to have some postschool education (p = 0.02 compared to women with ovarian cancer).
Table I. Selected Personal Characteristics of Study Subjects
|Mean age (years)||56.4||59.7||63.32||63.93|
|Highest level of education|
|High school||740 (49)||353 (56)||28 (62)||76 (59)|
|Technical||550 (37)||190 (30)||11 (25)||41 (32)|
|University||218 (14)||84 (14)||6 (13)||12 (9)|
|Ethnicity|| || || || |
|White||1,447 (96)||600 (96)||43 (96)||128 (99)|
|Asian||32 (2)||19 (3)||0 (0)||1 (1)|
|Other||25 (2)||6 (1)||1 (2)||0 (0)|
Ever having a term pregnancy was associated with a decreased risk of serous ovarian cancer (OR = 0.65, 95%CI 0.48–0.88) and there was a significant trend of decreasing risk with increasing parity (p = 0.02) (Table II). Although limited by the small sample size, very similar results were seen for fallopian tube cancers (OR = 0.47, 95%CI 0.22–1.11 for parous vs. nulliparous, p = 0.04 for trend). In contrast, ever having a term pregnancy was associated with a nonsignificantly increased risk of peritoneal cancer (OR = 1.78, 95%CI 0.77–3.88) and increasing parity did not lower risk.
Table II. Adjusted Odds Ratios and 95% Confidence Intervals for the Association Between Reproductive and Hormonal Factors and Serous Gynecological Cancers by Primary Site
|Parous|| || || || || || || |
| No||176 (12)||91 (15)||8 (18)||7 (6)||1.00||1.00||1.00|
| Yes||1,311 (88)||534 (85)||37 (82)||119 (94)||0.65 (0.48–0.88)||0.47 (0.22–1.11)||1.78 (0.77–3.88)|
|Parity|| || || || || || || |
| 0||176 (12)||91 (15)||8 (18)||7 (6)||1.00||1.00||1.00|
| 1||141 (9)||65 (10)||8 (18)||12 (9.5)||0.85 (0.56–1.29)||0.93 (0.32–2.76)||1.87 (0.70–4.99)|
| 2||497 (33)||191 (31)||14 (31)||38 (30)||0.66 (0.47–0.91)||0.50 (0.19–1.30)||1.67 (0.71–3.91)|
| 3||404 (27)||150 (24)||7 (15)||39 (31)||0.56 (0.40–0.80)||0.29 (0.10–0.85)||1.77 (0.75–4.17)|
| 4||185 (12)||74 (12)||4 (9)||18(14)||0.56 (0.38–0.85)||0.29 (0.08–1.06)||1.49 (0.58–3.82)|
| ≥5||84 (6)||53 (8)||4 (9)||12(9.5)||0.81 (0.51–1.29)||0.60 (0.16–2.19)||2.01 (0.73–5.49)|
| || || || || ||ptrend = 0.02||ptrend = 0.04||ptrend = 0.4|
|p-value compared to ovarian cancer2|| ||p = 0.3||p = 0.08|
|Per birth||0.93 (0.86–0.99)||0.85 (0.68–1.07)||1.06 (0.93 1.21)|
| Duration of breast-feeding (months)3|
| Never||204 (16)||115 (23)||11 (31)||16 (15)||1.00||1.00||1.00|
| ≤6||296 (23)||140 (27)||9 (25)||27 (25)||0.93 (0.68–1.28)||0.52 (0.20–1.35)||1.22 (0.63–2.36)|
| 7–12||269 (21)||91 (18)||10 (28)||25 (23)||0.66 (0.47–0.94)||0.83 (0.33–2.10)||1.26 (0.63–2.50)|
| 13–24||319 (25)||164 (21)||6 (17)||39 (36)||0.66 (0.47–0.93)||0.24 (0.08 0.73)||1.08 (0.55–2.11)|
| 25–35||133 (10)||45 (9)|| || ||0.63 (0.40–0.99)|| || |
| ≥36||65 (5)||12 (2)|| || ||0.26 (0.13–0.54)|| || |
| || || || || ||ptrend < 0.0001||ptrend = 0.03||ptrend = 0.9|
|p-value compared to ovarian cancer2|| ||p = 0.2||p = 0.3|
|Per month of breast feeding|| || || || ||0.98 (0.97–0.99)||0.96 (0.93–1.00)||1.00 (0.98–1.02)|
| Duration of hormonal contraceptive use (months)|
| Never||322 (22)||206 (33)||17 (40)||47 (37)||1.00||1.00||1.00|
| 1–12||130 (9)||79 (13)||3 (7)||16 (13)||1.02 (0.72–1.44)||0.63 (0.17–2.24)||1.08 (0.58–2.03)|
| 13–60||298 (20)||129 (21)||10 (23)||22 (17)||0.71 (0.53–0.95)||0.97 (0.41–2.31)||0.74 (0.42–1.32)|
| 61–120||314 (21)||103 (17)||8 (19)||21 (17)||0.52 (0.38–0.70)||0.69 (0.27–1.75)||0.67 (0.38–1.20)|
| 121–180||208 (14)||62 (10)||4 (9)||11 (7)||0.51 (0.36–0.73)||0.55 (0.17–1.76)||0.57 (0.28–1.17)|
| 181–240||127 (9)||27 (4)||1 (2)||4 (3)||0.36 (0.23–0.58)||0.12 (0.02–0.97)||0.39 (0.13–1.13)|
| >240||83 (6)||13 (2)|| ||6 (5)||0.22 (0.12–0.42)|| ||0.77 (0.30–1.94)|
| || || || || ||ptrend < 0.0001||ptrend = 0.03||ptrend = 0.06|
|p-value compared to ovarian cancer2|| ||p = 0.95||p = 0.30|
|Per year of use||0.95 (0.93–0.96)||0.94 (0.89–0.99)||0.97 (0.94–1.00)|
| Duration of hormone replacement therapy (months)|
| Never||952 (65)||361 (59)||31(69)||71 (58)||1.00||1.00||1.00|
| <24||125 (8)||50 (8)||2 (4)||13 (11)||0.94 (0.65–1.35)||0.47 (0.11–2.06)||1.20 (0.63–2.29)|
| 24–60||129 (9)||45 (7)||5(11)||16 (13)||0.81 (0.55–1.19)||0.95 (0.35–2.58)||1.34 (0.73–2.45)|
| >60–120||115 (8)||63 (10)||3 (7)||8 (7)||1.21 (0.85–1.73)||0.62 (0.18–2.13)||0.65 (0.30–1.43)|
| >120||149 (10)||95 (15)||4 (9)||14 (11)||1.43 (1.03–1.97)||0.55 (0.18–1.67)||0.69 (0.37–1.30)|
| || || || || ||ptrend = 0.02||ptrend = 0.29||ptrend = 0.14|
|p-value compared to ovarian cancer2|| ||p = 0.17||p = 0.04|
A similar pattern was seen for breast-feeding (Table II). We observed significant inverse trends with longer duration of breast-feeding for both ovarian and fallopian tube cancers (p < 0.0001 and p = 0.03 respectively), equivalent to risk reductions of 2% per month of breast-feeding for ovarian cancer (OR = 0.98, 95%CI 0.97–0.99) and 4% for fallopian tube cancer (OR = 0.96, 95%CI 0.93–1.00). We found no evidence that breast-feeding was associated with risk of primary peritoneal cancer.
Use of hormonal contraceptives was associated with risk reductions for all three cancers, with statistically significant trends of decreasing risk with increasing duration of contraceptive seen for ovarian and fallopian tube cancers (a reduction in risk of 5 and 6% per year of use respectively) and a slightly weaker effect for peritoneal cancer (3% per year, p = 0.05).
Long-term use of HRT was associated with a 40% increase in risk of serous ovarian cancer (OR = 1.43, 95%CI 1.03–1.97 for >120 months compared to no use, p for trend = 0.02), but nonsignificant inverse relations with both fallopian tube and peritoneal cancers (OR = 0.55, 95%CI 0.18–1.67 and OR = 0.69, 95%CI 0.37–1.30 respectively) although the associated confidence intervals were wide. The relationship with HRT differed significantly for peritoneal cancer compared to serous ovarian cancer (p = 0.04).
We observed weak to modest inverse relations with tubal sterilization across all three cancer types (Table III). Hysterectomy was associated with a small excess risk of ovarian and peritoneal cancers (statistically significant for the former), but a nonsignificant inverse association was seen with fallopian tube cancer.
Table III. Adjusted Odds Ratios and 95% Confidence Intervals for the Association Between Selected Medical, Personal and Lifestyle Factors and Serous Gynecological Cancers by Primary Site
| No||1,082 (73)||462 (74)||36 (80)||97 (76)||1.00||1.00||1.00|
| Yes||405 (27)||162 (26)||9 (20)||30 (24)||0.87 (0.69–1.09)||0.74 (0.34–1.63)||0.73 (0.46–1.14)|
|p-value compared to ovarian cancer2|| ||p = 0.95||p = 0.65|
|Hysterectomy|| || || || || || || |
| No||1,201 (81)||464 (74)||37 (82)||93 (73)||1.00||1.00||1.00|
| Yes||286 (19)||160 (26)||8 (18)||34 (27)||1.27 (1.00–1.60)||0.71 (0.32–1.58)||1.14 (0.74–1.75)|
|p-value compared to ovarian cancer2|| ||p = 0.2||p = 0.6|
|First degree relative with breast/ovarian cancer|
| No||1,296 (87)||495 (79)||38 (84)||102 (80)||1.00||1.00||1.00|
| Yes||191 (13)||129 (21)||7 (16)||25 (20)||1.71 (1.33–2.21)||1.11 (0.48–2.57)||1.41 (0.88–2.27)|
|p-value compared to ovarian cancer2|| ||p = 0.3||p = 0.8|
|Use of perineal talc|
| Never||820 (56)||315 (52)||19 (44)||64 (54)||1.00||1.00||1.00|
| Moderate||337 (23)||134 (22)||13 (30)||28 (23)||1.05 (0.82–1.35)||1.98 (0.93–4.22)||1.08 (0.67–1.74)|
| Substantial||305 (21)||161 (26)||11 (26)||27 (23)||1.31 (1.03–1.67)||1.64 (0.75–3.62)||0.99 (0.61–1.61)|
| || || || || ||ptrend = 0.03||ptrend = 0.43||ptrend = 0.9|
|p-value compared to ovarian cancer3|| ||p = 1.0||p = 0.26|
|Maximum BMI (kg/m2)|
| <25||492 (33)||212 (35)||13 (29)||27 (21)||1.00||1.00||1.00|
| ≥25–30||522 (36)||224 (37)||21 (47)||42 (34)||0.95 (0.75–1.20)||1.54 (0.74–3.22)||1.29 (0.78–2.14)|
| ≥30||458 (31)||167 (28)||11 (22)||56 (45)||0.82 (0.64–1.06)||0.93 (0.39–2.21)||2.10 (1.29–3.44)|
| || || || || ||ptrend = 1.0||ptrend = 0.8||ptrend = 0.0004|
|p-value compared to ovarian cancer3|| ||p = 0.9||p = 0.0003|
Having a first-degree relative with breast or ovarian cancer was associated with a 70% increase in risk of serous ovarian cancer (OR 1.71, 95%CI 1.33–2.21), and smaller nonsignificant excesses for fallopian tube and peritoneal cancers (Table III).
Substantial use of talc in the perineal area was associated with a significant 30% increase in risk of ovarian cancer and a nonsignificant increase in risk of fallopian tube cancer but was not associated with risk of peritoneal cancer (Table III).
Finally, while there appeared to be no relation between BMI and ovarian or fallopian tube cancer (Table III), obesity was associated with a 2-fold increase in risk of peritoneal cancer (OR 2.10, 95%CI 1.29–3.44) with a clear trend of increasing risk with increasing BMI (p = 0.0004). This relationship with BMI was significantly different for peritoneal cancer compared with serous ovarian cancer.
We have presented novel data from case-control analyses of primary peritoneal, fallopian tube and invasive serous ovarian cancers collected in the same study population. Given the close histological and clinical similarities of fallopian tube, peritoneal and serous ovarian cancers we expected the determinants of each cancer to be the same. However, while the risk factors for serous fallopian tube cancer closely paralleled those for serous ovarian cancer, the risk factors for serous peritoneal cancer differed somewhat in several key areas raising the possibility that peritoneal cancer has a different aetiological pathway.
Our findings with respect to the associations between parity, contraceptive use, breast-feeding, BMI, HRT use, tubal sterilization and talc use and risk of serous ovarian cancers are in concordance with previous research.8–10, 17–21
The few published data pertaining to fallopian tube cancer have also suggested that parity decreases the risk of fallopian tube cancer11, 14, 15 but, to our knowledge, relationships with contraceptive use and breast-feeding have not been reported previously. Four small case-case studies have compared women with peritoneal and ovarian cancers and our null finding in relation to parity is supported by the results of three of these. In one, women with peritoneal cancer had significantly more children than women with serous ovarian cancer.12 In two others, women with peritoneal cancer reported an almost 50% lower rate of nulliparity than women with ovarian cancer (14 vs. 24% and 6 vs. 11%, respectively).3, 13 The inverse association we found between contraceptive use and risk of peritoneal cancer concurs with findings of the previous case-case studies.3, 12, 13 We could find no published data evaluating the associations between breast-feeding or obesity and risk of peritoneal cancer.
A limitation of our study was the relatively low participation rate of 47% among controls: it was thus possible that those who took part were systematically different from those who did not. However, our findings in relation to invasive serous ovarian cancer concur with results from previously published studies indicating that selection bias has not materially affected our results. We further assessed this issue by comparing data from our control group to data from the 2001 Australian National Health Survey (NHS), a national population-based survey with a response rate of ∼90%.22 The distributions of education level, parity and BMI among our control women were almost identical to those from the NHS. Use of the oral contraceptives among women younger than 50 years was around 5% higher in our control women than in the NHS, which may have biased our estimates slightly away from the null. We also found that the age-standardized rate of hysterectomy in our controls was 5% lower than in the NHS possibly explaining why we found a small elevation in risk associated with hysterectomy while other studies have found either a null or small inverse effect for serous ovarian cancers. However, it should be noted that these small differences will not have affected the comparisons between the three cancer types.
Another possible source of error was misclassification of cancer subtype or site. Differentiation between primary peritoneal, fallopian tube and serous ovarian cancers can be difficult. However, there are defined criteria for the diagnosis of both fallopian tube and primary peritoneal cancers23 and the consistency across our pathology review suggested that misclassification has been minimal. Moreover we took a conservative approach and only classified tumors as fallopian tube or peritoneal if there was no or at most minimal ovarian involvement. To the extent that any such misclassification occurred, the likely effect would be to bias any differences between the tumor types towards the null and so this cannot explain the differences we have observed for peritoneal cancer.
We considered all known risk factors as potential confounders and it is unlikely that other unknown and/or unmeasured factors are so strongly associated with cancer risk that they could explain observed effects. Odds ratios calculated from fully adjusted models were not substantially different from age-adjusted estimates, suggesting that residual confounding by factors included in the models is unlikely to have been a major problem. Similarly, while the potential for chance-induced differences and similarities cannot be dismissed, the dose-response patterns we have found provide reassurance in this regard.
Our findings in relation to fallopian tube cancer have implications that challenge the incessant ovulation theory7 since we show that factors associated with ovulation also increase risk of fallopian tube cancer and tubal epithelium is not subject to breakdown and repair at ovulation. Our findings thus argue against incessant injury and repair being the primary mechanism of serous ovarian epithelial carcinogenesis. The repeated exposure of ovarian epithelium to follicular fluid at ovulation may be a carcinogenic stimulus however.7 Epithelium of the fimbrial end of the fallopian tube, where most cancers occur,24, 25 is intimately associated with the ovulatory follicle and is also exposed to the hormone and prostaglandin-rich follicular fluid at ovulation. Furthermore, oestrogen, found in high concentrations in follicular fluid, has been found to induce tubal cell growth via epidermal growth factor (EGF),26 while prolonged exposure to progestins, as in pregnancy and oral contraceptive use, induces atrophy of the tubal epithelium, irrespective of the presence of concomitant oestrogen.27 There is also some evidence that a loss of progesterone receptors in tubal epithelium may occur in the process of tubal carcinogenesis25 and these observations fit more closely with the hormonal aetiology and protective role of progestins in epithelial ovarian carcinogenesis proposed by Risch.28
The similarity in risk factors that we have found between serous ovarian and fallopian tube cancer also prompts questions about the cellular origins of these cancers. Some circumstantial evidence for a shared tubal origin for ovarian and fallopian tube cancers is accumulating. In prophylactic salpingo-oophorectomy specimens from women genetically predisposed to ovarian cancer, occult malignancies have been found in the tubes more frequently than in the ovaries.24, 29–31 Furthermore, gene expression in serous ovarian cancers is correlated with gene expression in fallopian tube epithelium, but not ovarian epithelium.32
Serous ovarian cancer is, however, diagnosed much more frequently than fallopian tube cancer. One explanation for this is that most serous cancers are at an advanced stage at diagnosis and, by convention, are labeled ovarian if there is more than minimal ovarian involvement.33 It may thus be that the true incidence of fallopian tube cancer is much higher than is generally recognized.33
In contrast to the results for fallopian tube cancer, our results suggest that risk factors for primary peritoneal cancers may differ from cancers of the ovary and fallopian tube casting doubt on a shared etiology among all three tumors. Studies of clonality provide further evidence of a difference in the pathways to the development of serous ovarian and peritoneal cancers. Although serous ovarian cancer has often spread extensively at the time of diagnosis, molecular studies at different tumor sites suggest that it is clonal in origin.34 In contrast, most studies have suggested that primary peritoneal cancer is multifocal in origin.35–37 Taken together, these potential differences between peritoneal and ovarian cancers imply that future inclusion of peritoneal cancers in studies of epithelial ovarian cancer may be inappropriate.
Serous cancers also occur in the endometrium but, like fallopian tube and peritoneal cancers, are relatively uncommon and their risk factors poorly understood. Women with endometrial cancer were not recruited as part of the current study, but studies comparing risk factors for uterine serous papillary cancers with those for serous ovarian may also provide useful insights into the origins and biology of serous gynecological cancers.
In conclusion, we found remarkably similar risk reductions for serous ovarian cancer and fallopian tube cancer associated with increasing parity, duration of breast-feeding and hormonal contraceptive use. In contrast, the pattern was somewhat different for primary peritoneal cancer, occurrence of which was not reduced by parity or breast-feeding, but was increased by obesity. These results suggest that peritoneal cancers may develop along a different pathway to that involved in development of serous ovarian and fallopian tube cancers and that, contrary to one of the prevailing theories of ovarian cancer pathogenesis, the physical effects of ovulation are unlikely to be the initiating event in serous ovarian carcinogenesis.
SJ is supported by an Australian Postgraduate Award (University of Queensland), DW and PW are supported by Senior Research Fellowships from the National Health and Medical Research Council of Australia and Queensland Cancer Fund, respectively. We gratefully acknowledge the cooperation of the New South Wales, Queensland, South Australian, Victorian and Western Australian Cancer Registries as well as the collaborating institutions represented within the AOCS Study Group (below). We would like to thank Ms. Nirmala Pandeya for conducting comparisons with data from the National Health Survey. We would also like to thank the women who participated in the study.