Strengths of the current study include its population-based design and relatively large size; nevertheless, some subgroup analyses were limited by sample size. Also, similar to most other studies that have assessed the relation of ovarian cysts with risk of ovarian cancer, the interpretation of our results is limited by the absence of information on the types of ovarian abnormalities that occurred, as well as the specific reason that ovarian surgery was performed. It is likely that a range of conditions (e.g., physiologic follicular and corpus luteum cysts; endometrioid and dermoid cysts; and benign serous and mucinous tumors) may have been reported as ovarian cysts. The majority of functional ovarian cysts are thought to resolve within 60–90 days; functional cysts that do not resolve within this time frame, as well as other types of ovarian cysts, are more likely to be treated surgically. The size of a cyst and the age and menopausal status of an affected woman may also influence the likelihood of surgical treatment.
We cannot discount the possibility of incomplete or differential recall by cases and controls, nor can we exclude the possibility that characteristics associated with willingness to participate in the study could have resulted in a non-representative sample of cases or controls. However, recall bias is unlikely to account for differences in risk across tumor subgroups. While differing response proportions across histologic types could influence our results if the exposures examined were linked to our ability to interview cases [16
], our response proportions were broadly similar and relatively high for women with borderline (80.6%) and invasive (75.4%) tumors. Also, response proportions were similar (ranging from 77% for mucinous invasive to 88% for endometrioid invasive disease) for all disease subtypes except those grouped as “other invasive” (largely undifferentiated) tumors, among whom we interviewed only 53% of eligible women. Thus, any bias relating to differential participation according to tumor type, if present, might be expected to most strongly influence findings in this latter subgroup.
The exclusion of women whose diagnosis of a benign gynecologic condition or ovarian surgery occurred within the year before the diagnosis/reference date was done to reduce possible bias resulting from 1) symptoms of an existing, as-yet undiagnosed, cancer leading to these diagnoses or surgery or 2) increased or early detection of an asymptomatic tumor during evaluation or treatment for a benign condition. The potential for such a bias to have otherwise influenced our findings is suggested by the larger proportion of cases than controls who reported a cyst diagnosis in that interval. However, an additional potential bias must also be considered: relative to women without a history of a pelvic condition and/ or surgery (in whom no ovarian evaluation is performed), women who undergo ovarian evaluation with no cancer identified will be less likely to be diagnosed with ovarian cancer for some time period. This sort of “healthy screenee” bias [17
] would be expected to lead to spuriously low estimates of ovarian cancer risk associated with the diagnosis and/or treatment of some pelvic conditions, at least temporarily. Cognizant of this, we performed analyses focusing on cancer risk associated with ovarian surgeries performed at least five years earlier, for by five years it is anticipated that little or no such bias should remain.
Only a few studies have assessed ovarian cancer risk in relation to prior ovarian surgery [18
], and none of these provide data regarding risk of disease subtypes or the possible indication for surgery. In a record linkage study in Canada, the risk of invasive or borderline ovarian cancer among 18,375 women who had undergone unilateral oophorectomy was substantially increased early in the follow-up period (6 months to <2 years after surgery; observed/expected cancers = 3.8, p<0.001), with no association observed from 2 to more than 10 years after surgery. In two other reports, however, unilateral oophorectomy was associated with 40% and 80% reductions in risk of ovarian cancer [19
]. These disparate results, together with the results of the current study, suggest that the indication for surgery-- as well as differences in risk according to tumor subtype-- should be considered in studies that examine the influence of ovarian surgery on cancer risk.
Prior studies that have assessed whether women with a history of benign ovarian cysts might be at altered risk of ovarian cancer provide inconsistent results and, for the most part, do not provide data for histology-specific risks and/or risks among women with and without subsequent ovarian surgery. Thus, direct comparisons to the current study are limited. A large, pooled analysis of case-control studies [21
] included only those cysts identified as a cause of infertility, and reported an overall increased risk of ovarian cancer (OR=1.9, 95% CI 0.8–4.5), with a nearly 4-fold elevation in risk of serous borderline tumors. In contrast, ORs of 0.7 and 1.3, both with wide confidence bounds, were reported for the association of ovarian cysts and invasive epithelial ovarian cancer in other case-control studies [22
]. In one of these [23
], subanalyses examined the relative frequency of a benign ovarian cyst identified during a histopathologic review of the malignant lesion. Mucinous [relative risk (RR) =12.9, 95% CI 4.2–39.7] and, to a lesser extent, endometrioid tumors were categorized as more often arising from a benign cyst than were serous tumors. In a Swedish record linkage study, Borgfeldt [24
] reported ORs for ovarian cancer of 0.9 (95% CI 0.7–1.1) and 1.2 (95% CI 0.8–1.9) among women with a hospital discharge diagnosis of a benign ovarian cyst/tumor and functional cysts, respectively, with no information on histology of the ovarian cancer; in contrast to the current study, women who underwent ovarian surgery were at a substantially increased risk (OR=8.8, 95% CI 5.2–14.8), again with no information on the histology of the malignant tumor. Crayford et al [25
] assessed the impact of removal of persistent benign cysts in a cohort of asymptomatic volunteers whose cysts were identified via ultrasonography, and reported no decrease in the proportion of expected deaths from ovarian cancer relative to other cancers during follow-up. However, this comparison did not allow examination of risk of ovarian cancer (overall or by histologic type) independently associated with such asymptomatic cysts and/or their removal; also, a large proportion of women who underwent surgical investigation for a cyst had both ovaries removed.
Prior studies assessing the impact of endometriosis are rather consistent in their finding of a moderately increased risk of invasive epithelial ovarian cancer, with no increase in risk of borderline tumors [4
]; however, the influence of ovarian surgery on cancer risk among women with endometriosis has not been assessed. Studies addressing risk in histologic subgroups also have observed consistently that the association is attributable to an increased risk of endometrioid and clear cell tumors [26
], and histopathologic examinations of ovarian tumors provide evidence of a tendency for endometriosis to be present at a higher frequency among women with endometrioid and clear cell ovarian tumors than among women with other ovarian tumor subtypes [28
]. Results of both Swedish [29
] and Japanese [27
] cohort studies suggest that the association of ovarian cancer risk and endometriosis may be greatest among women whose endometriosis is present on the ovaries. The results of the current study are thus consistent with prior studies and further raise the possibility (albeit with limited precision) that ovarian surgery in women with a diagnosis of endometriosis may lower subsequent cancer risk, relative to women with endometriosis who do not undergo ovarian surgery. Because, in the current study, we have no data regarding the location (i.e., ovarian or other) of endometriosis among women with this diagnosis, our results may 1) underestimate the association of ovarian cancer risk associated with the occurrence of ovarian endometriosis and 2) underestimate the extent to which ovarian surgery may lower risk among women with ovarian endometriosis.
We observed a moderate increase in risk of developing a borderline mucinous tumor among women with a prior ovarian cyst, consistent with the possibility that some benign ovarian abnormalities may serve as precursors of this tumor type. However, risk of this subtype was, if anything, further increased among women who reported ovarian surgery after their cyst diagnosis. These results contrast with a recent hypothesis that removal of benign precursor lesions may preferentially reduce risk of mucinous tumors [1
]. It is possible that benign ovarian abnormalities may serve as risk markers, rather than precursors, of subsequent borderline mucinous tumors; or that benign precursors of borderline mucinous tumors, if they exist, may tend to occur in multiple locations or to reoccur.
In contrast, history of a prior ovarian cyst was associated with only a modestly increased risk of invasive ovarian cancer, and among women with cysts and subsequent surgery, risk was reduced; this pattern was most evident for serous invasive tumors. While this common subtype of ovarian cancer has often been stated to arise “de novo”, indicating a rapid progression in the absence of an identifiable precursor, evidence is accumulating to support the existence of a precursor (possibly only detectable at the molecular level), either in the ovarian [8
] or distal tubal [9
] epithelium. Conceivably, such precursors may occur more commonly among women with some types of benign ovarian abnormalities, and may further be removed during the course of subsequent ovarian surgery. While our results require replication, they suggest differences in the relation of ovarian cysts and endometriosis with risk of specific subtypes of epithelial ovarian cancer, as well as the possibility that ovarian surgery after diagnosis of these conditions may reduce risk of invasive disease.