Benign Epithelial Ovarian Tumours—cancer Precursors or Markers for Ovarian Cancer Risk?
- Cite this article as:
- Jordan, S., Green, A. & Webb, P. Cancer Causes Control (2006) 17: 623. doi:10.1007/s10552-005-0370-y
- 95 Views
The natural history of the development of epithelial ovarian cancer remains obscure and no effective screening test exists. In several human malignancies progression from benign to invasive tumour occurs, but this sequence has not been established for epithelial ovarian cancer. We have reviewed epidemiological, histopathological and molecular studies of benign epithelial ovarian tumours to assess the evidence for and against such a progression in ovarian cancer. These data suggest that a diagnosis of a benign ovarian cyst or tumour is associated with an increased risk of ovarian cancer later in life. Current evidence also suggests that benign serous tumours can progress to low-grade serous cancer and that benign mucinous tumours can progress to mucinous cancer. The more common high-grade serous ovarian cancers are likely to arise de novo.
KeywordsEpithelial ovarian cancerBenign ovarian tumoursMalignant progression
Invasive epithelial ovarian cancer (EOC) is an aggressive malignancy that has no effective screening test, and is usually diagnosed late in the course of the disease when it is frequently fatal. One appealing proposition is that epithelial ovarian neoplasms, classified as benign, borderline (of low malignant potential, LMP), or malignant according to the degree of atypia, type of cellular proliferation, and presence/absence of invasion, represent stages on a continuum to invasive disease, as observed in several other human malignancies. Whether identifiable benign ovarian tumours can be precursors to malignancy or whether the neoplastic pathway diverges to produce either benign, borderline, or invasive epithelial ovarian tumours as distinct entities, is unknown.
The progression model is further complicated when the different histological subtypes of epithelial ovarian neoplasms are considered. Etiological differences appear to exist between the histological subtypes of invasive tumours [1, 2]. It may be, therefore, that the pathway to the development of invasive disease also varies according to histological subtype such that some subtypes progress from benign tumours while others develop de novo from the ovarian epithelium. In this review we have examined epidemiological, histopathological and molecular studies for evidence for and against malignant progression from benign epithelial ovarian tumours to epithelial ovarian cancer, specifically considering evidence by histological subtype.
Methods and search strategy
PubMed (1966–2004) was searched using the main search phrase “benign ovarian tumour” combined with other terms including “epidemiology”, “risk factors”, “incidence”, “prevalence”, “screening”, “histopathology”, “molecular”, “ovarian cancer” and other search terms appropriate to each section. Variations in key search phrases and their spelling were also included and abstracts from salient non-English language articles were perused. Articles relating to non-epithelial ovarian tumours were subsequently disregarded. Finally, lists of references cited by relevant research and review articles were examined to ensure that other relevant articles were included.
As we were particularly interested in the relationships within histological subtypes of ovarian tumours, the accuracy of histological diagnoses of the tumours included the individual studies is particularly important. Although the quality of pathological review is not easily established, where relevant all referenced studies included mention of pathology confirmation or review of included study materials.
Incidence and relative frequency of epithelial ovarian tumours
The annual age-standardised incidence of ovarian cancer is approximately 10–12 per 100,000 women in North America and Europe [3, 4]. However, the incidence of benign epithelial ovarian tumours, or adenomas, is difficult to determine because their diagnosis is not routinely or systematically registered. Estimates from case-series suggest that surgery for benign epithelial ovarian tumours occurs about 1.3 times as frequently as that for malignant (including borderline) epithelial tumours [5, 6]. The actual incidence is likely to be higher than suggested by surgical rates, however, as many benign tumours remain asymptomatic and thus untreated. For example, in a large UK screening program for self-referred asymptomatic women, benign epithelial ovarian tumours were detected in 1.7% of women, almost 20 times more frequently than epithelial ovarian cancers (0.09%) . Although this estimate of relative frequency may be distorted because women with benign tumours are more likely to be asymptomatic than women with cancers, the finding does imply that benign tumours frequently go undetected.
Two large randomised screening trials for ovarian cancer are ongoing in the United States and the United Kingdom, both using transvaginal ultrasound (TVU) examinations as part of their screening strategies . Results of these studies will provide more accurate population estimates of the prevalence of benign ovarian tumours and may provide further valuable information about the relationship between benign ovarian tumours and their invasive counterparts.
If stepwise tumour progression occurs, it would be expected that most tumours of a particular subtype would be benign, with fewer being borderline and fewer again being invasive. This is the pattern seen for mucinous ovarian tumours (Fig. 1) but invasive serous tumours occur almost as frequently as benign serous tumours and more than twice as frequently as borderline tumours suggesting that stepwise progression is less likely for this, the more common subtype.
Risk factors for epithelial ovarian tumours
If benign epithelial ovarian tumours are precursors to borderline and invasive tumours, it might be expected that the three tumour types would have common risk factors. Consistent evidence suggests that an increased risk of invasive EOC is associated with a family history of ovarian cancer, while decreased risk is associated with increasing parity, oral contraceptive (OC) use, breast-feeding, tubal ligation and hysterectomy [13–15]. Risk factors for borderline tumours appear to be similar to those for invasive tumours [2, 16, 17], but differences have been reported for different histological subtypes of ovarian cancer, suggesting particularly that the mucinous cancers differ etiologically from the serous and endometrioid cancers [2, 18–20]. Serous and endometrioid cancers appear to have a stronger association with family history and hormonal or reproductive factors while mucinous cancers may have a stronger relationship with smoking .
In contrast to the extensive research conducted to determine the risk factors for epithelial ovarian cancer, benign tumours have rarely been studied and consequently little is known about their etiology. The three earliest studies of benign tumours were small, hospital-based and included heterogeneous groups of tumours [21–23] so would be neither representative of all benign epithelial tumours, nor reflect subgroup differences. More recently, however, a larger US case-control study using controls drawn from the general population reported data on risk factors for benign tumours, including some results by histological subtype, which can be compared to risk factor data for ovarian cancer [24–26].
For a combined group of benign tumours (including teratomas, serous adenomas, endometriomas, and mucinous adenomas), parous women were found to have a 50% lower risk of benign tumours than nulliparous women . In addition, more than 5 years of OC use was found to be associated with a statistically significant 60% decrease in risk and a significant trend of decreasing risk with increasing duration of OC use was also reported . The inverse association with more than 5 years OC use was similar for serous and mucinous tumours (odds ratio (OR) 0.37 and 0.47 respectively), but was statistically significant only for endometriomas (OR 0.15) although these analyses were limited by small sample sizes. Subtype analyses were not reported for pregnancy. Studies of ovarian cancers have found a more marked protective effect of the OC on serous and endometrioid cancers compared to mucinous cancers [2, 18, 19].
Evidence from two ultrasound screening studies suggests that family history of ovarian cancer may be an important risk factor for benign ovarian tumours as well as cancer [27, 28]. The aim of both studies was to assess the effectiveness of ultrasound screening in detecting early ovarian cancer in asymptomatic women. In one, 776 British women with at least one first or second-degree relative with ovarian cancer were screened ; and in the second 5,479 healthy volunteers from the British general population were screened . The rate of detection of benign epithelial tumours in the women with a family history of ovarian cancer was almost four times the rate in women recruited from the general population (2.6% vs. 0.7%) suggesting that a genetic predisposition to ovarian cancer also predisposes to benign ovarian tumours . It must be noted, however, that the women screened in the second study were healthy volunteers recruited via advertising, and were therefore probably more health-conscious and perhaps at different risk of benign tumours compared to the general population; while those recruited into the screening study of high risk women were selected on the basis of their family history and thus had different motivations for participating in the study. It appears, however, that any hereditary disposition to benign (and borderline) ovarian tumours may be restricted to families that do not carry BRCA mutations. While between 2% and 11% of women with invasive cancer carry BRCA mutations depending on the population studied , mutations have rarely been reported among women with borderline or benign tumours [30, 31]. This is consistent with the observation that BRCA-associated ovarian tumours are, almost invariably, high-grade invasive serous cancers. It is also supported by results from a recent hospital-based study, which found that women from families with a hereditary predisposition to ovarian cancer unassociated with BRCA mutations were more likely to have benign cystadenomas (serous and mucinous) than women from BRCA families, or women without significant family histories of cancer (8% vs. 0%) .
While a relationship between benign epithelial ovarian tumours and family history and reproductive factors appears likely, and the scant data hint at type-specific differences in risk factors, the paucity of data and methodological concerns mean that firm conclusions cannot be drawn. In addition, there are a number of other factors which affect EOC risk (such as tubal ligation, hysterectomy, breastfeeding, talc use, and hormone replacement therapy) that have not been reported in relation to risk of benign tumours. Further comprehensive comparisons between etiological factors for benign ovarian tumours and invasive tumours, therefore, cannot be made. To address the deficiencies in our current knowledge, large population-based studies of benign ovarian tumours of sufficient size to detect differences between the histological subtypes are required.
Benign tumours as risk factors for ovarian cancer
There is some evidence that a diagnosis of a benign ovarian tumour or cyst increases a woman’s risk of ovarian cancer in later life. A pooled analysis of case-control studies found that women who reported a previous diagnosis of ovarian cysts had a non-significant increased risk of ovarian cancer overall (OR 1.9, 95% CI 0.8–4.5), and a significantly increased risk of serous borderline tumours (OR 3.8, 95% CI 2.2–6.5) . Studies from both Italy  and China  have also reported an increase in risk of ovarian cancer associated with a history of benign cysts (OR 1.3, 95% CI 0.9–1.8 and OR 12.0, 95% CI 2.5–57.7 respectively), while an earlier hospital-based case-control study found a non-significant decrease in risk (OR 0.7, 95% CI 0.4–1.1) . In all these studies the investigators were relying on the recall of participants and the type of cyst was not specified, so it is likely that the term “cyst” encompassed a variety of histological diagnoses from simple follicular cysts, to epithelial ovarian tumours. Recall bias may also have influenced the reported findings, and thus the relevance of these results is uncertain. In addition, a recent record linkage study found that women with a discharge diagnosis of an ovarian cyst (a heterogeneous group including benign tumours, unknown ovarian tumours and unspecified cysts in the ovary) were not at increased risk of developing ovarian cancer. However, those women who had received surgical treatment for their cysts, and thus were more likely to have had persistent benign epithelial tumours, were at markedly increased risk of subsequently developing ovarian cancer (OR 8.8, 95% CI 5.2–14.8) compared to matched controls . Apart from the problem of separating different cyst/tumours types, the study design allowed for only limited adjustment for potentially important confounding factors. Despite this, the fact that four of five diverse studies have now reported an increased risk of ovarian cancer among women who have had benign tumours/cysts suggests that women susceptible to benign tumours may also be susceptible to ovarian cancer. Analogous to benign cutaneous naevi and benign breast disease being markers of increased risk of melanoma [37, 38] and breast cancer  respectively, a diagnosis of a benign epithelial ovarian tumour may be a marker for increased risk of ovarian cancer in the future. It is thus important that this question be resolved by studying future cancer incidence in women with histologically diagnosed benign ovarian lesions to determine whether this risk is real, and if so, to what histological subtypes it applies.
Ultrasound evidence for progression
If benign tumours represent a pre-malignant condition, it would be expected that screening women for them and subsequently removing them should decrease the future incidence of ovarian cancer in those women. This theory was tested by following the British population-based ultrasound-screening cohort discussed previously . In all, 88 of the 5,479 screened women had benign epithelial tumours removed, often involving bilateral salpingo-oophorectomy. After an average of 15.5 years follow-up, the standardised mortality ratio (SMR) for ovarian cancer in the cohort was lower than expected compared to unscreened women from the British population but the difference was not statistically significant (SMR 90%, 95% CI 52%–127%). The results therefore suggest that the majority of ovarian cancer cases do not develop via tumour progression.
More recently, the clinical charts of 35 Japanese women who had had a TVU between 2 and 12 months preceding their diagnosis with invasive ovarian carcinoma were reviewed . Of these women, 19 had been followed up with serial ultrasounds for apparently benign cysts. The remaining 16 women had no ovarian abnormalities on TVU in the 12 months preceding diagnosis. Among the latter group, nine (56%) were subsequently diagnosed with serous ovarian cancer, four (25%) with mucinous cancer, and three (19%) with endometrioid cancer. None of the 16 had clear cell cancers. In contrast, the subtype distribution in the group with prior TVU abnormalities was two (11%) serous, two (11%) mucinous, seven (37%) endometrioid and eight (42%) clear cell cancers. On histopathological review, the majority (71%) of the non-serous tumours were found to contain benign and borderline portions as well as invasive cancer, but only three of the 11 serous tumours, including the two with previous TVU abnormalities, demonstrated this morphological heterogeneity. The authors concluded that these results were consistent with two pathways of carcinogenesis: de novo carcinogenesis particularly for serous cancers, and progression from adenoma to carcinoma for non-serous tumours. Without histological diagnosis of the ultrasound lesions it was impossible to be sure that the abnormalities seen prior to diagnosis were not, in fact, early malignancies rather than benign tumours. The normal-appearing ultrasounds recorded in the 12 months preceding diagnosis of women who presented with late-stage serous carcinomas do, however, offer some support for de novo carcinogenesis for the majority of serous type carcinomas.
Histopathological evidence for progression
Histopathological studies of epithelial ovarian tumours have provided the strongest evidence for tumour progression. The presence of areas of benign and borderline tumour tissue within specimens of invasive ovarian cancer, and evidence of transition between these areas suggest benign to malignant transformation occurs [41, 42]. Large population-based studies are required to precisely quantify morphological heterogeneity among the histological subtypes of ovarian cancer. However, several case-series of 111 , 189  and 255  tumour samples have been undertaken to examine this issue and all concluded that invasive mucinous tumours are much more likely than their serous counterparts to contain contiguous areas of benign tumour (74–90% vs. 15–56% respectively) [33, 43, 44]. An intermediate percentage of endometrioid (22–46%) and clear cell cancers (39%) appear to contain areas of benign tumour. It has also been observed that early-stage cancers contain areas of benign tumour tissue more frequently than later-stage cancers [33, 43] and that low-grade cancers contain areas of benign tumour tissue more frequently than high-grade cancers . As most serous cancers present at a later stage than other histological subtypes, it is possible that the malignant tumour obliterates any remnants of benign tissue, thereby accounting for the infrequent finding of benign tumour within invasive serous lesions.
Alternatively it has been proposed that low-grade and high-grade serous cancers represent two distinct entities: one an indolent cancer progressing slowly through stages, and the other an aggressive cancer with rapidly progressive behaviour . A recent study evaluating a new grading system for ovarian serous cancers found that 30 of 50 (60%) low-grade serous cancers but only 1 of 50 (2%) high-grade cancers were associated with borderline tumour . Low-grade tumours may progress from benign and borderline tumours, while the much more common high-grade tumours arise directly from the ovarian surface epithelium and spread rapidly throughout the pelvis and abdominal cavity. This latter pathway is further supported by the report of a case-series of 14 early epithelial ovarian cancers identified in specimens not suspected to contain carcinoma on gross examination pre- or post-surgery . Of these, 10 (71%) were serous tumours, one endometrioid, one clear cell, and two were undifferentiated. None were low-grade. In all 14 cases, dysplastic changes were noted on the ovarian surface epithelium or its inclusions but the cancer tissue was not related to pre-existing benign epithelial tumours.
The results of histopathological studies therefore support the hypothesis that most serous ovarian cancers, which are usually high-grade and late-stage at diagnosis, do not arise from benign tumours, but arise de novo from the surface epithelium of the ovary or its inclusions, however, the majority of mucinous cancers arise in pre-existing mucinous cystadenomas .
Although the above histopathology data appear to provide good evidence that at least some ovarian cystadenomas undergo malignant transformation, another possible explanation for the morphological heterogeneity of some ovarian tumours is that benign-appearing portions actually represent maturation, or areas of better differentiation, of malignant cells [44, 48]. Histopathological techniques are unable to distinguish the direction of cellular change in these situations and thus molecular studies are required to further clarify the issue.
Molecular evidence for progression
Molecular studies of morphologically heterogeneous ovarian tumours containing benign, borderline and malignant tissue
Zheng et al. 
Two MH tumours. 16 solitary benign tumours
LOH at 11p15.5 loci
Solitary benign and borderline tumours
Supports either: maturation from malignant tissue or benign tumour adjacent to malignant is not typical cystadenoma
Heterogeneous malignant tumours
LOH in benign, borderline and malignant tissue
Zheng et al. 
Six MH tumours. 11 solitary benign tumours
Supports either: maturation from malignant tissue or benign tumour adjacent to malignant is not typical cystadenoma
p53 mutations only in malignant tissue (50%)
p53 mutation in benign, borderline and malignant components
Kuwata et al. 
22 MH tumours. Solitary tumours: benign 27, borderline nine, invasive ten
p53 over expression and proliferative activity (PA)
At least some tumours progress from benign to malignant
Over-expression p53: malignant>borderline>benign
Increased PA: malignant>borderline>benign
Similar results, no over-expressed p53 and lower PA in benign tissue
Wolf et al. 
Ten MH serous tumours
Tumour aneuploidy (using FISH)
6/10 demonstrated aneuploidy
Findings imply progression
–Aneuploidy in benign tissue also seen in borderline/malignant
–Aneuploidy in borderline also in malignant
–Malignant relatively more chromosome gains than benign
Cuatrecasas et al. 
Ten borderline, five invasive MH mucinous tumours
Borderline 7/10—K-ras mutation benign & borderline
Suggest K-ras is an early event in mucinous tumourigenesis
Borderline 2/10—K-ras only in borderline tissue
Invasive 5/5—same mutation throughout
Mandai et al. 
One borderline, eight invasive MH mucinous tumours
7/8—K-ras mutation in benign, borderline, and invasive
Progression may occur in some tumours but others arise de novo and undergo maturation
1/8—K-ras in invasive/borderline not benign.
1/1 borderline—K-ras in borderline & some, but not, all benign portions
Garret et al. 
16 borderline, four invasive MH mucinous tumours
8/16 borderline had K-ras mutations
K-ras mutation may be an early genetic change in mucinous tumours; malignant and borderline tumours may arise clonally from benign tumours, i.e. progression
2/4 malignant had K-ras mutation
Mutation in borderline/malignant tissue frequently found in benign tissue but never in adjacent stromal tissue
One group examined six morphologically heterogeneous tumours with p53 mutations  and two with LOH on chromosome 11 at 11p15.5 . While both studies found the molecular change in both the morphologically benign portions of the tumours as well as the malignant-appearing portions, 16 solitary benign cystadenomas showed neither p53 mutations nor LOH on chromosome 11. This suggests that benign portions of heterogeneous tumours are not typical of benign cystadenomas. An alternative explanation is that they represent areas of better differentiated malignant tissue [49, 50] however this is not supported by a third study which observed abnormal p53 expression in invasive portions of 13 of 22 heterogeneous ovarian tumours but not in benign portions . Apart from the very small number of samples used, another drawback of the above studies is that they included a mix of different histological subtypes of epithelial ovarian cancer. As mucinous ovarian tumours probably differ from the other histological subtypes in etiology and developmental pathway, analysis of morphologically heterogeneous tumours according to histological subtype would be more informative.
Three molecular studies of morphologically heterogeneous mucinous ovarian tumours focussing on the K-ras oncogene [53–55] found that the majority contained the same K-ras mutation in benign, borderline and malignant portions of the tumour, although a few tumours had the mutation in borderline and malignant portions but not benign. These results suggest that if progression does occur, then K-ras mutation is an early event in tumourigenesis, or may be implicated in differentiation to the mucinous subtype [53, 55].
One study used fluorescence in situ hybridisation to look for numerical chromosomal abnormalities in ten serous epithelial ovarian carcinomas containing areas of benign or borderline tumour immediately adjacent to the malignant tumour . Six of ten tumours demonstrated aneuploidy, and within these tumours the frankly malignant tissue had relatively more chromosomal gains than the benign tissue. Abnormalities found in the benign tissue were, however, also seen in the borderline and malignant tissue. Although these results support the possibility of tumour progression, the grade of the tumours was not reported and, given the rarity of heterogeneous high-grade serous tumours, it is possible that these results are applicable to low-grade serous tumours only.
Low versus high-grade serous tumours
The following studies were not of morphologically heterogeneous tumours but provide further evidence for two pathways of serous carcinogenesis. K-ras mutations have been observed in 30–50% of serous borderline tumours, 30–50% of low-grade serous carcinomas, but rarely (0–12%) in high-grade serous cancers [45, 58, 59]. The BRAF gene, which along with K-ras is a member of the ERK/MAP kinase signalling pathway, is also frequently mutated in serous borderline tumours (30–50%) and low-grade serous carcinomas (35%) but rarely if ever in high-grade serous carcinomas [45, 59]. Although probably under-powered, one study of 30 serous benign cystadenomas found no BRAF or K-ras mutations and postulated that such mutations in these tumours may lead to the development of borderline tumours . The existing data therefore do not directly implicate benign serous tumours in the progression to low-grade serous tumours, but do not preclude the possibility. The discrepant molecular findings in high and low-grade serous tumours do, however, suggest that progression from borderline to low-grade serous cancer can occur, while the much more commonly encountered high-grade cancers appear to develop de novo [58, 61, 62].
Collectively, the studies described above neither provide consistent evidence of tumour progression, nor exclude this possibility. Most of the studies included only small numbers of samples and thus the power to detect low frequency molecular changes was limited. In addition, population-based sampling of tumours was not undertaken by any of the studies, thus the generalisability of the results could be questioned. Methodological issues aside, the validity of the conclusions drawn also depends upon the sequence of genetic alterations in carcinogenesis. Fearon and Vogelstein  in their model of colorectal tumourigenesis postulated that although genetic alterations often occur according to a preferred sequence, it is the total accumulation of changes that determines the biological behaviour of tumour rather than the order in which they occur. Thus future studies using larger, more representative samples of tumours that examine multiple molecular changes in the same sample, such as those using expression microarrays, are likely to be informative.
Current molecular evidence does, however, provide some support for the concept that more than one pathway of carcinogenesis exists for epithelial ovarian cancer, and as molecular techniques are further developed the role of benign epithelial tumours in carcinogenesis should be further clarified.
Benign and borderline epithelial ovarian tumours are both significant causes of morbidity in their own right, but clarification of their relationships with ovarian cancer may be the key to why some women develop this frequently fatal disease.
Dr Penelope Webb is supported by a Queensland Cancer Fund Senior Research Fellowship.
Dr Susan Jordan is supported by an Australian Postgraduate Award (University of Queensland).