Site of origin of epithelial ovarian cancer: the endometriosis connection
Correspondence: Dr P. Vercellini, Clinica Ostetrica e Ginecologica ‘Luigi Mangiagalli’, Via Commenda 12, 20122 Milano, Italy.
To investigate the left- and right-sided distribution of ovarian malignant surface epithelial tumours, data were collected on 209 women undergoing first-line surgery for Stage I and II disease. Considering the unilateral cancers, the observed proportion of left-sided lesions was 35/54 (65%) in the endometrioid, 20/45 (44%) in the serous, 19/35 (54%) in the clear cell, 13/29 (45%) in the mucinous, 2/8 (25%) in the mixed, and 2/5 (40%) in the undifferentiated histological type group. The proportion of left-sided unilateral endometrioid cancers was significantly different from the expected 50% (χ21, 4.74, P= 0.03) and very similar to that previously observed for benign endometriotic cysts, constituting further evidence in favour of a possible development of endometrioid cancers from the latter lesions.
Endometriosis is reported to be significantly more frequent in women with endometrioid, than in those with serous and mucinous ovarian, cancers1,2. However, since a direct transition from benign to malignant endometrial-type epithelium has been demonstrated infrequently, most endometrioid carcinomas are thought to derive from the modified surface epithelium of the ovaries without the interposition of pre-existing endometriosis3.
If metaplasia and neoplastic transformation of the coelomic mesothelium covering the gonadal surface is the common pathogenesis of endometrioid, serous and mucinous carcinomas, the lateral distribution of these epithelial ovarian cancers should be similar3. In fact, any carcinogenetic event leading to malignant change should have an equal chance of occurring on either side independently of histological type.
The proportion of left-sided ovarian endometriomas has been observed to be significantly higher than the expected 50%. This finding has been discussed in terms of the anatomical differences between the left and right hemipelvis (specifically, with respect to the presence of the sigmoid colon), which may favour implantation of regurgitated endometrial cells4. If endometrioid tumours develop from endometriotic cysts, their left- and right-sided distribution would also be expected to be unbalanced, whereas no such asymmetry should be found if they have the same origin as serous and mucinous carcinomas.
Given this background, we set out to investigate the left- and right-sided distribution of ovarian malignant surface epithelial tumours in a large series of women undergoing surgery for Stage I and II disease. Early forms were selected because only unilateral lesions are of interest to test the study hypothesis.
Clinical records were retrieved of consecutive women with Stage I–II epithelial ovarian cancers undergoing first-line surgical treatment between 1992 and 1999 at the First Department of Obstetrics and Gynaecology of the University of Milan and between 1996 and 1999 at the Division of Gynaecologic Oncology of the National Institute for Cancer Treatment in Milan. Data were collected on age, parity, menopausal status at surgery, stage of disease, side of ovarian tumours and presence or absence of associated endometriosis. Information on the histological type of ovarian cancer was obtained from pathological records. The side of origin of the tumour was computed for unilateral cases only, as no attempt was made to determine right or left dominance (i.e. the side with the larger lesion, in bilateral tumours).
Two authors (G.S. and G.S.), who did not know the study hypothesis, abstracted data independently. Discrepancies between reviewers were resolved by a third, unblinded evaluator (G.B.).
The frequency of left- and right-sided ovarian cancers was analysed with the χ2 test to compare observed and expected events. The 95% confidence interval of the proportion of left tumours was computed using the normal approximation.
In the study period 209 women with a mean age of 53 years were identified, 178 with Stage I and 31 with Stage II disease. One hundred and seven of these (51%) were premenopausal at diagnosis and 50 (24%) were nulliparous. Malignant endometrioid tumour was the most frequent diagnosis (n= 66 women [32%]), followed by serous (n= 61 [29%]), clear cell (n= 35 [17%]), mucinous (n= 30 [14%]), mixed (n= 10 [5%]), and undifferentiated (n= 7 [3%)] histological types. Agreement between reviewers on side of unilateral lesions was complete, whereas discrepancies occurred in determining substaging in 11 cases and menopausal status in 19, with corresponding kappa values of 0.87 and 0.82.
A total of 85 tumours (41%) were on the right-hand side, 91 (44%) on the left-hand side, and 33 (16%) were bilateral. In the women with unilateral cancers, the observed proportion of left-sided lesions was 35/54 (65%; 95% CI 51% to 77%; χ21 4.74; P= 0.03) in the endometrioid, 20/45 (44%) in the serous, 19/35 (54%) in the clear cell, 13/29 (45%) in the mucinous, 2/8 (25%) in the mixed and 2/5 (40%) in the undifferentiated histological type group. The inclusion of bilateral lesions in the endometrioid group gave a total of 47/78 (60%; 95% CI 49% to 71%; χ21 3.28; P= 0.07) left-sided and 31/78 right-sided tumours. The percentages of left and right unilateral endometrioid carcinomas were consistent in strata of age, parity, menopausal status and disease stage.
Endometriosis was found in 22/209 cases (10%), being more frequently associated with endometrioid (20%) and clear cell (14%) adenocarcinomas than with other histological types (Table 1). Three patients in the endometriod tumour group had concomitant endometrial cancer.
Table 1. Clinical characteristics and lateral distribution of Stage I and II epithelial ovarian cancer in 209 women studied. Values are given as mean (SD) or n [%].
|Age (years)||53 (10)||53 (13)||53 (12)||51(14)||54 (13)||49(13)|
|Parous||50 ||47 ||28 ||22 ||7 ||5 |
|Menopausal||31 ||33 ||17 ||13 ||5 ||3 |
|Stage|| || || || || || |
|I||51 ||50 ||34 ||28 ||9 ||6 |
|II||15 ||11 ||1||2 ||1 ||1 |
|Tumour|| || || || || || |
|Left-sided||35 ||20 ||19 ||13 ||2 ||2 |
|Right-sided||19 ||25 ||16 ||16 ||6 ||3 |
|Bilateral||12 ||16 ||0||1 ||2 ||2 |
|Associated endometriosis||13 ||2 ||5 ||1 ||1 ||0|
The oncogenic potential of endometriosis is controversial1–3. A large, nation-based, cohort study demonstrated an almost fourfold increase in the risk of ovarian cancer among women with a long history of ovarian endometriosis5. More specifically, an association between endometriosis and endometrioid ovarian carcinoma has been reported repeatedly1–3.
According to the results of our study, malignant endometrioid ovarian tumours arise more frequently on the left than the right gonad. The proportion of left-sided unilateral endometrioid cancers is very similar to that previously observed by us for benign endometriotic cysts (35/54, 65% and 255/403, 63%, respectively)4. Moreover, there was no important difference in the proportion of left- and right-sided lesions in the other histological types considered. In our view, this suggests that the pathogenesis of endometrioid cancer is distinct from that of other epithelial tumours of the ovary. Furthermore, it is tempting to speculate that a continuum may exist between ovarian endometriomas and malignant endometrioid tumours.
In the present study, as well as in previous studies1,3, endometriosis was found to be associated with both the endometrioid and the clear cell histological type, but we observed asymmetry in the lateral distribution of cancer lesions only in the former group. We have no plausible explanation for this apparent discrepancy, but consider that clear cell carcinomas represent histological variants not only of endometrioid but also of serous carcinomas3, which were distributed symmetrically. More importantly, endometrioid adenocarcinomas constitute 69% and clear cell carcinomas only 13% of malignancies arising in endometriosis6.
The serous histological type was relatively underrepresented in our series, whereas the endometrioid one was the most frequent. However, other authors observed percentages as low as 13% for serous carcinomas when only early stages were considered2, and the proportion of endometrioid tumours has been reported to be as high as 30% or more2,3. Moreover, a predominant serous component was observed in the 10 cancers classified as ‘mixed histological type’ by our pathologists.
We analysed a relatively large series of patients with early stage, epithelial ovarian cancers. Information was abstracted from hospital records by two independent investigators. All diagnoses were based on the pathological report and none of the cases selected according to the study protocol was excluded. Moreover, the surgeons who recorded the clinical data, the pathologists who performed the histological examination and two of the three reviewers were unaware of the hypothesis of the study. To determine lateral distribution we considered only the 176 unilateral lesions without attempts to identify right and left dominance in bilateral tumours. In our opinion, the assumption that the side with the larger tumour is the side of origin is merely speculative.
The side of origin of epithelial ovarian cancer has been a matter of debate, with a right-sided predominance observed in one series and a substantial lateral symmetry in others. However, distribution according to different histological types has never been reported. The present findings are further evidence suggesting that endometrioid ovarian tumours may originate from gonadal endometriosis, but a larger, multicentre study is needed to confirm this hypothesis.