Endometriosis: Pathogenesis and Treatment 2014 Ed.

22. The Association of Endometriosis with Ovarian Cancer: A Critical Review of Epidemiological Data

Sun-Wei Guo1, 2  


Shanghai OB/GYN Hospital, Fudan University, 419 Fangxie Road, Shanghai, 200011, China


Department of Biochemistry, Shanghai College of Medicine, Fudan University, Shanghai, China

Sun-Wei Guo

Email: hoxa10@gmail.com


Although endometriosis is well recognized as a benign gynecologic condition, its association with ovarian cancer has frequently been reported. Many review papers on this topic have been published, yet there seems to be no consensus as to whether or not endometriosis is a precursor of ovarian cancer and whether or not any actions should be taken based on our current knowledge on the endometriosis-ovarian cancer association. In this chapter, I shall critically review epidemiological and clinical prevalence data for and against the link, point out the challenges in proving the causal link, and, in the end, sketch some ways to solve this problem. I shall provide a tutorial review of epidemiological studies, measures of effect size, and confounding and point out the connection between the prevalence data and the relative risk estimate. Funnel plots are used to examine the asymmetry of risk estimates, in which the risk estimates such as OR are plotted on a logarithmic scale against the inverse of their corresponding measures of precision. Areas in need of further research are also outlined.


AssociationCase-control studiesEndometriosisEpidemiologyOvarian cancer

22.1 Introduction

Although endometriosis is well recognized as a benign gynecologic condition, its association with ovarian cancer has often been reported and mentioned in literature. A PubMed search with the keyword, endometriosis, yielded a total of 19,621 papers. Using the combination of “endometriosis” and “ovarian cancer,” it turned up 2,544 papers or 13.0 % of the total literature on endometriosis. In contrast, 3,722 (19.0 %), 3,542 (18.0 %), and 515 (2.6 %) papers were on endometriosis and infertility, endometriosis and pain, and endometriosis and inflammation, respectively (accessed June 18, 2013). Many review papers on this topic have been published, yet there is no consensus as to whether or not endometriosis is a precursor of ovarian cancer [1]. While many scientists sit on the fences, some investigators think that the ovarian endometriomas “could be viewed as a neoplastic process,” considering the malignant transformation of endometriosis as rather obvious [2]. Other investigators are more cautious, arguing that, among the nine criteria of causality, proposed by Hill [3], many are still unfulfilled [4].

However, a Lancet Oncology paper, published early last year, seemed to tip the balance of this debate towards the conclusion that endometriosis is a precursor of ovarian cancer. From a pooled analysis with primary data from 13 case-control studies, Pearce et al. reported that women with self-reported endometriosis were associated with a significantly increased risk of clear cell (odds ratio (OR) = 3.05), low-grade serous (OR = 2.11), and endometrioid invasive (OR = 2.04), but not high-grade serous invasive ovarian cancer [5]. The authors of that paper concluded that “we will develop a risk stratification model that combines genetic and epidemiological risk to better stratify women into high-risk, intermediate-risk, and low-risk categories, allowing better individualization of prevention and early detection approaches such as risk-reduction surgery and screening.” The lead author, Dr. Celeste Leigh Pearce, was even more optimistic, stating, in a journal news release, that “This breakthrough could lead to better identification of women at increased risk of ovarian cancer and could provide a basis for increased cancer surveillance of the relevant population, allowing better individualization of prevention and early detection approaches such as risk-reduction surgery and screening” (http://​health.​usnews.​com/​health-news/​news/​articles/​2012/​02/​22/​endometriosis-could-raise-risk-of-3-ovarian-cancers, accessed January 15, 2013). In other words, endometriosis is indeed a precursor of ovarian cancer and, as such, certain measures should be taken accordingly.

Ovarian cancer is by far the most lethal malignancy of the female reproductive system. Over 90 % of ovarian cancers arise from the surface epithelium [6]. Despite advances in radical surgery and chemotherapy, the overall survival has changed very little in the last 30 years [7]. With the advent of molecular biology, a great deal of efforts have been focused on early detection, yet this attempt often brings with roller-coaster experience and consequently it has so far not resulted in any tangible survival benefit to the patients. Faced with such an abject failure, it is important to identify the precursor(s) of ovarian cancer, even for some specific histotypes.

With this in perspective, it is perhaps understandable as why there have been so much attention on the endometriosis-ovarian cancer link. However, how solid is the evidence for this link? Would the evidence, gathered so far, warrant any actionable measures such as that suggested by Dr. Pearce? These are simple yet weighty issues that are worth careful examination.

In this chapter, I shall critically review the evidence in support of the link, point out the challenges in proving the causal link, and, in the end, sketch some ways to solve this problem. Due to space limitation, I shall restrict my attention to epidemiological and clinical data. While there is a growing body of literature documenting shared molecular aberrations between endometriosis and ovarian cancer, suffice to say that, due to the nature of these studies, many such aberrations are indicative of association, only suggestive for a causal link, simply because the temporality of the link is difficult to prove.

22.2 Methods

A systematic and comprehensive search of PubMed was performed for all studies published up to June 18, 2013, using the following combination of search terms of “endometriosis,” “ovarian cancer,” “epidemiology,” and “association.” The studies had to report epigenetic aberrations in endometriosis. The search was limited to publications written in English.

For each retrieved case-control studies, the OR and its 95 % confidence interval (CI) and the standard error (SE) of the log OR were extracted. The choice of log OR was simply due to the fact that, in contrast to the OR, its SE is unaffected by the magnitude of the log OR. For cohort studies, the standardized incidence ratio (SIR), rate ratio (RR), or hazard ratio (HR) and their SEs were extracted.

Funnel plots were used to examine asymmetry, in which the risk estimates such as OR were plotted on a logarithmic scale against the inverse of their corresponding standard errors, a measure of precision [8]. If bias is absent, small studies will have ORs that are widely scattered but symmetric about the OR estimates provided by larger, more precise studies. In this case, the plot would resemble an inverted funnel with the tip pointing roughly towards the true log OR. If publication bias is present, the plot will be asymmetric because some negative studies are not published

All computations were made with R statistics software system version 3.0.9 [9]. The statistical routine rmeta was used.

22.3 The Role of the Funding Source

The sponsor of the study had no role in study design, data collection, data analysis, data interpretation, or writing of this report.

22.4 Evidence from Clinical Series

22.4.1 Early Criteria

As early as 1925, Sampson proposed histopathological criteria for inferring that the malignancy arose from endometriosis, or the causal relationship between endometriosis and malignancy: (1) clear evidence of endometriosis close to the tumor (“proximity”); (2) the carcinoma must be seen to arise in endometriosis and not to be invading it from other sources (“arising from endometriosis”); and (3) presence of tissue resembling endometrial stroma surrounding characteristic glands (“endometrial stroma plus glands”) [10]. Scott later added one more criterion: the demonstration of a histology-proven transition from benign endometriosis to cancer (“transition”) [11].

Clearly, these criteria are all based on histological evidence, which, in turn, are based on tissue samples taken from patients. While the “proximity” and “endometrial stroma plus glands” criteria may be relatively easy to establish, the inference of “arising from endometriosis” and “transition” has, by necessity, to be based on a single snapshot, in contrast to serial observations, of the histological images or morphologic features during a presumably long period of tumorigenesis and, as such, can be challenging to establish. It is no wonder that these criteria, considered to be stringent, are rarely fulfilled [112].

The morphologic data can tell us something, but only to a certain extent. For example, ovarian cancer was once regarded as a single disease since by morphology the tumor seemingly originated from ovary but now a dualistic model of carcinogenesis of ovarian cancer based on distinctive clinicopathologic and molecular genetic features seems to have replaced the older view [713]. Also, just by morphology or histological data, it would be very difficult to find that there are 4 main subtypes of breast cancer caused by different subsets of genetic and epigenetic abnormalities [14]. In addition, since the choice of tissue sections entails certain degree of selection, it may be susceptible to attribution error, especially when the pathologists are inexperienced.

22.4.2 Prevalence Data: An Intimate Connection with the Odds Ratio

About a dozen reports on the prevalence of endometriosis in women with ovarian cancer have been published. If we denote the prevalence of endometriosis (E) as P(E), the prevalence of women with both endometriosis and ovarian cancer (O) (often called endometriosis-associated ovarian cancer or EAOC) as P(EO), and the proportion of E among women with O as the conditional probability P(E|O), then we can denote the relative risk of having Oin women with E vs. women without E ( $$ \overline{E} $$ ) as  $$ \mathrm{RR}= P\Big( O\left| E\Big)\right./ P\Big( O\left|\overline{E}\Big)\right. $$ , that is, the relative risk of having O when a woman has E vs. woman who does not. This is the quantity that most, if not all, epidemiological studies attempt to estimate. Here, the age dependency is often ignored for ease of exposition.

Since both P(O|E) and  $$ P\Big( O\left|\overline{E}\Big)\right. $$  are small (because the lifetime risk of developing O is about 1 % or 0.01 and even lower for some specific histotypes of ovarian cancers),  $$ 1- P\left( O\Big| E\right)\approx 1- P\Big( O\left|\overline{E}\Big)\right.\approx 1, $$

 $$ \mathrm{RR}=\frac{P\Big( O\left| E\Big)\right.}{P\Big( O\left|\overline{E}\Big)\right.}\approx \frac{\frac{P\Big( O\left| E\Big)\right.}{1- P\Big( O\left| E\Big)\right.}}{\frac{P\Big( O\left|\overline{E}\Big)\right.}{1- P\Big( O\left|\overline{E}\Big)\right.}}=\mathrm{OR} $$

That is, when ovarian cancer is rare (indeed it is), the OR obtained from case-control studies is merely an approximation to RR.

By Bayes’ theorem and after some simple algebra, it is easy to see that RR = P(E|O)[1−P(E)]/(P(E)/[1−P(E|O)]). That is, the true RR depends only on the prevalence of endometriosis (which we have a fairly good idea) and the prevalence of endometriosis among women with ovarian cancer (which can be estimated from data, at least in theory). Note that this RR does not depend on the prevalence of O. Hence we can calculate RR even for some specific histotypes of ovarian cancer, such as clear cell or endometrioid ovarian cancers. It is noted that RR can be expressed in a more revealing form, as RR = rO /r, where rO = P(E|O)/[1−P(E|O)], the odds of having endometriosis given that the woman has O, and r = P(E)/[1−P(E)], the odds of having endometriosis in the general population.

22.4.3 Prevalence Data from Published Studies

The prevalence data extracted from 13 studies reporting the prevalence of endometriosis in women with ovarian cancers, along with the estimated RR using either the 10 % prevalence for endometriosis or the 5 % for ovarian endometriomas, are listed in Table 22.1. It can be seen that there are enormous variations in the reported prevalence, and, consequently, the RR estimates also vary greatly from study to study. It should be noted that the reported prevalence of endometriosis in women with ovarian cancer is likely to be an underestimate of P(E|O), since some endometriotic lesions may not have been found by the surgeon who performed the operation, or pathologists who performed histologic examination may not have found any lesion which might still exist. On the other hand, P(E|O) is prone to be overestimated if one is single-mindedly trying to find all cases of ovarian cancer-associated endometriosis while dismissing or discarding cases with ovarian cancer but no endometriosis, consciously or otherwise. Anyhow, it can be seen from Table 22.1 that some RR estimates based on the reported prevalence still deviate greatly from the OR estimates reported from case-control studies, especially for clear cell and endometrioid histotypes. In fact, all, except one, OR estimates are pretty large. It is unclear as to whether some seemingly high prevalence estimates are genuine or a result of ascertainment bias, population idiosyncrasy, or chance events.

Table 22.1

Prevalence of endometriosis in women with ovarian cancers and the estimated relative risk of developing ovarian cancer in women with endometriosis vs. women without



Year of publication


Clear cell and endometrioid


Prevalence (%)

RR (P = 10 %)

RR (P = 5 %)

Prevalence (%)

RR (P = 10 %)

RR (P = 5 %)


Aure et al. [15]


4.2 (35/831)



12.5 (34/271)





Kurman et al. [16]



10.2 (5/49)





Russell [17]


11.3 (46/407)



34.3 (36/105)





Brescia et al. [18]



18.4 (14/76)





Vercellini et al. [19]


11.1 (52/466)



25.0 (38/152)





De La Cuesta et al.



40.0 (16/40)





Jimbo et al. [20]


14.5 (25/172)



35.6 (16/45)





Fukunaga et al. [21]


24.1 (54/224)



49.4 (40/81)





Ogawa et al. [22]


29.1 (37/127)



66.0 (33/50)





Vercellini et al. [23]


43.5 (91/209)



60.7 (54/89)



Only unilateral cancer and left-sided lesions were considered


Oral et al. [24]


7.7 (14/182)



26.3 (5/19)





Dzatic-Smiljkovic et al. [25]


11.0 (23/210)



33.3 (19/57)





Kondi-Pafiti et al. [26]


5.9 (1/17)



0.0 (0/16)




Based on all published prevalence estimates listed in Table 22.1, one can get a pooled estimate of prevalence for all histotypes of ovarian cancer weighted by the sample size, which is 12.9 %, yielding an RR estimate of 1.33 if a prevalence of 10 % is assumed, or 2.81 if a prevalence of 5 % is assumed. For clear cell and endometrioid ovarian cancers, the pooled estimate of prevalence is 28.0 % and the corresponding RR estimates are 3.51 and 7.40, respectively, depending on the endometriosis prevalence.

Note that, since RR = rO /r, where rO = P(E|O)/[1−P(E|O)] and r = P(E)/[1− P(E)], r is a constant when P(E) is assumed to be a fixed number (say, 5 %); hence we can calculate the standard error of log RR, which is equivalent to the standard error of log rO , or log[P(E|O)]log[1−P(E|O)]. By delta method, it is easy to see that the standard error of log RR is inversely proportional to the squared root of the sample size n. Hence, we can plot the log RR listed in Table 22.1 against the squared root of the sample size n reported in the study (Fig. 22.1).


Fig. 22.1

Funnel plot for the 13 log-transformed prevalence-based RRs, using data extracted from Table 22.1 and assuming a 5 % prevalence of endometriosis in the general population. The dashed line represents RR = 1. The alphabet-numeric combinations are the IDs shown in Table 22.1, and each ID represents one study. Note that for study K12, a Bayes estimate for binomial distribution assuming a non-informative prior (uniform distribution) was used, since otherwise the data would yield a RR of 0

It is interesting to see from Fig. 22.1 that when the estimated log RRs were plotted against the squared root of the sample size of the study, the seemingly large RRs seen in Table 22.1, when all placed in the funnel plot in Fig. 22.1, seem to gravitate to a point close to 1, but perhaps slightly greater than 1.

22.5 Epidemiological Evidence

22.5.1 Cohort Studies

In a typical cohort study, two cohorts or two groups of people, one with and one without a particular attribute (in our case, women with or without endometriosis), are identified and followed up longitudinally. The incidence of a particular event—quite often, disease (ovarian cancer, in our case)—in the two groups is evaluated and compared. In this way, whether having the particular attribute (exposure) would increase or decrease the incidence can be investigated.

Due to constraint in time and resources, cohort studies are seldom conducted concurrently or truly prospectively. Instead, many cohort studies are conducted retrospectively. In the latter case, the cohorts are identified and assembled in the past based on archived records. In this case, the occurrence of the event of interest is often retrieved from the records as well. While retrospective cohort studies require much less resources and time, their major disadvantage is their exclusive reliance on available information, sometimes the subjects’ own memory. Consequently, the quality of exposure or disease data can be compromised (e.g., recall bias).

In epidemiology, the outcome is frequently a disease. One common outcome measure in cohort studies is incidence, or the risk of developing certain disease within a specified period of time given no occurrence prior to that time period. It can be expressed either in cumulative incidence (incidence proportion) or incidence rate with a denominator (called incidence density rate or person-time incidence rate). In analytical epidemiology, one measure that is used very frequently is the standardized incidence ratio (SIR), which is a ratio or percentage quantifying the increase or decrease in incidence of a study cohort with reference to the general population.

While the SIR provides a succinct measure of the change from the incidence from the reference population or cohort, it becomes a bit cumbersome when there are some confounding factors that need to be controlled for, especially in retrospective cohort studies. In these circumstances, the rate ratio (RR, also called ratio of incidence densities) or hazard rate (HR) would be more convenient, and their use would render the use of some sophisticated statistical models such as the Cox regression model possible, facilitating elaborate statistical analysis. The RR gives the ratio of the event rate in the cohort of interest (say, women with endometriosis) vs. that in the reference group after adjustment for other known confounders. Both SIR and RR provide a measure of causality (or, rather, association, especially for retrospective cohort studies) between exposures and outcomes.

So far, seven cohort studies, with varying qualities, on endometriosis-ovarian cancer link can be identified (Table 22.2). Some of them were population-based and prospective studies, and others were hospital-based or retrospective cohort studies. These studies yielded either SIR or RR estimates, along with 95 % confidence intervals (CIs) (Table 22.2).

Table 22.2

Published cohort studies reporting ovarian cancer risk in women with endometriosis



Year of publication

Cohort size

Length of follow-up

# of cases



95 % CI


Brinton et al. [27]









Olson et al. [28]









Brinton et al. [29]









Brinton et al. [30]









Melin et al. [31]









Melin et al. [32]









Kobayashi et al. [33]









Aris [34]









Stewart et al. [35]



3.11* (for nulli-parous women)





1.52* (for parous women)



SIR standard incidence ratio, RR rate ratio, SE standard error, CI confidence interval

*“SIR or RR” column are RR estimates

From these SIRs and their CIs, their standard errors (SEs) can be easily calculated. Since all the RRs were calculated based on logistic or Poisson regression models after adjusting for possible confounding factors, the SE of the log-transformed RR can be calculated based on RRs and their CIs.

In evaluating risk estimates—be it SIR, RR, or OR if from case-control studies—from various studies, funnel plots are often employed [36]. In funnel plots, risk estimates, such as RRs or log RRs, are placed on the horizontal axis against some measure of study size or precision, such as their standard errors, on the vertical axis. The funnel plot is so named because of its shape: in the absence of selection biases (such as publication bias and bias in inclusion criteria), true heterogeneity (i.e., size of effect differs according to study size), and data irregularities (such as poor methodological design of small studies, inadequate analysis, and fraud), studies of large or small sample sizes should be more or less symmetrically scattered around the true log RR. Hence the plot should have the shape of a funnel with wide opening on the top (due to sampling variability), with the tip of the funnel pointing to the bottom and centering on the true log RR [36]. The choice of log OR instead of OR is due to the fact that the standard error of RR is related with the odds ratio, while the standard error of log OR is purely a function of sample sizes in different exposure-disease status combinations. The use of log OR also renders ORs that are greater than 1 or less than 1 symmetric about 1 (=0 on the log scale).

When plotting the 9 log-transformed SIRs/RRs against their SEs in a funnel plot (Fig. 22.2), two features can be noted. First, the plot looks like an asymmetric funnel, with its tip gravitating towards somewhere near log SIR or log RR = 0, i.e., SIR = 1 or RR = 1. Since there is no indication of bias in inclusion criteria or heterogeneity, this suggests that there may be a publication bias towards favoring positive studies and higher estimates of odds ratios may well be a chance variation. In addition, the plot seems to suggest that the true SIR or RR is near 1.


Fig. 22.2

Funnel plot for the 9 log-transformed SIRs/RRs, using data extracted from Table 22.2. The dashed line represents SIR = 1 or RR = 1. The alphabet-numeric combinations are the IDs shown in Table 22.2, and each ID represents one study

Second, the SIR estimate, K07 in the plot, provided by Kobayashi et al. [33] is situated at the rim of the funnel, suggesting that while it gave a larger SIR estimate, it is not a precise estimate.

The paper by Aris [34] gave P(E) = 0.107 and P(E|O) = 0.14, yielding RR = 1.36 as discussed above. This is very close to the RR estimate of 1.6 reported by the paper.

22.5.2 An Overview on Case-Control Studies

Case-control studies are an alternative to cohort studies for investigating the association between exposure (in our case, having endometriosis) and disease (ovarian cancer). The basic questions for such studies are the degree of association between risk for disease and the factor(s) under investigation, the extent to which the observed association may result from bias, confounding, and/or chance, and the extent to which they may be described as causal [37]. A case-control study compares cases (in our case, women with ovarian cancer) and controls (women without ovarian cancer, say) with respect to their exposure (or lack thereof) or levels of exposure to a suspecting risk factor (in our case, having endometriosis). When the risk factor at hand is a dichotomous variable, such as having endometriosis or not, the outcome measure is typically the odds of exposure in cases as compared with that in controls, or OR. When the occurrence of disease is rare, such as ovarian cancer, the OR estimated from case-control studies becomes an acceptable approximation to the relative risk. Case-control studies can be a powerful tool in the investigation of exposure-disease relationship when both the disease and the exposure are rare. A prime example is the uncovering of the relationship between in utero exposure to diethylstilbestrol (DES) and vaginal adenocarcinomas in the daughters [38]. That study was based on just eight cases, each with four matched controls. Seven out of eight cases had been exposed to DES in utero, but in contrast none of the 32 controls had.

As with SIR or RR used in cohort studies, the OR or relative risk (RR) used in case-control studies is the measure of association between disease and exposure. However, the association could be causal but also could be merely a correlation. For women with endometriosis (E), or with ovarian cancer (O), the association between E and O could be due to a variety of scenarios. Figure 22.3 shows several scenarios in which E and O can be found to be associated. Scenario a is the case where factor X has a causal relationship with both E and O. E and O are associated simply because of the presence of the common risk factor X, which may or may not be measured in a study. It should be noted that E and O share at least one common risk factor, that is, the incessant ovulation/menstruation. Incessant ovulation or unopposed estrogen exposure is a known major risk factor for ovarian cancer [39]. Similarly, incessant menstruation is a known major and consistently identified risk factor for endometriosis [40]. Figure 22.4 shows two numerical examples, perhaps somewhat extreme but nonetheless not unusual cases. In example a, failure to control for the confounding factor gives rise to spurious results. It is interesting to point out that, while the OR for the O-E association in each stratum of factor X is 1, the OR for the association with pooled levels of the factor is 1.35 > 1. Of course, the failure to control for confounders can also go to the other direction, in which the pooled OR can be smaller than ORs in each stratum (example b).


Fig. 22.3

Diagrams showing 4 different scenarios in which E and O can be found associated


Fig. 22.4

Two hypothetical examples showing that failure to control for confounding can lead to spurious results. (a) Artificially inflated OR; (b) underestimated OR

In Fig. 22.3, scenario b shows the case in which both factors X and E represent the same underlying cause for O, such as the case when X and E represent different aspects of the same factor. Scenario c is the case where Eleads to X, which, in turn, has a causal relationship with O. In the case of E-O association, it is possible that the diagnosis of endometriosis may result in the use of danazol, an androgenic agent, which could increase the risk of Oin light of the “androgen hypothesis” of ovarian cancer [4142]. In other words, it could be the exposure to an androgenic agent, once a popular therapeutic for endometriosis, that increases the risk of ovarian cancer, not the endometriosis itself. Scenario d is the case in which E-O has a causal relationship.

It should be noted that factor X in scenario a is considered a confounding factor. Confounding is the distortion of a disease/exposure association brought about by the association of other factors with both disease (O) and exposure (E) [37].

The magnitude of OR is a measure of the strength of association, or the effect size. In general, when an OR is large, say greater than 10, the association is likely to be genuine. For example, the relative risk of having cervical cancer in women with HPV positivity vs. negativity is about 1000. Depending on the number of daily cigarettes consumed, the OR for smoking-lung cancer association ranges from 7 to about 27. The OR for the association between the DES exposure and vaginal cancer is about 40. In contract, for an association with an OR < 2, it is likely that the OR estimate could be a result of confounding or bias and needs to be scrutinized rigorously even though the association could also be genuine [43].

22.5.3 Case-Control Studies

Eleven case-control studies can be identified (Table 22.3). As expected, these studies vary in the types of ovarian cancer and the selection of controls (women with endometriosis or endometriomas, or infertility). Note that the last study listed in Table 22.3 is somewhat different from the rest, since both cases and controls had endometriosis, but the risk (protective) factor of interest was whether or not the subject had all visible endometriotic lesions removed [53]. Regardless, the forest plot revealed that the pooled (raw) OR estimate is 1.54 (95 % CI = 1.43–1.66, excluding the last study; Fig. 22.5), suggesting that overall, the OR value is moderate. In addition, there is little heterogeneity (p = 0.47 for heterogeneity test).

Table 22.3

Published case-control studies reporting ovarian cancer risk in women with endometriosis



Year of publication

# of cases

# of controls

# case w/ endo

# controls w/ endo

Adjusted OR (95%CI)



Ness et al. [44]






1.7 (1.2–2.4)



Ness et al. [45]






1.73 (1.10–2.71)

Originally considered infertility as a possible risk factor but also looked at patients with infertility due to endometriosis


Modugno et al. [46]






1.32 (1.06–1.65)

This study also included some data from Ness et al. [44]


Borgfeld et al. [47]






1.34 (1.03–1.75)



Merritt et al. [48]






1.31 (0.97–1.78)



Nagle et al. [49]






3.0 (1.5–5.9)

For endometrioid and clear cell ovarian cancers


Rossing et al. [50]






1.6 (1.1–2.3)

For invasive ovarian cancer


Wu et al. [51]






1.66 (1.01–2.75)



Pearce et al. [5]






1.46 (1.31–1.63)

Invasive clear cell and endometrioid ovarian cancers


Merritt et al. [52]






1.92 (1.36–2.71)

Low-grade serous, endometrioid/mixed, mucinous and clear cell


Melin et al. [53]






0.30 (0.12–0.74)

All cases and controls had endometriosis. The factor of interest was whether or not the subject had a radical extirpation of all visible endometriosis

OR odds ratio, SE standard error, CI confidence interval, # numbers, endo endometriosis, w/ with


Fig. 22.5

Forest plot summarizing the results from 10 case-control studies using data from Table 22.3

As with the SIR/RR estimates, the funnel plot of the log ORs from the ten studies indicates that the plot also looks like an asymmetric funnel, with its tip pointing towards somewhere near log OR = 0, i.e., OR = 1 (Fig. 22.6). Since there is no indication of bias in inclusion criteria or heterogeneity, this suggests that there may be a publication bias towards favoring positive studies and higher estimates of odds ratios may well be a chance variation. In addition, the plot seems to suggest that the true OR is quite moderate.


Fig. 22.6

Funnel plot for the log ORs, using data extracted from Table 22.3. The dashed line represents OR = 1. The alphabet-numeric combinations are the IDs shown in Table 22.3, and each ID represents one study

The study by Melin et al. [53] is of particular interest since, unlike other case-control studies, it examined the effect of surgical treatment on the endometriosis-ovarian cancer association. By linkage to the National Swedish Cancer Register, it identified all women diagnosed with epithelial ovarian cancer at least 1 year after the endometriosis diagnosis (cases). Two controls per case with no ovarian cancer before the date of cancer diagnosis of the case were randomly selected from the study base and matched for the year of birth. It found an OR of 0.30 (95 % CI = 0.12–0.74) for women who received a complete removal of all visible endometriosis. That is, for a woman with endometriosis, her risk of developing ovarian cancer could be cut by 70 % if she had all visible endometriotic lesions removed. It is worth noting that so far no other case-control studies have taken surgical completeness into consideration, since the study by Melin et al. strongly suggests this can be a protective factor.

22.5.4 Other Considerations

The control for shared risk (and/or protective) factors between E and O appears to be a big challenge in sorting out the relationship between E and O association. Besides the scenario depicted in Fig. 22.4a, it is known that E and O share some other common risk/protective factors, for example, the use of oral contraceptives (OC) and age at menarche. These factors are very likely to be causally associated with both E and O, effectively making them confounding factors when assessing the E-O association in case-control studies. However, while some studies did control for OC use, few, if any, controlled for the number of ovulations/menstrual cycles.

While the mean age at onset of ovarian cancer is about 56 years [4], the onset of endometriosis occurs mostly and typically during women’s reproductive age. This has been taken as a support for temporality requirement in the Hill’s 9 criteria of causality [4]. Indeed, the reported mean age of EAOC cases is often significantly younger than ovarian cancer patients without endometriosis but older than women with endometriosis alone [34].

However, the case-control studies published so far have not demonstrated a clear, graded temporal relationship between endometriosis and ovarian cancer. Most epithelial tumors take a latent period of at least 15 years to develop [4]. If endometriosis is a precursor of certain types of ovarian cancer, then it should take a certain latent period, likely to be shorter than 15 years, for ovarian cancer to develop. Consequently, one would see that after excluding some cases with endometriosis, say, ≤3 years of interval between the diagnosis of endometriosis and of ovarian cancer, the OR would go up since this would effectively remove many “noisy” cases which would dilute the association signal. Unfortunately, we actually see the opposite from the study by Pearce et al. [5]. Figure 22.7 is a graphical rendition of its sensitivity analysis (Table 4 in [5]). One can see that once the cases who had at least 3, 5, or 10 years of interval between the diagnosis of endometriosis and of ovarian cancer were removed, the OR estimate goes down considerably.


Fig. 22.7

A graphical rendition of the sensitivity analysis for the association of endometriosis and risk of invasive ovarian cancer based on timing (time interval) of diagnosis between the two diseases, as reported by Pearce et al. [5] (their Table 4). When patients with the time interval less than or equal to 3 years, 5 years, and 10 years are excluded, the decrease in the OR estimate is seen

22.6 Should Any Action Be Taken?

Given the somewhat consistent but rather moderate increase in OR, some investigators believe that ovarian cancer originates from endometriosis, at least for clear cell carcinoma and endometrioid adenocarcinoma [54]; hence, screening, laboratory, and imaging evaluation should be “recommended for early detection of malignant disorders in women with endometriosis” [55]. Some even show that patients with EAOC actually had a more favorable prognosis [5658]. However, other studies do not find such evidence [5960].

Due to the low incidence of ovarian cancer and the rather moderate increase in risk, extreme caution needs to be exercised when conveying the message to the public and also in the context of screening. For clear cell ovarian cancer, the prevalence is reported to be 13 per 100,000 women (Surveillance Epidemiology and End Results: http://​seer.​cancer.​gov/​statfacts/​html/​ovary.​html, accessed January 17, 2013). Assuming, perhaps too optimistically, that a screening test exists that is 99 % sensitive and 99 % specific. Even with this rosy scenario, the corresponding positive predictive value is a disappointing 3.7 %. In other words, out of 100 women who have tested positive, fully 96 would have a false positive result and be likely to be subjected to invasive procedures. Therefore, given the low incidence and also the moderate increase in OR, it is perhaps premature to talk about screening.

22.7 Conclusion

From the funnel plots for the SIR/RRs reported from cohort studies and the ORs from case-control studies, it seems that there may be a publication bias towards favoring positive studies. In addition, the plots seem to suggest that the true effect size is very moderate. Yet the vast discrepancy between RRs estimated from prevalence of endometriosis in women with ovarian cancer and ORs reported from published case-control studies is puzzling. Since the prevalence is likely an underestimate, the true RR is likely to be higher, which would highlight the discrepancy even more. It is unclear as to what factors contributed to the discrepancy. Have all epidemiological studies published so far underestimated the effect size due to failure to control for some, yet to be identified, confounders or certain biases of unknown sources? Or have many studies reporting the prevalence of endometriosis in ovarian cancer somehow overreported, perhaps unwittingly, because of ascertainment or selection bias and population idiosyncrasy or have simply fallen into the trap of attribution error? There is no answer as of now, and to address these questions would warrant more studies.

While the presence of ovarian endometriomas may generate a pro-inflammatory microenvironment that may be conducive to the development of ovarian cancer, it is noted that most, if not all, diseases, especially those associated with pain, have signs of inflammation. Even obesity has signs of inflammation. What is unclear is how the pro-inflammatory milieu in endometriosis per se leads to ovarian cancer. It is also unclear as to whether the peritoneal or vagino-rectal deep infiltrating endometriosis would also increase the risk of ovarian cancer more than that of other gynecological cancers. Moreover, the failure in providing or adjustment for information on treatment in many published epidemiological studies raises the question as to whether a surgery or drug treatment can actually reduce the risk of ovarian cancer. It also raises the question as to whether the use of danazol, an androgenic agent and once a popular therapeutics, could increase the risk of ovarian cancer. Finally, due to the nature of case-control studies, it cannot rule out that both endometriosis and ovarian cancer (especially clear cell or endometrioid type) may simply share some common risk and/or protective factors—such as the “incessant menstruation” and OC use, or yet to be identified—so that an elevated OR is still an association, but the relationship is by no means causal.

The finding reported by Melin et al. [53] that the risk of developing ovarian cancer in women with endometriosis could be cut by 70 % if they had all visible endometriotic lesions removed is particularly interesting. So far almost all other case-control studies published failed to control the effect of surgical treatment on the endometriosis-ovarian cancer association, even though the diagnosis of endometriosis is usually established by laparoscopic visualization of lesions, which is almost always followed by surgical removal of the lesions. Of course, some subtypes of endometriotic lesions, such as deep infiltrating endometriosis, can be challenging to remove completely. But does this mean that those women who had a complete removal of all their visible lesions are those who had less severe endometriosis? How does this surgical completeness or radicality interact with the extensiveness or severity of endometriosis and impact the risk of developing ovarian cancer? There are no data to answer these questions.

While younger age at diagnosis of endometriosis compared to that of ovarian cancer is often taken as a proof of temporality in the causal link, many epidemiological studies have not clearly demonstrated a temporal relationship between endometriosis and ovarian cancer. Since it is now well documented that, similar to cancer, endometriotic lesions are monoclonal in origin [61], one way to prove the temporal relationship, perhaps once for all, and to provide a convincing proof that some histotypes of ovarian cancer originate from endometriosis is to reconstruct a phylogenetic trees delineating the relationship between endometriotic lesions and ovarian cancer based on molecular clock. A proof-of-concept study demonstrating the utility of the molecular clock in reconstructing the geological relationship among pieces of endometrial fragments demonstrates that the phylogenetic approach is feasible using today’s genetic technology [62].

In summary, while published clinical and epidemiological studies strongly implicated the risk, though moderate, of developing certain histotypes of ovarian cancer in women with endometriosis, many stones are still left unturned. Future studies need to determine whether the association is causal with a clear temporal relationship or merely association due to exposure to shared risk factors. Given the moderate association, it is perhaps premature to institute any actionable measures as of now. Future studies also need to resolve an apparent discrepancy in estimated effect size between the clinical data and epidemiological data and to further delineate the molecular pathways linking endometriosis and ovarian cancer. Care should be taken to avoid making “just-so” stories.


I would like to thank Professor Paolo Vercellini for stimulating the discussion when preparing for this chapter. This research was supported in part by grant 81270676 from the National Science Foundation of China and financial support from Fudan University and the Shanghai Key Laboratory of Female Reproductive Endocrine-Related Diseases.

Conflict of Interest Statement

None declared.



Sayasneh A, Tsivos D, Crawford R. Endometriosis and ovarian cancer: a systematic review. ISRN Obstet Gynecol. 2011;2011:140310. doi:10.​5402/​2011/​140310.PubMedCentralPubMedCrossRef


Kobayashi H, Kajiwara H, Kanayama S, Yamada Y, Furukawa N, Noguchi T, Haruta S, Yoshida S, Sakata M, Sado T, Oi H. Molecular pathogenesis of endometriosis-associated clear cell carcinoma of the ovary (review). Oncol Rep. 2009;22(2):233–40.PubMed


Hill AB. The environment and disease: association or causation? Proc R Soc Med. 1965;58:295–300.PubMedCentralPubMed


Vigano P, Somigliana E, Parazzini F, Vercellini P. Bias versus causality: interpreting recent evidence of association between endometriosis and ovarian cancer. Fertil Steril. 2007;88(3):588–93.PubMedCrossRef


Pearce CL, Templeman C, Rossing MA, Lee A, Near AM, Webb PM, Nagle CM, Doherty JA, Cushing-Haugen KL, Wicklund KG, Chang-Claude J, Hein R, Lurie G, Wilkens LR, Carney ME, Goodman MT, Moysich K, Kjaer SK, Hogdall E, Jensen A, Goode EL, Fridley BL, Larson MC, Schildkraut JM, Palmieri RT, Cramer DW, Terry KL, Vitonis AF, Titus LJ, Ziogas A, Brewster W, Anton-Culver H, Gentry-Maharaj A, Ramus SJ, Anderson AR, Brueggmann D, Fasching PA, Gayther SA, Huntsman DG, Menon U, Ness RB, Pike MC, Risch H, Wu AH, Berchuck A. Association between endometriosis and risk of histological subtypes of ovarian cancer: a pooled analysis of case-control studies. Lancet Oncol. 2012;13(4):385–94.PubMedCentralPubMedCrossRef


Godwin AK, Testa JR, Hamilton TC. The biology of ovarian cancer development. Cancer. 1993;71(2 Suppl):530–6.PubMed


Vaughan S, Coward JI, Bast Jr RC, Berchuck A, Berek JS, Brenton JD, Coukos G, Crum CC, Drapkin R, Etemadmoghadam D, Friedlander M, Gabra H, Kaye SB, Lord CJ, Lengyel E, Levine DA, McNeish IA, Menon U, Mills GB, Nephew KP, Oza AM, Sood AK, Stronach EA, Walczak H, Bowtell DD, Balkwill FR. Rethinking ovarian cancer: recommendations for improving outcomes. Nat Rev Cancer. 2011;11(10):719–25. doi:10.​1038/​nrc3144.PubMedCentralPubMedCrossRef


Light R, Pilleman DB. Summing up. The science of reviewing research. Cambridge: Harvard University Press; 1984.


Team RDC. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna; 2009 (ISBN 3-900051-07-0, http://​www.​R-project.​org.).


Sampson JA. Endometrial carcinoma of the ovary arising in endometrial tissue in that organ. Arch Surg. 1925;10(1–72).


Scott RB. Malignant changes in endometriosis. Obstet Gynecol. 1953;2(3):283–9.PubMed


Somigliana E, Vigano P, Parazzini F, Stoppelli S, Giambattista E, Vercellini P. Association between endometriosis and cancer: a comprehensive review and a critical analysis of clinical and epidemiological evidence. Gynecol Oncol. 2006;101(2):331–41.PubMedCrossRef


Kurman RJ, Shih Ie M. Molecular pathogenesis and extraovarian origin of epithelial ovarian cancer–shifting the paradigm. Hum Pathol. 2011;42(7):918–31.PubMedCentralPubMedCrossRef


Network TCGA. Comprehensive molecular portraits of human breast tumours. Nature. 2012;490(7418):61–70.CrossRef


Aure JC, Hoeg K, Kolstad P. Carcinoma of the ovary and endometriosis. Acta Obstet Gynecol Scand. 1971;50(1):63–7.PubMedCrossRef


Kurman RJ, Craig JM. Endometrioid and clear cell carcinoma of the ovary. Cancer. 1972;29(6):1653–64.PubMedCrossRef


Russell P. The pathological assessment of ovarian neoplasms. I: Introduction to the common ‘epithelial’ tumours and analysis of benign ‘epithelial’ tumours. Pathology. 1979;11(1):5–26.PubMedCrossRef


Brescia RJ, Dubin N, Demopoulos RI. Endometrioid and clear cell carcinoma of the ovary. Factors affecting survival. Int J Gynecol Pathol. 1989;8(2):132–8.PubMedCrossRef


Vercellini P, Parazzini F, Bolis G, Carinelli S, Dindelli M, Vendola N, Luchini L, Crosignani PG. Endometriosis and ovarian cancer. Am J Obstet Gynecol. 1993;169(1):181–2.PubMedCrossRef


Jimbo H, Yoshikawa H, Onda T, Yasugi T, Sakamoto A, Taketani Y. Prevalence of ovarian endometriosis in epithelial ovarian cancer. Int J Gynaecol Obstet. 1997;59(3):245–50.PubMedCrossRef


Fukunaga M, Nomura K, Ishikawa E, Ushigome S. Ovarian atypical endometriosis: its close association with malignant epithelial tumours. Histopathology. 1997;30(3):249–55.PubMedCrossRef


Ogawa S, Kaku T, Amada S, Kobayashi H, Hirakawa T, Ariyoshi K, Kamura T, Nakano H. Ovarian endometriosis associated with ovarian carcinoma: a clinicopathological and immunohistochemical study. Gynecol Oncol. 2000;77(2):298–304.PubMedCrossRef


Vercellini P, Scarfone G, Bolis G, Stellato G, Carinelli S, Crosignani PG. Site of origin of epithelial ovarian cancer: the endometriosis connection. BJOG. 2000;107(9):1155–7.PubMedCrossRef


Oral E, Ilvan S, Tustas E, Korbeyli B, Bese T, Demirkiran F, Arvas M, Kosebay D. Prevalence of endometriosis in malignant epithelial ovary tumours. Eur J Obstet Gynecol Reprod Biol. 2003;109(1):97–101.PubMedCrossRef


Dzatic-Smiljkovic O, Vasiljevic M, Djukic M, Vugdelic R, Vugdelic J. Frequency of ovarian endometriosis in epithelial ovarian cancer patients. Clin Exp Obstet Gynecol. 2011;38(4):394–8.PubMed


Kondi-Pafiti A, Papakonstantinou E, Iavazzo C, Grigoriadis C, Salakos N, Gregoriou O. Clinicopathological characteristics of ovarian carcinomas associated with endometriosis. Arch Gynecol Obstet. 2012;285(2):479–83.PubMedCrossRef


Brinton LA, Gridley G, Persson I, Baron J, Bergqvist A. Cancer risk after a hospital discharge diagnosis of endometriosis. Am J Obstet Gynecol. 1997;176(3):572–9.PubMedCrossRef


Olson JE, Cerhan JR, Janney CA, Anderson KE, Vachon CM, Sellers TA. Postmenopausal cancer risk after self-reported endometriosis diagnosis in the Iowa Women’s Health Study. Cancer. 2002;94(5):1612–8.PubMedCrossRef


Brinton LA, Lamb EJ, Moghissi KS, Scoccia B, Althuis MD, Mabie JE, Westhoff CL. Ovarian cancer risk associated with varying causes of infertility. Fertil Steril. 2004;82(2):405–14.PubMedCrossRef


Brinton LA, Sakoda LC, Sherman ME, Frederiksen K, Kjaer SK, Graubard BI, Olsen JH, Mellemkjaer L. Relationship of benign gynecologic diseases to subsequent risk of ovarian and uterine tumors. Cancer Epidemiol Biomarkers Prev. 2005;14(12):2929–35.PubMedCrossRef


Melin A, Sparen P, Persson I, Bergqvist A. Endometriosis and the risk of cancer with special emphasis on ovarian cancer. Hum Reprod. 2006;21(5):1237–42.PubMedCrossRef


Murta EF, Nomelini RS, Ferreira FA, Lima MA. Ovarian clear cell carcinoma associated with endometriosis: a case report with immunohistochemical study. Eur J Gynaecol Oncol. 2007;28(5):403–5.PubMed


Kobayashi H, Sumimoto K, Moniwa N, Imai M, Takakura K, Kuromaki T, Morioka E, Arisawa K, Terao T. Risk of developing ovarian cancer among women with ovarian endometrioma: a cohort study in Shizuoka, Japan. Int J Gynecol Cancer. 2007;17(1):37–43.PubMedCrossRef


Aris A. Endometriosis-associated ovarian cancer: a ten-year cohort study of women living in the Estrie Region of Quebec, Canada. J Ovarian Res. 2010;3:2.PubMedCentralPubMedCrossRef


Stewart LM, Holman CD, Aboagye-Sarfo P, Finn JC, Preen DB, Hart R. In vitro fertilization, endometriosis, nulliparity and ovarian cancer risk. Gynecol Oncol. 2013;128:260–4


Sterne JAC, Egger M, Smith GD. Investigating and dealing with publication and other biases. In: Egger M, Smith GD, Altman DG, editors. Systematic reviews in Health Care: meta-analysis in context. London: BMJ Books; 2001. p. 189–208.CrossRef


Breslow NE, Day NE. Statistical methods in cancer research. vol. 1: The analysis of case-control studies. Lyon: IARC Scientific Publications No. 32. International Agency for Research on Cancer; 1980.


Herbst AL, Ulfelder H, Poskanzer DC. Adenocarcinoma of the vagina. Association of maternal stilbestrol therapy with tumor appearance in young women. N Engl J Med. 1971;284(15):878–81. doi:10.​1056/​NEJM197104222841​604.PubMedCrossRef


Casagrande JT, Louie EW, Pike MC, Roy S, Ross RK, Henderson BE. “Incessant ovulation” and ovarian cancer. Lancet. 1979;2(8135):170–3.PubMedCrossRef


Vercellini P, Crosignani P, Somigliana E, Vigano P, Buggio L, Bolis G, Fedele L. The ‘incessant menstruation’ hypothesis: a mechanistic ovarian cancer model with implications for prevention. Hum Reprod. 2011;26(9):2262–73.PubMedCrossRef


Risch HA. Hormonal etiology of epithelial ovarian cancer, with a hypothesis concerning the role of androgens and progesterone. J Natl Cancer Inst. 1998;90(23):1774–86.PubMedCrossRef


Cottreau CM, Ness RB, Modugno F, Allen GO, Goodman MT. Endometriosis and its treatment with danazol or lupron in relation to ovarian cancer. Clin Cancer Res. 2003;9(14):5142–4.PubMed


Rothman KJ, Poole C. A strengthening programme for weak associations. Int J Epidemiol. 1988;17(4):955–9.PubMedCrossRef


Ness RB, Grisso JA, Cottreau C, Klapper J, Vergona R, Wheeler JE, Morgan M, Schlesselman JJ. Factors related to inflammation of the ovarian epithelium and risk of ovarian cancer. Epidemiology. 2000;11(2):111–7.PubMedCrossRef


Ness RB, Cramer DW, Goodman MT, Kjaer SK, Mallin K, Mosgaard BJ, Purdie DM, Risch HA, Vergona R, Wu AH. Infertility, fertility drugs, and ovarian cancer: a pooled analysis of case-control studies. Am J Epidemiol. 2002;155(3):217–24.PubMedCrossRef


Modugno F, Ness RB, Allen GO, Schildkraut JM, Davis FG, Goodman MT. Oral contraceptive use, reproductive history, and risk of epithelial ovarian cancer in women with and without endometriosis. Am J Obstet Gynecol. 2004;191(3):733–40.PubMedCrossRef


Borgfeldt C, Andolf E. Cancer risk after hospital discharge diagnosis of benign ovarian cysts and endometriosis. Acta Obstet Gynecol Scand. 2004;83(4):395–400.PubMedCrossRef


Merritt MA, Green AC, Nagle CM, Webb PM. Talcum powder, chronic pelvic inflammation and NSAIDs in relation to risk of epithelial ovarian cancer. Int J Cancer. 2008;122(1):170–6. doi:10.​1002/​ijc.​23017.PubMedCrossRef


Nagle CM, Olsen CM, Webb PM, Jordan SJ, Whiteman DC, Green AC. Endometrioid and clear cell ovarian cancers: a comparative analysis of risk factors. Eur J Cancer. 2008;44(16):2477–84.PubMedCrossRef


Rossing MA, Cushing-Haugen KL, Wicklund KG, Doherty JA, Weiss NS. Risk of epithelial ovarian cancer in relation to benign ovarian conditions and ovarian surgery. Cancer Causes Control. 2008;19(10):1357–64.PubMedCentralPubMedCrossRef


Wu AH, Pearce CL, Tseng CC, Templeman C, Pike MC. Markers of inflammation and risk of ovarian cancer in Los Angeles County. Int J Cancer. 2009;124(6):1409–15.PubMedCrossRef


Merritt MA, De Pari M, Vitonis AF, Titus LJ, Cramer DW, Terry KL. Reproductive characteristics in relation to ovarian cancer risk by histologic pathways. Hum Reprod. 28(5):1406–17.


Melin AS, Lundholm C, Malki N, Swahn ML, Sparen P, Bergqvist A. Hormonal and surgical treatments for endometriosis and risk of epithelial ovarian cancer. Acta Obstet Gynecol Scand. 2013;92(5):546–54. doi:10.​1111/​aogs.​12123.PubMedCrossRef


Kuhn E, Kurman RJ, Shih IM. Ovarian cancer is an imported disease: fact or fiction? Curr Obstet Gynecol Rep. 2012;1(1):1–9.PubMedCentralPubMedCrossRef


Baldi A, Campioni M, Signorile PG. Endometriosis: pathogenesis, diagnosis, therapy and association with cancer (review). Oncol Rep. 2008;19(4):843–6.PubMed


Erzen M, Rakar S, Klancnik B, Syrjanen K. Endometriosis-associated ovarian carcinoma (EAOC): an entity distinct from other ovarian carcinomas as suggested by a nested case-control study. Gynecol Oncol. 2001;83(1):100–8.PubMedCrossRef


Wang S, Qiu L, Lang JH, Shen K, Huang HF, Pan LY, Wu M, Yang JX, Guo LN. Prognostic analysis of endometrioid epithelial ovarian cancer with or without endometriosis: a 12-year cohort study of Chinese patients. Am J Obstet Gynecol. 2013;209:241.PubMedCrossRef


Melin A, Lundholm C, Malki N, Swahn ML, Sparen P, Bergqvist A. Endometriosis as a prognostic factor for cancer survival. Int J Cancer. 2011;129(4):948–55. doi:10.​1002/​ijc.​25718.PubMedCrossRef


Cuff J, Longacre TA. Endometriosis does not confer improved prognosis in ovarian carcinoma of uniform cell type. Am J Surg Pathol. 2012;36(5):688–95.PubMedCrossRef


Noli S, Cipriani S, Scarfone G, Villa A, Grossi E, Monti E, Vercellini P, Parazzini F. Long term survival of ovarian endometriosis associated clear cell and endometrioid ovarian cancers. Int J Gynecol Cancer. 2013.


Wu Y, Basir Z, Kajdacsy-Balla A, Strawn E, Macias V, Montgomery K, Guo SW. Resolution of clonal origins for endometriotic lesions using laser capture microdissection and the human androgen receptor (HUMARA) assay. Fertil Steril. 2003;79 Suppl 1:710–7.PubMedCrossRef


Wu Y, Guo SW. Reconstructing cellular lineages in endometrial cells. Fertil Steril. 2008;89(2):481–4. doi:10.​1016/​j.​fertnstert.​2007.​03.​028.PubMedCrossRef