Endometriosis and cancer: a systematic review and meta-analysis

Abstract

BACKGROUND Endometriosis is an often chronic, inflammatory gynaecologic condition affecting 190 million women worldwide. Studies have reported an elevated cancer risk among patients with endometriosis. However, prior research has included methodologic issues that impede valid and robust interpretation. OBJECTIVE AND RATIONALE We conducted a meta-analysis of studies investigating the association between endometriosis and cancer risk and analysed the results by methodologic characteristics. We discuss the implications of cancer screening in patients and management challenges faced by clinicians. SEARCH METHODS We searched PubMed and Embase databases for eligible studies from inception through 24 October 2019. We included cohort and case-control studies examining the association between endometriosis and cancer risk; cross-sectional studies and case reports were excluded. Publications had to present risk/rate/odds estimates with 95% CI. Random effects meta-analysis was used to estimate summary relative risks (SRR) and CIs. Heterogeneity across studies was assessed by the Q test and I2 statistics, and publication bias using Egger's and Begg's tests. Risk of bias and quality of the included studies were assessed using the risk of bias in non-randomized studies of interventions (ROBINS-I) tool. OUTCOMES Forty-nine population-based case-control and cohort studies were included. Twenty-six studies were scored as having a ‘serious’/‘critical’ risk of bias, and the remaining 23 ‘low’/‘moderate’. Cancer-specific analyses showed a positive association between endometriosis and ovarian cancer risk (SRR = 1.93, 95% CI = 1.68–2.22; n = 24 studies) that was strongest for clear cell (SRR = 3.44, 95% CI = 2.82–4.42; n = 5 studies) and endometrioid (SRR = 2.33, 95% CI = 1.82–2.98; n = 5 studies) histotypes (Pheterogeneity < 0.0001), although with significant evidence of both heterogeneity across studies and publication bias (Egger’s and Begg’s P-values < 0.01). A robust association was observed between endometriosis and thyroid cancer (SRR = 1.39, 95% CI =1.24–1.57; n = 5 studies), a very small association with breast cancer (SRR = 1.04, 95% CI =1.00–1.09; n = 20 studies) and no association with colorectal cancer (SRR = 1.00, 95% CI =0.87–1.16; n = 5 studies). The association with endometrial cancer was not statistically significant (SRR = 1.23, 95% CI =0.97–1.57; n = 17 studies) overall and wholly null when restricted to prospective cohort studies (SRR = 0.99, 95% CI =0.72–1.37; n = 5 studies). The association with cutaneous melanoma was also non-significant (SRR = 1.17, 95% CI =0.97–1.41; n = 7 studies) but increased in magnitude and was statistically significant when restricted to studies with low/moderate risk of bias (SRR = 1.71, 95% CI = 1.24–2.36, n = 2 studies). The most robust finding both in terms of statistical significance and magnitude of effect was an inverse association with cervical cancer (SRR = 0.68, 95% CI =0.56–0.82; n = 4 studies); however, this result has a high potential to reflect heightened access to detection of dysplasia for women who reached an endometriosis diagnosis and is thus likely not causal. Several additional cancer types were explored based on <4 studies. WIDER IMPLICATIONS Endometriosis was associated with a higher risk of ovarian and thyroid, and minimally (only 4% greater risk) with breast cancer, and with a lower risk of cervical cancer. However, this meta-analysis confirms that: a majority of studies had severe/critical risk of bias; there is impactful heterogeneity across studies—and for ovarian cancer, publication bias; and causal inference requires temporality, which in many studies was not considered. We discuss the implications of these potential associations from the perspectives of patients with endometriosis, clinicians involved in their care, and scientists investigating their long-term health risks.

Letter to the Editor: Endometriosis and malignancy—The intriguing relationship

“…Finally, we would like to stress that although close to doubled, the absolute risk of ovarian malignancy in people with endometriosis remains very low: the absolute risk of ovarian cancer increases from 1.3% in the general population to 2.5% in endometriosis patients 3,7 . Based on this low risk and the known risk of surgical menopause, it is not warranted to recommend performing bilateral salpingo‐oophorectomy in women with endometriosis for prevention of ovarian cancer.…”

“…The analysis did not consider the temporality of the association (e.g. ensuring that the endometriosis diagnosis preceded the cancer diagnosis by at least a biologically plausible window of cancer initiation, promotion, and detection); thus, it solely gives information on the cross‐sectional coexistence of endometriosis and malignancy, rather than malignancy risk in women with endometriosis 3 . This is an extremely important distinction for interpretation and public health messaging.…”

“…In the absence of the time‐varying component of the analysis, the results reflect correlated cross‐sectional factors rather than predictive factors. Further, these factors are largely related to the likelihood of endometriosis diagnosis, introducing diagnostic bias 3 . As an example, the higher levels of CA‐125 in patients with both endometriosis and ovarian cancer reflect the known higher levels of this tumor marker in ovarian cancer patients and women with endometriosis; however, without consideration of the timing of the measure, it is uncertain whether they predict cancer in endometriosis patients or are simply evidence of the presence of cancer in endometriosis patients.…”

“…In 2019, Li et al 6 reported a SRR of 1.96 (95% CI 1.69–2.29) based on 25 studies published between 1997 and 2017. Finally, in 2020 we reported the results of a meta‐analysis of 24 studies published between 1999 and 2019, which yielded a SRR of 1.93 (95% CI 1.68–2.22) for the association between endometriosis and cancer 3 . Based on the available literature, the risk of ovarian malignancy is thus up to 2 times higher—not 4—in people with endometriosis.…”

“…ensuring that the endometriosis diagnosis preceded the cancer diagnosis by at least a biologically plausible window of cancer initiation, promotion, and detection); thus, it solely gives information on the cross-sectional coexistence of endometriosis and malignancy, rather than malignancy risk in women with endometriosis. 3 This is an extremely important distinction for interpretation and public health messaging. However, the data available to the authors would allow for the scientifically-appropriate time-varying analysis, by performing survival analyses using Cox proportional hazards regression modeling and ensuring that endometriosis precedes cancer diagnosis by at least 1 year.…”