Prevalence of endometriosis in Mayer-Rokitansky-Küster-Hauser syndrome variants: a systematic review and meta-analysis

‘There is nothing as practical as a good theory’ ( Lewin, 1943 ) ‘There is nothing as practical as a good theory’ ( Lewin, 1943 ) The strong association between obstructive Müllerian anomalies and endometriosis has usually been interpreted as evidence that an increase in pelvic exposure to menstrual blood and endometrium leads to a proportional increase in the likelihood of endometriosis development, thus supporting the validity of the retrograde menstruation (RM)/implantation theory ( Viganò et al. , 2024 ). On the other hand, the absence of such an increased risk of endometriosis in patients with non-obstructive Müllerian anomalies argues against the coelomic metaplasia/embryonic remnants theory ( Vercellini et al. , 2024b ). However, patients with both obstructive and non-obstructive Müllerian anomalies have a functional, normally menstruating uterus. This could potentially confound the findings, as such dichotomization does not allow for partial obstruction, which is difficult to detect and quantify. Moreover, the clinical characteristics of the two subpopulations are very different in terms of age, symptomatology, and indication for surgery ( Vercellini et al. , 2024b ). Therefore, to further verify the above assumptions, it would be important to assess the prevalence of endometriosis in women without a uterus as a manifestation of a Müllerian anomaly, i.e. in patients with the so-called Mayer–Rokitansky–Küster–Hauser syndrome (MRKHS), in whom the interruption of the development of the paramesonephric ducts during early embryogenesis usually results in agenesis of the uterus and the cranial two-thirds of the vagina. Two determinant conditions would here coexist: the theoretical absence of a functional endometrium (FE) and the presence of a major defect in the development of the Müllerian ducts during embryogenesis. The absence of endometriosis in MRKHS patients would further support the notion that the disease requires eutopic endometrium refluxing through the tubes to be transplanted into the pelvis ( Viganò et al. , 2024 ). On the other hand, the presence of endometriosis in patients without a uterus due to a congenital defect would provide undeniable evidence that endometriosis can develop in the absence of eutopic FE, strongly supporting the coelomic/embryonic remnants theory ( Batt and Mitwally, 2003 ; Mok-Lin et al. , 2010 ). Indeed, the finding of pelvic endometriosis in patients with MRKHS, who are presumed to lack FE, has traditionally been cited as the most obvious demonstration of the coelomic metaplasia/embryonic remnants theory and against a role for RM in the pathogenesis of endometriosis ( Uğur et al. , 1995 ; Kaya et al. , 2004 ; Balci et al. , 2008 ; Cho et al. , 2009 ; Mok-Lin et al. , 2010 ; Yan et al. , 2011 ; Troncon et al. , 2014 ). Nevertheless, the situation is not straightforward, as 48–99% of patients with MRKHS, have unilateral or bilateral uterine remnants (URs), in which FE may be found ( Brucker et al. , 2017 ; Konrad et al. , 2019 ). Both the ESHRE/ESGE classification of congenital anomalies of the female genital tract ( Grimbizis et al. , 2013 ) and the ASRM classification of Müllerian anomalies ( Pfeifer et al. , 2021 ) categorize MRKHS variants based on the presence or absence of FE within UR. The importance of such a distinction is illustrated by the case reported by Pedro Acién in 1986. He described a 26-year-old patient with complete uterovaginal agenesis, except for a small rudimentary left myometrial bud, which continued with a normal tube observed at laparoscopy. No UR or salpinx was seen on the right pelvic side, but an ovarian endometrioma was present. The authors argued that in this woman without FE, the RM theory had to be ruled out and the development of endometriosis ‘might be explained by celomic metaplasia or Müllerian remains [adjacent to or in the ovary] developed under endogenous or exogenous hormonal stimuli’. However, Acién et al. (1988) subsequently reported that the same patient underwent laparotomy 2 years later and a large right endometrioma was found adjacent to a small, occult retroperitoneal rudimentary horn with a cavity lined by FE and continuous with a patent tube. Consequently, the authors reasoned ‘that other cases reported as Rokitansky–Küster–Hauser syndromes with endometriosis might have a similar Müllerian anomaly with a functioning endometrium’. Therefore, the detection of endometriosis in patients with MRKHS and an UR containing FE cannot be considered as evidence against the RM hypothesis. Instead, comparing endometriosis prevalence in patients with MRKHS and UR with FE (MRKHSFE+) or without FE or UR (MRKHSFE−) may provide a better modality to verify the pathogenic role of the ‘eutopic’ endometrium in the context of the same, major Müllerian anomaly. In other words, a substantially higher endometriosis prevalence in MRKHSFE+ than in MRKHSFE− patients would support the RM theory, also considering that hypoplastic or normal tubes are generally present and connected with UR. Moreover, as MRKHSFE+ and MRKHSFE− patients would be selected from the same study population, i.e. individuals with a major developmental Müllerian defect, confounding should be limited. Conversely, a similar or only slightly different endometriosis prevalence in MRKHSFE+ and MRKHSFE− patients would reasonably support the coelomic metaplasia/embryonic remnants hypothesis. However, the limited number of patients described in individual published studies is insufficient to reliably characterize the relationship between endometriosis and MRKHS variants ( Tian et al. , 2021 ), and some investigators suggested that it would be pathogenically interesting and clinically useful to benefit from the results of a meta-analysis on this specific issue ( Dabi et al. , 2020 ). Given this background, we decided to systematically review the available evidence on the prevalence of endometriosis in women with MRKHS since 1980. In particular, in light of the above considerations, we aimed to compare the frequency of endometriosis in MRKHSFE+ versus MRKHSFE− patients.

The flow diagram of the literature search results is shown in Fig. 1 . From a total of 834 records screened, 800 studies that did not meet the inclusion criteria were excluded. The main reason for exclusion was that these studies did not specifically address the presence or absence of endometrium (e.g. Acién, 1986 ; Pati and Mukhopadhyay, 2009 ; Gweon et al. , 2016 ). A further 72 studies were identified and included using the ‘similar articles’ and ‘cited by’ functions of PubMed. Four studies ( Fedele et al. , 2000 , 2006b , 2008 ; Brucker et al. , 2022 ) describing patients included in a later larger series were also excluded. A total of 102 studies published between 1980 and 2023, 29 of which were consecutive clinical case series ( Rock et al. , 1980 ; Olive and Henderson, 1987 ; Kvint and Wilhelmsson, 1988 ; Fedele et al. , 1990 , 1996 , 2007 ; Strübbe et al. , 1992 ; Uğur et al. , 1995 ; Soong et al. , 1996 ; Creatsas et al. , 2001 ; Plevraki et al. , 2004 ; Oppelt et al. , 2006 ; Darwish, 2007 ; Khen-Dunlop et al. , 2007 ; Marsh et al. , 2013 ; Parikh et al. , 2013 ; Perez-Medina et al. , 2013 ; Rall et al. , 2013 ; Will et al. , 2013 ; Preibsch et al. , 2014 ; Grimbizis et al. , 2015 ; Brucker et al. , 2017 ; Dabi et al. , 2020 ; Deng et al. , 2020 ; Paul et al. , 2020 ; Piriyev and Römer, 2020 ; Kisu et al. , 2021 ; Tian et al. , 2021 ; Steinmacher et al. , 2022 ) and 73 were case reports ( Neinstein and Castle, 1983 ; Purwar et al. , 1983 ; Winer-Muram et al. , 1984 ; Acién et al. , 1988 , 2010 ; Esakowitz and Yates, 1988 ; Metzger et al. , 1988 ; Powell et al. , 1988 ; Griggs et al. , 1990 ; Reed et al. , 1990 ; Yeko et al. , 1992 ; Kaminski et al. , 1994 ; Chapron et al. , 1995 ; Malik et al. , 1997 ; Russ et al. , 1997 ; Giatras et al. , 1998 ; Chatwani et al. , 1999 ; Forsnes et al. , 1999 ; Connell et al. , 2000 ; Dandu et al. , 2000 ; Fisher et al. , 2000 ; Schult et al. , 2000 ; Tsin et al. , 2000 ; Bellver-Pradas et al. , 2001 ; Dokmeci et al. , 2002 ; Pandey and Hamdi, 2003 ; Sönmezer et al. , 2003 ; Steinkampf et al. , 2003 ; Kaya et al. , 2004 ; Deligeoroglou et al. , 2005 ; Griesinger et al. , 2005 ; Fedele et al. , 2006a ; Goluda et al. , 2006 ; Dietrich et al. , 2007 ; Harland et al. , 2007 ; Balci et al. , 2008 ; Rajeswari et al. , 2008 ; Raudrant et al. , 2008 ; Cho et al. , 2009 ; Lamarca et al. , 2009 ; Parkar and Kamau, 2009 ; Fukuda et al. , 2010 ; Mok-Lin et al. , 2010 ; Terada et al. , 2010 ; Deka et al. , 2011 ; Elliott et al. , 2011 ; Yan et al. , 2011 ; Kondo et al. , 2013 ; Bacopoulou et al. , 2014 ; Bhuyar, 2014 ; Dragusin et al. , 2014 ; Kawano et al. , 2014 ; Kundu et al. , 2014 ; Shoar et al. , 2014 ; Troncon et al. , 2014 ; Valecha et al. , 2014 ; Hasegawa et al. , 2015 ; Sheikh et al. , 2015 ; Yakasai et al. , 2015 ; Salem Wehbe et al. , 2016 ; Ajiboye et al. , 2017 ; Amaratunga et al. , 2017 ; Karthik et al. , 2017 ; Htay et al. , 2019 ; Villa et al. , 2019 ; Albahlol et al. , 2020 ; Harzif et al. , 2021 ; Jha et al. , 2021 ; Romano et al. , 2021 ; Samantray et al. , 2021 ; Yang et al. , 2021 ; Dai et al. , 2023 ; Zhao et al. , 2023 ), were finally included in the assessment of endometriosis prevalence in patients with MRKHSFE+ versus MRKHSFE−. Preferred reporting items for systematic reviews and meta-analyses 2020 flow diagram of the literature search and study selection process for a systematic review of and endometriosis. The characteristics of the selected studies are shown in Supplementary Table S2 . A total of 666 MRKHS patients were included. The median [IQR] number of individuals included in case series was 10 [6–23], ranging from 2 to 106. The mean±SD age of patients at clinical diagnosis of MRKHS was 20.0 ± 6.3 years in the whole study group, 19.1 ± 3.4 in the case series subgroup, and 20.2 ± 6.9 in the case report subgroup. The corresponding ages at surgical diagnosis of endometriosis were 22.5 ± 7.3, 23.2 ± 7.6, and 22.4 ± 7.4 years, respectively. The MRKHS anatomical details in the study population are described in Table 1 . Half of the patients (334/666, 50.2%) had bilateral UR, 23 (3.5%) had unilateral UR, and 155 (23.3%) had no UR. In a further 154 (23.3%) participants UR was present, but it was not specified whether it was unilateral or bilateral. Fallopian tubes were absent in only 29 (4.4%) cases. Endometriosis was present in 71/666 participants (10.7%; 95% CI, 8.5–13.2%). A total of 51/593 (8.6%; 95% CI, 6.6–11.2%) individuals had endometriosis in the MRKHS patients included in case series. The prevalence estimate in case reports (20/73, 27.4%; 95% CI, 18.4–38.6%) was substantially higher and significantly different from that observed in case series ( P <0.0001). When considering only the 19 case series with ≥10 participants and regardless of preoperative evaluation for the presence of FE ( Olive and Henderson, 1987 ; Strübbe et al. , 1992 ; Uğur et al. , 1995 ; Fedele et al. , 1996 , 2007 ; Soong et al. , 1996 ; Creatsas et al. , 2001 ; Oppelt et al. , 2006 ; Khen-Dunlop et al. , 2007 ; Perez-Medina et al. , 2013 ; Rall et al. , 2013 ; Preibsch et al. , 2014 ; Grimbizis et al. , 2015 ; Brucker et al. , 2017 ; Dabi et al. , 2020 ; Deng et al. , 2020 ; Kisu et al. , 2021 ; Tian et al. , 2021 ; Steinmacher et al. , 2022 ) the aggregate proportion of MRKHS patients with endometriosis was 3.4% (41/1219; 95% CI, 2.5–4.5%). However, heterogeneity between studies was high, with individual proportions ranging from 0% in 12 studies ( Strübbe et al. , 1992 ; Fedele et al. , 1996 , 2007 ; Soong et al. , 1996 ; Creatsas et al. , 2001 ; Oppelt et al. , 2006 ; Khen-Dunlop et al. , 2007 ; Perez-Medina et al. , 2013 ; Rall et al. , 2013 ; Preibsch et al. , 2014 ; Grimbizis et al. , 2015 ; Kisu et al. , 2021 ) to 38% in one ( Dabi et al. , 2020 ). In addition, 31 of the 51 endometriosis cases observed in the total case series subgroup were reported in the two studies with the highest disease proportion ( Dabi et al. , 2020 , 8/21; Tian et al. , 2021 , 23/136). Anatomical details of patients with Mayer–Rokitansky–Küster–Hauser syndrome and with (n   =   71) or without (n   =   595) coexisting endometriosis. Literature data, 1980–2023. UR, uterine remnant. Of the 71 MRKHS patients with endometriosis, 64 (90.1%) had coexisting FE, and only seven (9.9%) had no evidence of FE within UR or no UR. Of these seven cases, UR was absent in four, bilateral in two, and unilateral in one. The mean±SD age at clinical diagnosis of MRKHS was 17.1 ± 2.7 years in patients with FE, and 16.8 ± 3.6 years in patients without FE. The age at surgical diagnosis of endometriosis was 23.1 ± 8.0 and 20.7 ± 4.5 years, respectively. The percentage of patients with endometriosis was 32.0% in the subgroup with FE (64/200; 95% CI, 25.9–38.8%) and 1.5% (7/466; 95% CI, 0.7–3.1%) in the subgroup without FE within UR/without UR. At meta-analysis, considering the data from case series, the estimated overall prevalence of endometriosis in MRKHSFE+ and MRKHSFE− patients was 16.8% (95% CI, 1.8–38.5%) ( Fig. 2 ) and 0% (95% CI, 0–0%) ( Fig. 3 ), respectively. In the sensitivity analysis that excluded the studies by Dabi et al. (2020) and Tian et al. (2021) the overall estimates of endometriosis prevalence in MRKHSFE+ and MRKHSFE− patients were 8.4% (95% CI, 0–24.6%) ( Supplementary Fig. S1 ) and 0% (95% CI, 0–0%) ( Supplementary Fig. S2 ), respectively. To account for the quality score of the considered studies ( Supplementary Figs S3 and S4 ), we computed a subgroup analysis by dividing the studies with quality score <4 and those with quality score ≥4. The overall prevalence estimates of endometriosis among the MRKHSFE+ patients were 0% (95% CI, 0–29.9%) for the study of Strübbe et al. (1992) , i.e. the only study with a quality score <4, and 19.0% (95% CI, 2.4–42.3%) for the other studies, all with a quality score ≥4 ( Supplementary Fig. S5 ). Among the MRKHSFE− patients, the overall prevalence estimates of endometriosis were 0% (95% CI, 0–1.1%) and 0% (95% CI, 0–0%) for studies with a quality score <4 and ≥4, respectively ( Supplementary Fig. S6 ). Heterogeneity was significant in all MRKHSFE+ meta-analyses ( P   0.980). However, estimating heterogeneity in the MRKHSFE− subgroup seems irrelevant with such low endometriosis prevalence figures. Forest plot relative to the meta-analysis of endometriosis prevalence in Mayer–Rokitansky–Küster–Hauser syndrome with functional endometrium (MRKHSFE+) patients. Results are reported as an overall estimate and for the case series and case reports subgroups separately. Forest plot relative to the meta-analysis of endometriosis prevalence in Mayer–Rokitansky–Küster–Hauser syndrome lacking functional endometrium (MRKHSFE−) patients. Results are reported as an overall estimate and for the case series and case reports subgroups separately. Diamonds are not visible as the CIs are null. Among women from the case report studies, the prevalence estimates of endometriosis in patients with and without FE were 51.8% (14/27; 95% CI, 34.0–69.3%) and 13.0% (6/46; 95% CI, 5.7–26.0%), respectively. Considering the data from case series that reported the prevalence of endometriosis for both MRKHSFE+ and MRKHSFE− patients, the risk of endometriosis was significantly higher in FE+ compared with FE− patients (OR = 12.0; 95% CI, 5.1–28.3) ( Fig. 4 ), and the result was confirmed in the sensitivity analysis that excluded Dabi et al. (2020) and Tian et al. (2021) (OR = 4.3; 95% CI, 1.5–12.4) ( Supplementary Fig. S7 ), with significant heterogeneity in both cases ( P  < 0.01). Forest plot showing the odds ratios of endometriosis risk of Mayer–Rokitansky–Küster–Hauser syndrome with versus lacking functional endometrium (MRKHSFE+ vs MRKHSFE−) patients. Only the 14 case series reporting both MRKHSFE+ and MRKHSFE− individuals are considered (E+, endometriosis present; E−, endometriosis absent; MRKHSFE+, Mayer–Rokitansky–Küster–Hauser syndrome with functional endometrium; MRKHSFE−, Mayer–Rokitansky–Küster–Hauser syndrome without functional endometrium. The characteristics of the endometriotic lesions observed in 71 MRKHS participants are described in Table 2 . The most common lesions were ovarian endometriomas (n   =   36) and superficial peritoneal implants (n   =   33), whereas only six deep infiltrating lesions (DIEs) were reported. The sum does not add up to the total because some patients had more than one type of lesion. Characteristics and stage of endometriotic lesions detected in 71 patients with Mayer–Rokitansky–Küster–Hauser syndrome. Literature data, 1980–2023. r-ASRM, revised American Society for Reproductive Medicine classification of endometriosis. Lesion laterality can only be reliably assessed for ovarian endometriomas. Endometriotic cysts were located on the right side of the pelvis in 16 cases, on the left side in 15, and on both sides in 5. In no case did a right endometrioma develop in the absence of FE on the right side of the pelvis. Three endometriomas developed on the left ovary in the absence of FE, and in one case bilateral endometriomas were reported, but FE was detected on the left pelvic side only. Overall, FE was present on the same pelvic side as ovarian endometriomas in 32/36 (88.9%) cases. The median [IQR] methodological quality score of the 102 included reports was 3 [2; 4], range, 1–7. Three studies were in Q1, 40 in Q2, 50 in Q3, and 9 in Q4. The corresponding figures were 4 [3–5], min 2, max 7, for the case series subgroup, and 3 [2–4], min 1, max 5, for the case report subgroup, respectively ( Supplementary Figs S3 and S4 ). Of relevance, the quartile distribution was 3 in Q1 (10%), 18 in Q2 (62%) and 8 in Q3 (28%) for the 29 case series, and 22 in Q2 (30%), 42 in Q3 (58%), and 9 in Q4 (12%) for the 73 case reports. Two studies that reported unexpectedly high proportions of individuals with endometriosis deserve closer scrutiny. Dabi et al. (2020 , France) retrospectively reviewed the medical records of all the patients with a diagnosis of MRKHS confirmed by pelvic US and/or MRI evaluated at their tertiary referral centre for Müllerian anomalies between 1991 and 2013. Mainly symptomatic patients with cyclic pelvic pain and with at least one uterine horn remnant were included, whereas patients with complete utero-vaginal aplasia (i.e. without UR) were excluded. Histopathological examination revealed FE in 11 of the 14 patients who underwent UR removal. The study selection criteria probably explain the unusual frequency of endometriosis (8/21, 38%). However, the diagnosis of peritoneal endometriosis was confirmed histologically in only five of the eight cases reported. Tian et al. (2021) , conducted a retrospective review of data of 511 patients with a US and/or MRI diagnosis of MRKHS evaluated at the Peking Union Medical College Hospital, China, between January 2009 and January 2020. A total of 136 participants underwent surgery, 57 without UR resection, and 79 with UR resection. In the latter group, 45 people had neither FE nor endometriosis. Of the remaining 34 cases with FE in UR, 23 had coexisting histopathologically confirmed ovarian and/or peritoneal endometriosis. Endometriosis was never observed in patients without UR or without FE. Considering the entire surgical series, the proportion of patients with endometriosis was 17% (23/136). Uğur et al. (1995 , Turkey) published the findings of a large series of participants with Müllerian anomalies. ‘Histologically proven’ (no histological features given), stage I endometriosis was identified in a single patient who was described as having combined uterovaginal hypoplasia/agenesis. The authors claimed that FE was absent (presumably on preoperative US), but did not provide anatomical or imaging details. Kaya et al. (2004 , Turkey) described a 14-year-old female referred for primary amenorrhoea. A 7-cm left ovarian endometrioma was excised, and histopathology showed endometrial glands and stroma with haemosiderin-laden macrophages. The uterus was initially reported to be absent and only remnants of rudimentary tubes were described. However, the presence of a ‘hypoplastic uterus’ is noted later in the article. A pelvic MRI was not performed. Balci et al. (2008 , Turkey) excised an 11 cm×8 cm, retroperitoneal, histologically proven (endometrial tissue), endometriotic, right perirenal cyst in a 17-year-old patient with a right pelvic kidney who complained of amenorrhoea. No UR was found on preoperative US and MRI and at laparotomy. The authors state that the tubes were absent in the abstract but present and normal in the main text. Cho et al. (2009 , South Korea) excised a left 6-cm endometrioma laparoscopically in a 26-year-old, symptomatic patient without UR on preoperative US and at surgery. Pelvic adhesions were present but were lysed to allow adequate visualization. Histology revealed endometrial glands and stroma lining the cyst with haemosiderin-laden macrophages. Mok-Lin et al. (2010 , USA) performed laparoscopies for chronic pelvic pain in a woman without UR and tubes at the ages of 20 and 25 years. On both occasions, stage I endometriosis was found in the pouch of Douglas but, unfortunately, it was electrocoagulated without biopsy. As a baseline MRI revealed the possible presence of UR, the patient was initially placed on continuous oral contraceptive use ‘to minimize the risk of hematometra in the event that the uterine horns were indeed present’. Yan et al. (2011 , China) found a 5-mm superficial endometriotic nodule on the right ovary at laparotomy for sigmoid colon neovaginoplasty in a 23-year-old asymptomatic patient with a single UR without evidence of FE on preoperative transabdominal US. Pelvic MRI was not performed. Moreover, the photomicrograph of haematoxylin and eosin-stained slide of the ovarian lesion (page 231, Fig. 1 of the original publication) is of poor quality and does not allow identification of endometrial glands or endometrial stromal cells (see Acknowledgements section below). Troncon et al. (2014 , Brazil) excised a left ovarian endometrioma in a 24-year-old patient with bilateral UR who had undergone sigmoid colon neovaginoplasty 7 years previously. Preoperative MRI and US had erroneously failed to detect the presence of UR. Thus, it seems unclear how the presence of FE within UR could be excluded, as laparoscopy is not informative in this regard. ‘Biopsy confirmed cystic endometriosis’ was described, but histological features were not reported. The seven studies above were considered to be at high risk of bias, mainly because of selection, ascertainment, and reporting issues. In particular, the exposure and/or outcome were unclear. Preoperative MRI was either not performed, uninformative, or incorrect (Uğur et al. , 1985; Kaya et al. , 2004 ; Mok-Lin et al. , 2010 ; Yan et al. , 2011 ; Troncon et al. , 2014 ); endometriosis was not biopsied ( Uğur et al. , 1995 ; Mok-Lin et al. , 2010 ); the histological diagnosis of endometriosis was questionable ( Yan et al. , 2011 ); or the histological details were not reported ( Balci et al. , 2008 ; Troncon et al. , 2014 ). In two cases, the anatomical description was inconsistent ( Kaya et al. , 2004 ; Balci et al. , 2008 ); Based on the risk of bias assessment tool by Murad et al. (2018) , the median [IQR] score for this selected subset of studies was 2 [1–2], min 1, max 4.

This systematic review and meta-analysis was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 (PRISMA 2020) recommendations ( Page et al. , 2021 ). The study protocol was registered on PROSPERO (registration number, CRD42024512351). Our primary aim was to answer the following PECO research question ( Morgan et al. , 2018 ): in patients with MRKHS (population), is the presence of FE in UR (exposure), compared with the absence of FE in UR/absence of UR (comparison), associated with a higher prevalence of endometriosis (outcome)? The hypothesis was that MRKHSFE+ patients may be affected by RM and may be more likely to develop endometriosis than MRKHSFE− individuals. The secondary aim was to assess whether the quality of the evidence provided in case series and case reports was comparable. The hypothesis was that case reports as a whole may have a higher risk of bias than case series regarding the frequency of endometriosis in general and in patients without FE in particular. A markedly higher estimate in the case report subgroup compared with the case series subgroup could indicate publication bias. This review was exempt from Institutional Review Board approval as exclusively published de-identified data were used. This review was restricted to full-length, English-language articles published in peer-reviewed journals between 1 January 1980 and 1 June 2024. We considered studies published before 1980 to be less relevant based on the hypothesis that the imaging performance of equipment used in the 1970s to detect FE within UR was probably suboptimal and not comparable to modern US and magnetic resonance imaging (MRI) equipment. A systematic search of the electronic PubMed and Embase databases was performed in June 2024. The keyword ‘endometriosis’ was used in combination with ‘Mayer–Rokitansky–Küster–Hauser syndrome’, ‘Müllerian agenesis’, ‘uterine agenesis’, ‘vaginal agenesis’, ‘Müllerian anomalies’, and ‘female genital malformations’. References from relevant publications were systematically screened, and further articles were searched using PubMed’s ‘similar articles’ and ‘cited by’ functions. No automation tool was used in any phase of the whole process. Studies were selected if they reported both the presence versus absence of FE within UR/absence of UR assessed on preoperative US or MRI or histology after UR removal, and the presence versus absence of surgically confirmed endometriosis. Given the rarity of the MRKHS, case series and case reports were deemed eligible for inclusion. There are differing opinions on the minimum number of participants required for a study to be defined as a case series. For the present review, the definition of Murad et al. (2018) was adopted, i.e. ‘c ase series of more than one patient may allow narrative or quantitative synthesis ’. Moreover, the distinction between case series and descriptive cohort studies is not always straightforward ( Dekkers et al. , 2012 ; Esene et al. , 2014 ; Guo et al. , 2016 ; Munn et al. , 2020 ). Here we consistently adhered to the definition of Grimes and Schulz (2002) , i.e. ‘A case-series aggregate individual cases in one report’, and that of Fletcher et al. (1996) , i.e. ‘A case series is a prevalence survey of a group of individuals with a particular disease, performed at a single point in time. It is a particularly common way of delineating the clinical picture of rare diseases’. If a study included patients with different types of female genital malformation, only findings regarding MRKHS cases were extracted ( Olive and Henderson, 1987 ; Uğur et al. , 1995 ; Plevraki et al. , 2004 ; Parikh et al. , 2013 ; Paul et al. , 2020 ; Piriyev and Römer, 2020 ). Studies that did not investigate the presence of FE within UR, did not report the results of US or MRI, or histological examinations, or without abdominal exploration were excluded, as were editorials, opinions, reviews, and abstracts presented at meetings. We initially decided to exclude patients with concomitant bilateral fallopian tube agenesis, as this would prevent RM even in the presence of FE within UR. However, we subsequently decided to include these cases also, on the basis that the detection of endometriosis in the absence of an anatomical continuum between UR and the pelvis would provide valuable evidence for the coelomic metaplasia/embryonic remnants hypothesis. When studies reported data from patients with a US and/or MRI diagnosis of MRKHS but without systematic direct pelvic visualization, only those participants who underwent surgery were selected. Moreover, all patients who underwent surgery were considered for the overall estimate of endometriosis prevalence in the MRKHS study population, but only those cases with specific investigation of presence versus absence of FE were included in the analyses regarding the primary research objective. Two observers (F.C. and M.P.) independently screened titles and abstracts after removing duplicates and assessed all relevant full-text articles. Disagreements between the observers were resolved by discussion or, if necessary, by a third reviewer (A.D.) adjudication (e.g. in cases where endometriosis was diagnosed only visually and not histologically). The two reviewers independently extracted the following study characteristics and descriptive data of the included patients, as well as outcome data, using a standardized form: study identifiers, study design, setting, Müllerian anomalies classification adopted, endometriosis classification used, number and age of participants, presence versus absence of UR, side of unilateral UR, presence versus absence of FE within UR, diagnostic modality adopted to detect FE, presence versus absence of fallopian tubes, type of endometriotic lesions and, if available, disease stage, and presence of additional pelvic abnormalities. If patients underwent surgery with a description of MRKHS characteristics and no endometriotic lesions were reported, we assumed that endometriosis was absent, even if this was not formally stated. The methodological quality (risk of bias) of the selected reports was assessed using the tool for appraising and synthesizing the evidence derived from case reports and case series devised and proposed by Murad et al. (2018) . The four bias domains of the tool (selection; ascertainment; causality; reporting) were evaluated independently by two reviewers (F.C. and M.P.), and any disagreements were resolved by discussion with a third reviewer (A.D.). The tool includes eight leading explanatory questions, all with binary responses, with a resulting possible summary score ranging from 0 to 8. However, as the occurrence of a challenge/rechallenge phenomenon is irrelevant to the present study, we excluded question 5 from the causality domain. We assigned one point to question 6 (was there a dose-response effect?) only when an ovarian endometrioma was detected on the same side of the pelvis as an UR containing FE ( Supplementary Table S1 ). Consequently, the possible summary score here ranged from 0 to 7. If the answer to a specific question was unclear, this was treated as a negative response and a score of 0 was assigned. The score of each included study was divided into quartiles (6–7 points, Q1, best quartile; 4–5 points, Q2; 2–3 points, Q3; 0–1 points, Q4, worst quartile). The overall prevalence of endometriosis in the total MRKHS population and in each study subgroup (MRKHSFE+ and MRKHSFE−) was reported as a percentage with corresponding 95% CI. As the presence of FE is considered here as exposure, classic epidemiological 2×2 tables were created in which exposure (presence vs absence of FE) and outcome (presence vs absence of endometriosis) were calculated both for the entire MRKHS study population and separately for case series and case reports. The association between FE and endometriosis was evaluated using the χ 2 1 test. The risk of endometriosis in MRKHSFE+ patients compared to MRKHSFE− patients was reported by calculating odds ratios (OR) and their 95% CIs in the total study population and in each subgroup. The methodological quality of the case series and case reports subgroups was described using median values and interquartile ranges [IQRs]. Meta-analysis of the prevalence of endometriosis was performed using Metaprop, a command implemented in Stata to compute proportional meta-analyses ( Barker et al. , 2021 ; StataCorp., 2023). The Freeman Tukey arcsine square root transformation was applied to the data before pooling for meta-analysis ( Munn et al. , 2015 ). Subgroup meta-analyses were conducted to obtain overall estimates of endometriosis prevalence for case series and case report studies separately. To evaluate heterogeneity across studies, heterogeneity I 2   P -value was also reported. To compare the prevalence of endometriosis between MRKHS+ and MRKHS−, a meta-analysis was performed and overall OR with 95% CI was reported. To account for low prevalence rates and studies without events, meta-analysis using the Mantel–Haenszel method was applied ( Efthimiou, 2019 ). To evaluate heterogeneity across studies, heterogeneity P -value was also reported. For both prevalence and ORs, we performed also sensitivity meta-analyses by excluding the two outlier case series by Dabi et al. (2020) and Tian et al. (2021) , as particular selection criteria were adopted (e.g. asymptomatic individuals were excluded in Dabi et al. , 2020 ), endometriosis diagnosis was not always based on histopathological examination ( Dabi et al. , 2020 ), or identification of FE was possible for only some of the MRKHS patients with UR ( Tian et al. , 2021 ). Subgroup meta-analysis by study quality score (<4 vs ≥4) was also performed to summarize the prevalence. However, this subgroup analysis was not applied to estimate the overall OR as all the studies in which both MRKHSFE+ and MRKHSFE− patients were included had a score ≥4. All analyses were performed with Stata software, version 18.0 ( StataCorp, 2023 ).

Exactly a century ago, John Sampson (1924) in Albany, New York described the potential anatomical determinants of transtubal menstrual reflux, and Robert Meyer (1924) , in Berlin, proposed coelomic metaplasia, i.e. the transdifferentiation of the peritoneal mesothelium into endometrial epithelium, as the origin of endometriosis. After all these years, the pathogenesis of endometriosis is still not fully understood, and the RM/implantation and the coelomic metaplasia/embryonic remnants theories remain the most popular and controversial. It remains uncertain whether or not the aggregation of descriptive findings performed with our review can prove that endometriosis in MRKHS patients is caused by RM. In fact, more than two-thirds (136/200) of MRKHSFE+ did not have endometriosis. Thus, it could be argued that, similar to the general population of women of reproductive age, also in this specific patient population other pathogenic factors (e.g. immuno-endocrinological influences, environmental exposures, lifestyle changes) or individual susceptibility (e.g. genetic and epigenetic modifications) are likely to act as disease promoters and determine who will or will not develop endometriosis, over and above the occurrence of RM as the disease initiator and the conditio sine qua non for its onset ( Arab et al. , 2022 ; Crespi and Evans, 2023 ; Rahmioglu et al. , 2023 ; Bedrick et al. , 2024 ; Chandrakanth et al. , 2024 ; Knez et al. , 2024 ; Vallée et al. , 2024 ). In addition, although direct comparisons are questionable, the prevalence of endometriosis in MRKHSFE+ patients appears to be lower than in patients with other Müllerian anomalies classified as ‘obstructive’ ( Vercellini et al. , 2024b ). However, the amount of RM in individuals with normal uteri and classic obstructive Müllerian anomalies is expected to be substantially greater than that in MRKHSFE+ patients who have rudimentary URs containing an unknown, but probably limited, quantity of FE. Moreover, other mechanistic variables, such as frequency and strength of myometrial contractions ( Salmeri et al. , 2024 ) and patency and intramural diameter of the fallopian tubes ( Vercellini et al. , 2024a ), may determine how much cryptomenstruation can reach the pelvis. In conclusion, the prevalence of endometriosis in MRKHS patients was substantially higher in those with UR with FE compared to those with UR without FE/without UR. Moreover, when considering only the 19 case series with ≥10 participants, the proportion of MRKHS patients with endometriosis was 3.4% (41/1219; 95% CI, 2.5–4.5%). Endometriosis was found in only seven MRKHS patients without UR, one in case series and six in case reports. Considering the high risk of bias, the detection of endometriosis in MRKHS patients allegedly without FE in the few relevant case reports published in the last four decades should no longer be interpreted tout court as proof for the coelomic metaplasia/embryonic remnants theory. Our findings should raise awareness of the importance of accurately investigating and reporting the presence or absence of FE within UR, and of systematically performing biopsies of visually diagnosed endometriosis in MRKHS patients.

According to the results of this systematic review, about 1 in 10 patients with MRKHS also had pelvic endometriosis. The frequency was three times higher in case reports (27.4%) than in case series (8.6%). Importantly, UR was present in approximately three out of four MRKHS patients (511/666; 76.7%), and more than one-third of these participants had FE within UR (200/511; 39.1%). As fallopian tubes were almost always identified (637/666; 95.6%), it could be argued that all the anatomical–histological components that could potentially give rise to the RM phenomenon were present in a large proportion of the MRKHS study population. Almost a third (32%) of the MRKHSFE+ subjects had coexisting endometriosis compared to only 1.5% of the MRKHSFE− subjects. The answer to our original PECO research question is therefore positive. Indeed, the simple possibility that a critical volume of FE in UR is needed, together with a functional fallopian tube, to originate a sufficient exposure to RM, is an additional anatomical factor that should be considered as a primary variable potentially affecting the risk of endometriosis development among FE+ patients. This is supported by evidence in the baboon model, where a positive linear correlation has been observed between the weight of endometrium injected in the pelvis and the number of endometriosis lesions identified at follow-up laparoscopy ( D’Hooghe et al. , 1995 ), and by the biologically plausible evidence demonstrating a positive association between increased amount of menstruation (hypermenorrhoea) and endometriosis risk ( Cramer et al. , 1986 ; Darrow et al. , 1993 ; Matalliotakis et al. , 2008 ). Unfortunately, insufficient information on the volume of FE in UR was available in the studies included in the present review. Regarding our second research question, it should be considered that due to the uncontrolled and non-experimental study design, both case series and case reports are associated with an increased risk of selection, performance, detection, attrition, and reporting bias ( Guo et al. , 2016 ; Munn et al. , 2020 ). For example, investigators may not have reported all evaluated MRKHS patients; the presence of FE and/or endometriosis may not always have been adequately ascertained; some cases may not have been reported with sufficient anatomical and pathological details. This is one of the reasons why the results of case series and case reports are generally considered to be of low quality. Based on the tool proposed by Murad et al. (2018) , the risk of bias in our review was substantially higher in the case report subgroup than in the case series subgroup. Overall, the methodological quality score of 72% of the reports included in the case series was in the top half of the total possible score, whereas less than a third of the case reports scores were in Q1–Q2. Although we did not perform a formal statistical comparison between groups, as suggested by Murad et al. (2018) , the difference in the distribution of scores is very clear. Publication bias may particularly affect case reports when the unexpected association between MRKHS and endometriosis is identified. Indeed, clinicians dealing with an MRKHS patient with coexisting endometriosis may be more motivated to investigate the case, collect and record the findings, and submit it to a medical journal, and editors may be more inclined to accept the manuscript, than in situations where a ‘usual’ case of MRKHS without endometriosis is observed. This also seems suggested by the otherwise difficult to explain statistically significant difference in endometriosis prevalence between case series and case reports. Indeed, six out of the seven cases of endometriosis in patients without FE are case reports, whereas only one such case was reported in case series. As the aggregate estimates derived from case series differ substantially from those derived from case reports, the doubt arises as to where the truth lies. If the judgement is to be based on the ascertainment of the methodological quality of the reports, it is tempting to speculate that the results from case series are more likely to reflect the true patterns of association between different MRKHS variants and endometriosis. Finally, according to Dekkers et al. (2012) , in cohort studies ‘participants should be initially free of the outcome and be followed over time to assess the occurrence of the outcome’. This was not the case in the studies included in our review, where the presence of UR and FE (exposure), and endometriosis (outcome) were assessed at a single time point. As temporality cannot be assessed (i.e. in MRKHS patients, endometriosis developed after endometrial cycling within UR because of the onset of ovarian activity), association, but not causality can be established. Although positions of epidemiologists vary and we cannot exclude possible design misclassification (case series instead of descriptive cohort study), we do not believe that this could have influenced the reported estimates and comparisons. A detailed assessment of the seven cases of endometriosis in patients supposedly without detected FE raises perplexities. Preoperative MRI was not systematically performed or was misinterpreted, the anatomical description at surgery was incomplete or inconsistent, the histopathological diagnosis of endometriosis was absent or questionable, and precise microscopic features were not always reported. According to some researchers ( Fernando et al. , 2013 ), the visual diagnosis of superficial peritoneal endometriotic lesions is inaccurate, and this may be particularly relevant when dealing with potential cryptomenorrhoea in patients with Müllerian anomalies ( Fedele et al. , 1992 ). Thus, it seems uncertain whether FE (exposure) was really absent in all of the above cases and/or whether all of the lesions diagnosed as endometriosis (outcome) were truly disease or, for example, mere haemosiderin deposits ( Fedele et al. , 1992 ). This ascertainment bias seems to be crucial, because ultimately these seven supposedly FE-free cases represent the only evidence supporting the coelomic metaplasia/embryonic remnants theory in this specific study population, whereas all the other MRKHSFE+ cases would only merely confirm that RM is a biologically plausible mechanism for the development of endometriosis. However, the overall methodological quality of the evidence in these seven cases appears to be very low. The concordance between unilateral FE and ipsilateral presence of endometriotic lesions appears to provide further convincing, albeit indirect, evidence for RM as the main pathogenic driver of endometriosis development in MRKHS patients. According to Sönmezer et al. (2003) , when bilateral UR are discordant for the presence of endometrium, endometriosis develops on the side with UR containing endometrium, but not on the side with UR without endometrium. Indeed, our findings confirm this assumption as FE was almost always present in ipsilateral UR, when an endometrioma was detected (e.g. Tian et al. , 2021 ). Lesion laterality can be assessed in ovarian cysts, but not as reliably in superficial peritoneal implants and DIEs. The literature search, selection of studies, data extraction, and assessment of risk of bias were performed independently by two observers and discrepancies were resolved by discussion with a third adjudicating reviewer. Various evaluation tools have recently been proposed to critically appraise the methodological quality of case series ( Guo et al. , 2016 ; Murad et al. , 2018 ; Munn et al. , 2020 ). Although there are similarities between these instruments, some authors believe that tools designed exclusively for case series include questions that are specific to this study design ( Munn et al. , 2020 ). However, the joint tool for case series and case reports proposed by Murad et al. (2018) , allowed us to assess the risk of bias in both study types using the same checklist, thus providing the possibility to compare the quality of the evidence presented in the two subgroups. In this regard, specifying an a priori cut-off score to distinguish between high- and low-quality descriptive observational reports with the aim of including only the former ones in systematic reviews, is considered inappropriate by several epidemiologists and methodologists, as different questions may not have the same ‘weight’ ( Murad et al. , 2018 ). Alternatively, Munn et al. (2020) suggest that systematic reviewers themselves decide on cut-off values/scores to be used in specific circumstances to determine whether the quality of each considered study is high, moderate, or low. Following this advice, we have summarized the scores derived from the included studies and described their distribution on a scale of 0–7 (one irrelevant question was excluded) in quartiles, without performing a formal statistical comparison between case series and case reports. As only two electronic databases were searched and only English-language articles were screened, it cannot be excluded that additional cases of MRKHSFE− with endometriosis remained undetected. More in general, case series and case reports represent the lower level in the hierarchy of evidence. In addition, estimates based on limited numbers of patients imply wide CIs and much uncertainty. Therefore, the generalizability of our findings derived from a small study population can be questioned. However, MRKHS is a rare condition affecting approximately 1/5000 women, and the data included in our review comprise the only available evidence on which to base pathogenic considerations and clinical decision-making (e.g. removal of UR containing FE regardless of cyclical pain symptoms to prevent endometriosis onset; consideration of long-term hormonal suppressive therapy if surgery is not an option). Indeed, case series may be the only realistic design to study rare disorders ( Esene et al. , 2014 ). In such rare conditions, where population-based studies are not feasible, valuable knowledge can be gained from the evaluation of case reports and case series ( Martyn, 2002 ). A guideline has been developed specifically for the reporting of case reports, with the aim of facilitating their collection and combination into large datasets, thus allowing analysis of systematically aggregated information ( Gagnier et al. , 2013 ). According to Martyn (2002) , retrospective reviews of case notes are rarely complete. Furthermore, if the series is not consecutive or incomplete, it is impossible to know whether the study outcomes might have been very different in missing cases. In addition, Grimes and Schulz (2002) warn that, when considering the evidence from case series and case reports, causal inferences about the temporal sequence of the events should be dealt with great caution, since both exposure and outcome are determined simultaneously. The authors argue that descriptive studies can only be used to develop hypotheses about causes ( Grimes and Schulz, 2002 ). However, Kiene et al. (2013) challenge the notion that case reports cannot be used to establish causality and propose ‘cognition-based medicine’ to appraise the evidence from individual observations. According to this ‘expanded evidence’ approach, insight into the mechanisms of why a condition may determine an outcome allows ‘cognitional causality assessment’ as a form of vertical bridging between cause and effect.