Prevalence of adenomyosis in women with subfertility: systematic review and meta‐analysis

To determine the prevalence of adenomyosis in women with subfertility.


INTRODUCTION
Adenomyosis is a benign condition characterized by the presence of ectopic endometrial glands and stroma in the myometrium 1 . It has been associated typically with multiparity in women older than 40 years with menorrhagia, in whom a definitive diagnosis was possible only through the histological analysis of hysterectomy specimens [2][3][4] . Recently, however, there has been a growing body of evidence suggesting that adenomyosis may present earlier in life with abnormal uterine bleeding, subfertility, pelvic pain or even without symptoms and that its diagnosis is possible using non-invasive techniques 2,5,6 . Yet, epidemiological data on the burden of adenomyosis remain scarce, with the reported prevalence varying widely between 5% and 70% 7 .
Adenomyosis is often diagnosed concurrently with endometriosis and/or fibroids. The prevalence of ultrasound-diagnosed adenomyosis has been reported to be as high as 89.4% in women with endometriosis [27][28][29] . Similarly, the prevalence of histologically diagnosed adenomyosis with concurrent fibroids (20%) was noted to be higher compared with that of isolated adenomyosis (8%) on specimens of supracervical hysterectomies 7,30 . However, the burden of isolated adenomyosis remains understudied.
In the most recent systematic review on the prevalence of adenomyosis in women with subfertility, conducted a decade ago, the investigators could not draw definitive conclusions because of limited data 9 . Additional studies have since reported on the prevalence of adenomyosis in women with subfertility, yet there has been no recent review evaluating pooled prevalence estimates. The aim of this study was to provide a comprehensive up-to-date synthesis of the data on prevalence of isolated adenomyosis and adenomyosis occurring concurrently with endometriosis and/or fibroids in women with subfertility.

Protocol registration
The protocol of this systematic review was developed and registered prospectively with PROSPERO (CRD42021255140). The protocol outlined the search strategy, inclusion and exclusion criteria for study selection, quality assessment and strategy for data extraction and synthesis. We have reported the systematic review in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 checklist 31 .

Search strategy
A comprehensive and systematic search of the literature was performed on 4 November 2022 in the following databases: MEDLINE, EMBASE, CINAHL Plus, Google Scholar, PsycINFO and Web of Science Core Collection. We searched for publications in any language from inception until October 2022 using a combination of medical subject headings and the following keywords: 'prevalence' OR 'incidence' OR 'epidemiology' OR 'frequency' OR 'occurrence, adenomyosis' OR 'adenomyoma' OR 'adenomyosis uteri' OR 'endometrial adenoma' OR 'endometriosis interna' OR 'endometriosis, stroma' OR 'internal endometriosis' OR 'stroma endometriosis' OR 'stromal endometriosis' OR 'uterine adenomyomatosis' OR 'uterine adenomyosis' OR 'uterus adenomyosis, infertile' OR 'subfertile' OR 'fertile' (Table S1). The search strategy was reviewed by a medical sciences librarian. A manual search of the reference lists of selected articles was also performed to identify any missing papers with relevant data not identified by the electronic search.

Study selection and eligibility criteria
Two authors (I.M. and P.M.) screened independently the title and abstract of identified papers. Full-text articles were obtained after screening the abstracts that met the eligibility criteria of the study, and any disagreement was resolved by discussion until a consensus was reached. Cohort and cross-sectional studies examining the prevalence of adenomyosis in women with subfertility were included. Subfertility was defined as failure to establish a clinical pregnancy after 12 months of regular, unprotected sexual intercourse or impairment of a woman's capacity to reproduce as an individual or with her partner 32 . The term subfertility was used interchangeably with infertility. Subfertile women undergoing and those who were not undergoing ART treatment were included. We included articles without any language restrictions. Studies were included if the authors used any of the previously published and recognized ultrasound and magnetic resonance imaging (MRI) diagnostic features for the diagnosis of adenomyosis 2,[33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48][49] . A participant was considered to have adenomyosis when diagnosed on ultrasound, MRI or a combination of ultrasound and MRI. Diagnosis using ultrasound was made based on any of the direct (myometrial cysts, hyperechogenic islands, echogenic subendometrial lines and buds) and/or indirect (asymmetrical myometrial thickening, globular uterus, fan-shaped shadowing, translesional vascularity, irregular and interrupted junctional zone (JZ)) features of adenomyosis. MRI features included any of the following: maximum (JZmax) or average JZ thickness, JZmax-to-myometrial thickness ratio, high-signal-intensity myometrial spots and low-signal-intensity mass. We excluded studies in Systematic review on adenomyosis prevalence in subfertile women 25 which data were not available separately for women with isolated adenomyosis. Studies in which adenomyosis was diagnosed visually by hysteroscopy, laparoscopy and laparotomy were also excluded due to lack of validation of these tools for diagnosis of adenomyosis. Finally, we also excluded studies in which diagnostic criteria were not mentioned or could not be retrieved after contacting the investigators (Table S2). We planned to report data separately for women with isolated adenomyosis and for those with adenomyosis with concurrent endometriosis and/or fibroids because these pathologies are known to result in a significant variation in the prevalence of adenomyosis 7,30,50 .
The primary outcome was the prevalence of adenomyosis in women with subfertility. This was expressed as a proportion, with the numerator defined as the number of women with subfertility and adenomyosis with or without endometriosis and/or fibroids, while the denominator represented the total number of women with subfertility in the study group. Subgroup analyses were conducted to identify variation in the prevalence of isolated adenomyosis according to geographical location, diagnostic modality, diagnostic criteria, type of ultrasound, ultrasound features of adenomyosis and the use of ART.

Data extraction
Two reviewers (I.M. and P.M.) independently extracted study data, including title, first author and year of publication of the paper, country in which the study was conducted, study design, participant characteristics,

Statistical analysis
Data analysis was performed using Stata statistical software, release 17 (StataCorp. LLC, College Station, TX, USA). The prevalence of adenomyosis was reported as a proportion with 95% CI. The Stata Metaprop statistical command and double arcsine transformation (Freeman-Tukey transformation) were used to pool proportions from individual studies, allowing inclusion of studies with proportions of zero or one 53 . Heterogeneity was assessed graphically, using forest plots, and statistically, using the tau 2 and I 2 statistics. Effect estimates were pooled using a random-effects model to allow for differences in prevalence estimates between different studies. The tau 2 statistic represented the extent of variation in the prevalence observed in different studies (between-study variance) 54 . I 2 > 50% was considered indicative of substantial heterogeneity 54 .
The primary analysis included studies reporting on the prevalence of adenomyosis in women with subfertility. Secondary analyses were also conducted to identify variation in the prevalence of isolated adenomyosis according to geographical location, mode of diagnosis, diagnostic criteria, type of ultrasound, ultrasound features of adenomyosis and the use of ART. To determine the possible impact of these factors on the observed heterogeneity across studies, we conducted metaregression analysis within the above subgroups.

Study selection
The PRISMA flowchart outlines the study selection process in detail ( Figure 1 23 11/10 268 meet the inclusion criteria. The title and abstract of 891 articles were screened after exclusion of 371 duplicates. Following title and abstract screening, 76 records potentially meeting the eligibility criteria were included. We obtained the full text of 71 records, of which 33 studies were excluded because they did not meet the inclusion criteria. Of these, 23 studies had an inappropriate denominator (eight studies that assessed women without subfertility, 10 studies in which all women had endometriosis and five studies in which all women had adenomyosis), two studies had an inappropriate numerator (all women had endometriosis), one study did not report on the prevalence of adenomyosis and focused on pregnancy outcome instead, one study was a review article, one study excluded cases of adenomyosis and five studies used an inappropriate diagnostic modality (Table S2).
Additionally, 17 studies with missing information/data were excluded due to unanswered correspondence by the authors or lack of contact details (Table S3). When possible, we attempted to contact the authors of 26 studies to obtain relevant study details and missing data (  Figure 3 Forest plot showing prevalence of adenomyosis with concurrent endometriosis in women with subfertility. Only first author is given for each study. Random-effects model and Freeman-Tukey formula were used.
ultrasound 22,24,64,68 . Severity was defined differently in all four studies and was based on the number of diagnostic features identified 22,24,68 , the number of reviewers making the diagnosis of adenomyosis 22 and the effect of adenomyosis on the shape of the uterine cavity 64 . The number of features required for adenomyosis to be classified as severe varied as follows: two features 22 , three features 24 and all features 68 .

Overall prevalence of adenomyosis with coexisting fibroids
Prevalence estimates across eight studies ranged from 0% in women aged 23-40 years with subfertility in a prospective study by Silva et al. 63 to 5% in two studies, including a cross-sectional study of women with a mean age of 38 years undergoing ART treatment in Spain 64 and a prospective study of women attending a general gynecology clinic with subfertility in the UK 27 . The study of Silva et al. 63 applied MUSA criteria 48 to diagnose adenomyosis using ultrasound, whereas the study of Naftalin et al. 27 and Puente et al. 64 used the presence of at least one or more direct (myometrial cysts, hyperechogenic islands, echogenic subendometrial lines and buds) and/or indirect (asymmetrical myometrial thickening, globular uterus, fan-shaped shadowing, irregular and interrupted JZ) ultrasound features to diagnose adenomyosis. The overall random-effects pooled prevalence of adenomyosis with fibroids in women with subfertility was 1% (95% CI, 0-4%) with a high level of heterogeneity (I 2 = 95.8%, tau 2 = 0.03) ( Figure 2).

Overall prevalence of adenomyosis with coexisting endometriosis
Prevalence estimates across 18 studies ranged from 0% in women aged 23-40 years with subfertility in the study by Silva et al. 63 to 56% in women aged 17-46 years with subfertility in Germany 50 . Both studies were prospective; one study 63 applied sonographic MUSA criteria, while the other 50 used average JZ thickness of more than 10 mm on MRI 50 to diagnose adenomyosis. The overall random-effects pooled prevalence of adenomyosis with coexisting endometriosis in women with subfertility was 6% (95% CI, 3-11%) with a high level of heterogeneity (I 2 = 98.6%, tau 2 = 0.12) (Figure 3).

Overall prevalence of adenomyosis with coexisting endometriosis and/or fibroids
Prevalence estimates across nine studies ranged from 0% in a prospective cohort of women aged 23-40 years with subfertility in Portugal 63 to 35% in a prospective study of women aged 26-41 years presenting with dysmenorrhea, menorrhagia and subfertility in the UK 8 and in a cross-sectional study of women aged 18-42 years with subfertility and undergoing surgery for benign gynecological disease in France 56 . The study in Portugal 63 used well-defined MUSA criteria, whereas the other two studies 8,56 used MRI to diagnose adenomyosis. The MRI diagnostic criteria in terms of the cut-off for JZ thickness varied between the two studies. De Souza et al. 8 used increased thickness of the JZ (8-12 mm) and increased JZ-to-outer-myometrial ratio with or without high-signal foci to diagnose diffuse adenomyosis, and the presence of a localized ill-defined mixed-signal-intensity mass to diagnose focal adenomyosis. In contrast, Bourdon et al. 56 used the presence of increased thickness of the JZ ≥ 12 mm and JZmax/myometrial thickness > 40% with or without presence of high-signal-intensity myometrial spots to diagnose diffuse adenomyosis, and the presence of a low-signal-intensity mass with ill-defined margins to diagnose focal adenomyosis. The overall random-effects pooled prevalence of adenomyosis with coexisting endometriosis and/or fibroids in women with subfertility was 7% (95% CI, 2-13%) with a high level of interstudy heterogeneity (I 2 = 98.3%; tau 2 = 0.09) (Figure 4).

Overall prevalence of isolated adenomyosis
The prevalence of isolated adenomyosis across 21 included studies ranged from 0% in women aged 23-40 years with subfertility in the study of Silva et al. 63 to 29% in women aged 29-49 years undergoing oocyte recipient IVF cycles in Spain 61 . The study of Silva et al. 63 was prospective and used MUSA criteria to diagnose adenomyosis, whereas the study of Martínez-Conejero et al. 61 was retrospective and used ultrasound features of a

Figure 5
Forest plot showing prevalence of isolated adenomyosis in women with subfertility according to diagnostic modality. Only first author is given for each study. Random-effects model and Freeman-Tukey formula were used. I 2 and tau 2 were not generated for subgroups including fewer than four studies. MRI, magnetic resonance imaging.
hypoechogenic and heterogeneous myometrium associated with elliptic intramyometrial lakes of more than 2 mm in diameter in a globular-appearing uterus to diagnose adenomyosis. The mean age of women with a diagnosis of adenomyosis ranged from 32 years 14 to 41 years 61 . The overall random-effects pooled prevalence of isolated adenomyosis was 10% (95% CI, 6-15%) with a high level of heterogeneity (I 2 = 99.1%; tau 2 = 0.12). We performed subgroup analyses according to geographical area, diagnostic modality, diagnostic criteria, type of ultrasound, ultrasound features of adenomyosis and population of women with subfertility for the isolated adenomyosis group (Figures 5-10).

Prevalence of isolated adenomyosis according to diagnostic modality
The prevalence of adenomyosis according to diagnostic method ranged from 0% to 29% in studies using ultrasound, 4% to 19% in studies using MRI and 4% to 15% in studies using a combination of ultrasound and MRI. The pooled prevalence of isolated adenomyosis was 10% (95% CI, 5-16%) using ultrasound, 10% (95% CI, 6-14%) using MRI and 7% (95% CI, 5-9%) using a combination of ultrasound and MRI as a diagnostic tool ( Figure 5). We performed a metaregression analysis including pooled prevalence estimates for subgroups based on the mode of diagnosis. This suggested that the mode of diagnosis had little effect on the prevalence of adenomyosis (P-value ranging from 0.74 to 0.86) when comparing ultrasound with other diagnostic modalities (Table 3).
Metaregression analysis suggested that there was little effect of ultrasound diagnostic criteria (P = 0.85), type of ultrasound (P = 0.49) or direct/indirect features of adenomyosis (P = 0.13) on the prevalence of adenomyosis (Table 3).

MRI as diagnostic modality
Five studies used MRI as the diagnostic modality of choice 8,15,50,56,66 and two studies used MRI in combination with ultrasound to confirm adenomyosis diagnosed on ultrasound 65 or in uncertain cases 18 . The MRI criteria used for the diagnosis of adenomyosis were uniform across four studies 8,56,65,66 and were based on previously published criteria 71 . Adenomyosis was diagnosed based on increased JZmax and JZmax-to-myometrial-thickness ratio, with or without high-signal-intensity myometrial spots. Focal adenomyosis was diagnosed based on the presence of a low-/mixed-signal-intensity mass with ill-defined margins. One study used MRI features of JZ thickness ≥ 12 mm and presence of subendometrial cysts for diagnosis of adenomyosis 18 . Importantly, the diagnostic criteria varied for studies conducted more than a decade ago, in which diagnosis was based on the mean JZ thickness, defined as an average of three measurements of JZ in the midsagittal section of the uterus (anterior, posterior and fundal walls) 15,50 . This varied from > 7 mm 15 to > 10 mm 50 .
Of seven included MRI-based studies, the prevalence of isolated adenomyosis ranged from 4% to 19% across three studies in which MRI was performed by two radiologists with expertise in gynecological MRI 8,50,56 , and from 4% to 11% in two studies in which MRI was performed by one radiologist with expertise in gynecological imaging 15,65 . The prevalence was 14% in one study in which MRI data were rereviewed by the study investigator and three radiologists with expertise in pelvic MRI 66 . The operators were blinded to clinical data in three studies 15,50,56 and to previous imaging in three studies 8,56,65 . There was no operator information available for one study 18 .

Quality assessment of included studies
All included studies were assessed for methodological quality using the CASP 2018 checklist. We judged all included studies to be of good quality. The summary of the quality assessment of 21 included studies can be found in Table S4.
However, these findings may represent only the tip of the iceberg due to lack of appropriate use and awareness of standardized diagnostic criteria, leading to underdiagnosis. This is evident from the great difference noted in the prevalence of adenomyosis between the study using only one indirect sign 23 , in which the prevalence of isolated adenomyosis was 0.45%, and those using a combination of direct and indirect signs of adenomyosis, in which the pooled prevalence of isolated adenomyosis was 11% (95% CI, 7-16%) ( Figure 8).
Furthermore, the prevalence of isolated adenomyosis varied across different geographical locations, with Australia exhibiting the highest pooled prevalence of adenomyosis (19%), which was significantly higher compared with that in Asia (5%). This variation may be attributed to the following factors: a region-specific increase in adenomyosis in Australia, for unknown reasons; the use of a combination of multiple direct and indirect signs to diagnose adenomyosis; and the expertise of the operator performing the imaging assessment.
This systematic review confirms that transvaginal ultrasound remains the most widely available first-line diagnostic tool of choice for adenomyosis, followed by MRI 2 . The prevalence of isolated adenomyosis was 10% on both ultrasound and MRI and 7% using a combination of ultrasound and MRI. However, the diagnostic criteria varied across the included studies. The MUSA consensus published in 2015 provides a standardized terminology for describing ultrasound features associated with adenomyosis 48 . Although all included studies which started recruitment or analysis after the publication of MUSA guidelines adopted the MUSA criteria to diagnose adenomyosis, the number of features required to establish

Figure 10
Forest plot showing prevalence of isolated adenomyosis in women with subfertility according to type of population (undergoing vs those not undergoing assisted reproductive technology (ART)). Only first author is given for each study. Random-effects model and Freeman-Tukey formula were used. a diagnosis varied widely (Table 1) 22,24,59,63,65,67 . Though the MUSA criteria are based on a combination of 2D and 3D signs, two studies used only 2D features of MUSA criteria to diagnose adenomyosis 59,65 . All of this increases the risk of underdiagnosis.
In this review, variation was noted between the prevalence of isolated adenomyosis and that of adenomyosis with coexisting endometriosis and/or fibroids. Sonographic features of adenomyosis are highly prevalent in women with endometriosis 28 . Although there is a degree of pathophysiology and symptom overlap between the two conditions, they are different gynecological entities and often coexist [27][28][29][72][73][74] . However, this should be interpreted with caution because, in most of the included studies, it was difficult to gather information on whether endometriosis was confirmed histologically following surgical visualization.

Comparison with existing literature
The prevalence of sonographic signs of adenomyosis in women attending a gynecology clinic has been estimated to be 20.9% 27 . In this systematic review, the pooled prevalence of isolated adenomyosis in women with subfertility was 10%. The different rate observed in this study may be attributed to a different population of interest (women with subfertility), inclusion of both symptomatic and asymptomatic women and exclusion of endometriosis and fibroids.

Strengths and limitations
To our knowledge, this is the first systematic review to highlight the lack of cohesive data on the prevalence of adenomyosis in women with subfertility. We used a comprehensive search strategy of major bibliographic databases with no language restriction to maximize the global representativeness of data. This resulted in a large study population of 25 600 women across different continents. Furthermore, we employed robust methodology to analyze comprehensively our data. We determined the prevalence of adenomyosis in four clinically relevant groups and elaborated on the prevalence of isolated adenomyosis according to geographical location, diagnostic modality, diagnostic criteria, type of ultrasound, ultrasound features of adenomyosis and the use of ART.
One of the limitations of this review is the inherent heterogeneity of the included studies that were pooled in the meta-analysis of prevalence estimates. This stemmed from variation in participant characteristics, diagnostic modality and diagnostic criteria. Second, the quality assessment of the included studies was limited by the lack of a validated tool for assessing the methodological quality of prevalence studies. We used the CASP tool to reduce the subjectivity of the quality assessment 51 . The high heterogeneity of studies warrants caution when interpreting the pooled prevalence estimates. *Metaregression used to assess difference between subgroups and to determine P-values. I 2 and tau 2 were not generated for subgroups including fewer than four studies. 2D, two-dimensional; 3D, three-dimensional; MRI, magnetic resonance imaging; MUSA, morphological uterus sonographic assessment.

Implications for clinical practice
The variation in the prevalence of adenomyosis in various subgroups and the high heterogeneity between studies reflects the lack of use of standardized criteria by studies conducted prior to publication of the MUSA criteria and inappropriate use after publication. The inappropriate use of MUSA criteria by the included studies, which is evident from the variation in the number of MUSA features required to establish the diagnosis of adenomyosis 22,24,59,63,65,67 , increases the risk of underdiagnosis due to potentially missed cases of mild adenomyosis. Appropriate training of clinicians and sonographers to diagnose adenomyosis using standardized criteria should be part of a basic gynecological ultrasound curriculum across the world. Identification of adenomyosis using a uniform ultrasound set of criteria would provide insight into the true burden of this condition.

Implications for research
There is a need for future studies to use standardized and uniform diagnostic criteria (MUSA ultrasound consensus criteria) to diagnose adenomyosis. This would not only reduce interobserver variability, but also mitigate interstudy heterogeneity in future meta-analyses. It would also enable robust and uniform reporting of data on the burden of adenomyosis. The impact of adenomyosis on pregnancy outcome can be evaluated appropriately only if the disease is correctly and uniformly identified; otherwise, the evidence regarding this association may remain conflicting due to variation in the denominator.

Conclusion
One in 10 women with subfertility have a diagnosis of isolated adenomyosis. The prevalence of adenomyosis varies according to the presence of concurrent endometriosis and/or fibroids.

SUPPORTING INFORMATION ON THE INTERNET
The following supporting information may be found in the online version of this article: