The association between endometriosis and migraine: a systematic review and meta-analysis of observational studies

This is a systematic review and meta-analysis performed according to the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) [ 23 ] and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines [ 24 ]. The study protocol was prospectively registered on PROSPERO (CRD42023449492). Ethical approval and informed consent were not needed in accordance with 45CFR§ 46. The following databases were searched from inception up to July 22nd, 2024: Cochrane Central Register of Controlled Trials (CENTRAL), Ovid MEDLINE and Embase via OvidSP, and CINAHL. ClinicalTrials.gov was used to search for ongoing and registered trials; OpenGrey and Google.com were used to search for grey literature. The electronic search algorithm consisted of MeSH terms “endometriosis”, “headache”, and “migraine” with all subheadings explored; the terms were then combined with “and”. The full search strategy is presented in Supplementary Appendix. Reference lists of relevant articles were backwards searched to identify papers not captured by the electronic searches. No restrictions were applied to publication date or language. The literature search and article eligibility were independently assessed by two authors (G.E.C. and S.M.), with any disagreements resolved through discussion with a third author (D.R.K.). The PICOS framework was used to assess study eligibility as follows: (i) Population: individuals with endometriosis, diagnosed by surgery, imaging, clinical assessment, International Classification of Disease (ICD) codes, or self-report [ 25 ]; (ii) Intervention/Exposure: migraine, diagnosed by clinical assessment by a neurologist, ICD codes, questionnaire, interview, or self-report, with the preferred diagnostic method being the application of the International Classification of Headache Disorders, 3rd edition criteria created by the Headache Classification Committee of the International Headache Society (IHS) [ 26 ]; (iii) Comparator: individuals without a diagnosis of endometriosis and with no chronic pelvic pain, although no specific exclusion of endometriosis was required; (iv) Outcome: pooled risk of migraine in endometriosis compared to controls; Study type: peer-reviewed cohort, case-control, or cross-sectional studies. Case reports, case series, or any studies without a comparison population (control group), animal studies, conference abstracts, review articles, editorials, or commentaries were excluded. In the case of multiple publications from the same source population, only the most recent publication was included to avoid amplifying the effect of the same cohort on the pooled estimate. Data were extracted by two authors (G.E.C. and S.M.). The following data were collected and tabulated: first author, publication year, study country, study design, study period, sample size (number of endometriosis cases and number of controls), frequency of migraine in cases and controls, diagnostic modalities for endometriosis and migraine, baseline characteristics of study participants (including mean age, body mass index (BMI), ethnicity, cigarette smoking status, co-morbidities, parity, and infertility history), menopausal status, treatment of endometriosis and migraine, confounding factors adjusted for in the analysis. Additionally, we collected data on the localization and stage of endometriosis lesions in accordance with the revised American Society for Reproductive Medicine (rASRM) classification [ 27 ], categorizing stages I-II as minimal-to-mild disease, and stages III-IV as moderate-to-severe disease [ 27 ]. When data were missing or unclear, the authors of the included studies were contacted for clarification. The Risk Of Bias In Non-randomized Studies - of Exposure (ROBINS-E) tool [ 28 ] from the updated Cochrane collaboration guidelines was utilized for the assessment of methodological quality of included studies. Seven domains of bias were assessed, including:1) confounding, 2) measurement of the exposure, 3) the selection of participants into a study, 4) post-exposure interventions, 5) missing data, 6) measurement of the outcome, and 7) selection of the reported result. In the domain of bias due to confounding, a study was rated as “very high” risk of bias if it did not report any effect estimates, “high” risk of bias if it reported unadjusted risk estimates, “moderate” risk of bias if it reported effect estimates adjusted for a single confounding factor, and “low” risk of bias if it reported effect estimates adjusted for multiple confounding factors. A study was considered at low risk of bias if it was rated ‘low’ in all domains; a rating of some concerns if it was rated as ‘some concerns’ in one domain; high risk of bias if rated as ‘high’ in one domain, and very high risk of bias if rated as ‘very high’ in one domain [ 28 ]. Each study was assessed independently by two authors (G.E.C. and S.M.); when disagreement occurred, consensus was achieved via discussion. The certainty of the evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach [ 29 ]. Guidance from the Grade Handbook for grading quality of evidence and strength of recommendations [ 30 ] was utilized to make the assessment, and the GradePRO Guideline Development Tool software [ 31 ] was utilized to create a Summary of Findings table. Starting from a certainty of “high” for randomized controlled studies, GRADE allows for downgrading if one or more of five domains suggests limitations in the certainty of the evidence (risk of bias, imprecision, inconsistency, indirectness, and publication bias). For nonrandomized studies, the initial certainty starts at low, then five downgrading domains and three upgrading domains (large effect, dose–response gradient, and plausible confounding) are assessed. Original data on binary outcome measures were extrapolated from included studies so that Log Odds ratios (ORs) with 95% Confidence Intervals (95% CI) were computed and pooled together. A random effects model was applied to estimate the pooled effect size for every investigated outcome [ 32 ]. The overall effect size was estimated as the weighted average of study-specific effect sizes, with larger studies having greater weights and smaller studies having lesser weights. Forest plots were used to present study-specific and overall effect sizes together with 95% CI. Between-study heterogeneity was represented by the I 2 index which estimates the percentage of total variation across studies that is due to between-study heterogeneity rather than due to sampling variation. I 2 index values were interpreted as follows: a value of up to 50% was considered low heterogeneity, 51–75% moderate heterogeneity, and > 75% high heterogeneity [ 33 ]. Publication bias was identified using the linear regression-based method according to Egger et al. [ 34 ]. Meta-analysis was conducted using Review Manager (RevMan) for Mac (Version 5.3) [ 35 ]. The primary analysis to answer the research question was obtained by comparing pooled data on the risk of migraine in endometriosis patients versus controls. Secondary outcomes included the risk of migraine with aura in individuals with endometriosis compared to controls. Additionally, we planned to provide pooled risk estimates according to endometriosis disease stage (by rASRM) and localization (deep lesions, ovarian endometriomas and superficial endometriosis), as outlined in the PROSPERO study protocol. Sensitivity analyses were performed to assess the robustness of the results, accounting for the ROBINS-E risk of bias rating, any adjustment for confounding factors (e.g. age, BMI) and for hormonal therapy. Subgroup analyses were implemented to explore potential sources of heterogeneity in the estimates, considering the endometriosis diagnostic method (laparoscopic or non-laparoscopic diagnosis), and migraine diagnostic method (neurologist, ICD codes, questionnaire, interview, self-report). All sensitivity and subgroup analyses were a priori planned; however, an additional sensitivity analysis not outlined in the PROSPERO study protocol was conducted for the specific confounder of hormonal therapy, given the availability of sufficient data for pooling.

The screening process identified 1,536 records, of which 220 were duplicates, leaving 1,316 records to be screened. Following title and abstract screening, 1,283 records did not meet the inclusion and exclusion criteria, and 33 studies were eligible for full-text screening (Table S1 ). Fourteen studies were included in the qualitative synthesis [ 4 – 7 , 10 , 15 , 36 – 43 ]. Thirteen studies were included in the meta-analysis, accounting for a total of 331,655 individuals (32,489 with endometriosis vs. 299,166 controls). The PRISMA flow diagram showing the process of literature selection is shown in Fig.  1 . Fig. 1 PRISMA Flowchart of study selection. Abbreviations : CENTRAL Cochrane Central Register of Controlled Trials; ICTRP International Clinical Trials Registry Platform; PRISMA Preferred Reporting Items for Systematic reviews and Meta-Analyses PRISMA Flowchart of study selection. Abbreviations : CENTRAL Cochrane Central Register of Controlled Trials; ICTRP International Clinical Trials Registry Platform; PRISMA Preferred Reporting Items for Systematic reviews and Meta-Analyses Of the fourteen studies included in the qualitative synthesis, most were cross-sectional ( n  = 6), with the remaining being case-control ( n  = 4), retrospective cohort ( n  = 3), or prospective cohort ( n  = 1) studies. Most of the included studies ( n  = 5) were conducted in the United States. The baseline characteristics are summarized in Table  1 . Table 1 Baseline characteristics of included studies Author, year, location Design Number of endometriosis cases Endometriosis diagnostic method Number of controls Migraine diagnostic method Result N (%) with migraine Adjustment factors Ferrero et al. 2004 , Italy [ 7 ] Cross-sectional 133 Laparoscopy and histology 166 Clinical by neurologist Exposure: 51 (38.3%) Control: 25 (15.1%)* Migraine without aura : Exposure: 33 (24.8%) Control: 23 (13.9%) Migraine with aura : Exposure: 18 (13.5%) Control: 2 (1.2%) N/A Gete et al. 2023 , Australia [ 42 ] Prospective cohort 1,149 Self-reported 6,457 Self-reported Exposure: 682 (59%) Control: 3,312 (51%)* Age, residence, marital status, education, income, smoking, alcohol intake, physical activity, BMI, parity, use of hormonal therapy Karamustafaoglu et al. 2019 , Turkey [ 44 ] Cross-sectional 114 Laparoscopy 97 Questionnaire Exposure: 51 (69%) Control: 26 (43%)* N/A Maitrot et al. 2020 , France [ 49 ] Case-control 182 Laparoscopy and histology 132 Questionnaire Exposure: 64 (35.2%) Control: 23 (17.4%)* Migraine without aura : Exposure: 55 (30.2%) Control: 17 (12.9%) Migraine with aura : Exposure: 9 (4.9%) Control: 6 (4.5%) Age, BMI, and family history of migraine Miller et al. 2018 , USA [ 5 ] Cross-sectional 296 Laparoscopy 95 Self-reported Exposure: 205 (87.2%) Control: 30 (12.8%) Age, race, use of hormonal therapy Mirkin et al. 2007 , USA [ 41 ] Retrospective cohort 40,150 MY ICD codes 7,144,896 MY ICD codes Exposure: 8.9% Control: 4%* N/A Nyholt et al. 2009 , Australia [ 40 ] Cross-sectional 767 Self-reported 164 Telephone interview Exposure: 285 (37%) Control: 50 (30%) N/A Stratton et al. 2015 , USA [ 45 ] Cross-sectional 18 Laparoscopy and histology 20 Structured interview Exposure: 11 (61%) Control: 0* N/A Sultana et al. 2024 , Bangladesh [ 43 ] Case-control 190 Laparoscopy or laparotomy 190 Clinical by neurologist Exposure: 27 (14.2%) Control: 5 (2.6%)* Age, BMI Tietjen et al. 2007 , USA [ 46 ] Cross-sectional 46 Self-report of laparoscopy 229 Clinical by neurologist Exposure: 36 (78.3%) Control: 127 (55.5%) N/A Tietjen et al. 2006 , USA [ 4 ] Case-control 16 Self-report of laparoscopy 86 Clinical by neurologist Exposure: 14 (87.5%) Control: 36 (41.9%) N/A Wu et al. 2018 , Taiwan [ 47 ] Retrospective cohort 9,191 ICD codes 27,573 ICD codes Exposure: 398 (4.3%) Control: 36 (41.9%)** N/A Wu et al. 2022 , China [ 48 ] Case-control 167 Laparoscopy 190 Clinical by neurologist Exposure: 50 (29.9%) Control: 23 (12.1%)* Migraine without aura : Exposure: 49 (29.3%) Control: 23 (12.1%) Migraine with aura : Exposure: 1 (0.6%) Control: 0 N/A Yang et al. 2012 , Taiwan [ 6 ] Retrospective cohort 20,220 ICD codes 263,767 ICD codes Exposure: 985 (4.9%) Control: 7,701 (2.9%)* Age, use of hormonal therapy N/A – not applicable; NR – not reported, OR – odds ratio; BMI – Body mass index; ICD – International Classification of Disease; MY – member years (total days of eligibility of an individual divided by 365) * p  < 0.01; ** p  < 0.05 Baseline characteristics of included studies Exposure: 51 (38.3%) Control: 25 (15.1%)* Migraine without aura : Exposure: 33 (24.8%) Control: 23 (13.9%) Migraine with aura : Exposure: 18 (13.5%) Control: 2 (1.2%) Exposure: 682 (59%) Control: 3,312 (51%)* Exposure: 51 (69%) Control: 26 (43%)* Exposure: 64 (35.2%) Control: 23 (17.4%)* Migraine without aura : Exposure: 55 (30.2%) Control: 17 (12.9%) Migraine with aura : Exposure: 9 (4.9%) Control: 6 (4.5%) Exposure: 205 (87.2%) Control: 30 (12.8%) Exposure: 8.9% Control: 4%* Exposure: 285 (37%) Control: 50 (30%) Exposure: 11 (61%) Control: 0* Exposure: 27 (14.2%) Control: 5 (2.6%)* Exposure: 36 (78.3%) Control: 127 (55.5%) Exposure: 14 (87.5%) Control: 36 (41.9%) Exposure: 398 (4.3%) Control: 36 (41.9%)** Exposure: 50 (29.9%) Control: 23 (12.1%)* Migraine without aura : Exposure: 49 (29.3%) Control: 23 (12.1%) Migraine with aura : Exposure: 1 (0.6%) Control: 0 Exposure: 985 (4.9%) Control: 7,701 (2.9%)* N/A – not applicable; NR – not reported, OR – odds ratio; BMI – Body mass index; ICD – International Classification of Disease; MY – member years (total days of eligibility of an individual divided by 365) * p  < 0.01; ** p  < 0.05 The population size of the included studies was highly variable, with n  = 283,987 individuals derived from the largest study [ 6 ] and n  = 38 individuals included in the smallest [ 40 ]. In half of the included studies the diagnosis of endometriosis was based on laparoscopic confirmation of the disease [ 5 , 7 , 10 , 38 – 40 , 43 ]; four studies (28.5%) diagnosed endometriosis via self-report [ 4 , 15 , 37 , 41 ], and the remaining three studies (21.5%) diagnosed endometriosis via ICD codes [ 6 , 36 , 42 ]. Only Wu et al., 2022 [ 43 ] reported data on endometriosis severity according to the r-ASRM staging system [ 27 ]. Four of the included studies [ 7 , 10 , 38 , 43 ] performed laparoscopy to rule out endometriosis in controls. Migraine diagnostic method was highly heterogeneous, with 5 studies achieving diagnosis via interview with a neurologist applying IHS criteria, 3 studies adopting ICD codes in insurance databases, 2 studies via questionnaires applying IHS criteria, 2 studies via self-reported diagnosis, one study via telephone interview applying IHS criteria, and one study via self-report and clinical interview. Three studies provided risk estimates adjusted for use of hormonal therapy [ 5 , 6 , 37 ]. Since Tietjen et al. 2006 [ 4 ] and Tietjen et al. 2007 [ 41 ] were conducted one year apart, the author was contacted and confirmed that the study populations were distinct. The risk of bias assessment found a serious risk of bias in the majority of the included studies, with 50% of studies ( n  = 7) being at very high risk of bias, 21% ( n  = 3) at high risk of bias, 14.5% of studies ( n  = 2) rated as ‘some concerns’, and 14.5% ( n  = 2) judged as at low risk of bias (Figure S1 ). Bias due to confounding was the section most often ranked as very high risk. Nyholt et al., 2009 [ 15 ] was ranked as high risk of bias in three domains: confounding, exposure assessment, and missing data. Mirkin et al., 2007 [ 36 ] and Tietjen et al., 2007 [ 41 ] were ranked as very high risk of bias due to confounding and high risk of bias due to exposure assessment. Gete et al., 2023 [ 37 ] was ranked as very high risk of bias in selection of participants into the study and high risk of bias in selection of the reported result. No other studies were ranked as very high or high risk of bias in more than one category. Meta-analysis of the 13 included studies revealed a greater risk of migraine in individuals with endometriosis compared to controls (OR 2.25; 95% CI = 1.85–2.72; n  = 13 studies) with high heterogeneity (I 2  = 81%) and non-significant publication bias (Egger’s, p  = 0.83) (Fig.  2 ). The certainty of the evidence according to GRADE was graded as Low (Table  2 ): downgraded two levels due to very high risk of bias, downgraded one level due to indirectness (multiple diagnostic methods were accepted, the population settings varied, and not all studies excluded endometriosis in the control group), and upgraded one level for strong association (OR > 2). Fig. 2 Forest plot for the association between endometriosis and migraine diagnosis Forest plot for the association between endometriosis and migraine diagnosis Table 2 Assessment of certainty of evidence using the GRADE evidence profile Certainty assessment No of patients Effect Certainty No of studies Study design Risk of bias Inconsistency Indirectness Imprecision Other considerations endometriosis control Relative (95% CI) Absolute (95% CI) Migraine 13 Observational studies very serious a not serious serious b not serious strong association 2,859/32,489 (8.8%) 12,030/299,166 (4.0%) OR 2.25 (1.85 to 2.72) 46 more per 1 , 000 (from 32 more to 62 more) ⨁⨁◯◯ Low a, b Migraine without aura 3 Observational studies very serious c not serious serious b not serious strong association 137/482 (28.4%) 63/488 (12.9%) OR 2.64 (1.89 to 3.69) 152 more per 1 , 000 (from 90 more to 224 more) ⨁◯◯◯ Very low b, c Migraine with aura 3 Observational studies very serious c serious d serious b very serious e limited evidence, heterogeneity of findings 28/482 (5.8%) 8/488 (1.6%) OR 3.47 (0.53 to 22.89) 37 more per 1 , 000 (from 7 fewer to 255 more) ⨁◯◯◯ Very low b, c,d, e CI : confidence interval; OR : odds ratio Explanations a. 50% of studies ( n  = 7) at very high risk of bias, 21% ( n  = 3) at high risk of bias, 14.5% of studies ( n  = 3) rated as ‘some concerns’, and 14.5% ( n  = 3) at low risk of bias b. Multiple diagnostic methods accepted, various population settings (e.g. neurology clinic), not all studies excluded endometriosis in the control group (e.g. negative laparoscopy) c. 67% ( n  = 2) rated as ‘some concerns’ and 33% ( n  = 1) at very high risk of bias d. One study found similar rates in both groups e. 95% CI 0.53–22.89 Assessment of certainty of evidence using the GRADE evidence profile OR 2.25 (1.85 to 2.72) 46 more per 1 , 000 (from 32 more to 62 more) ⨁⨁◯◯ Low a, b 63/488 (12.9%) OR 2.64 (1.89 to 3.69) 152 more per 1 , 000 (from 90 more to 224 more) ⨁◯◯◯ Very low b, c 28/482 (5.8%) 8/488 (1.6%) OR 3.47 (0.53 to 22.89) 37 more per 1 , 000 (from 7 fewer to 255 more) ⨁◯◯◯ Very low b, c,d, e CI : confidence interval; OR : odds ratio Explanations a. 50% of studies ( n  = 7) at very high risk of bias, 21% ( n  = 3) at high risk of bias, 14.5% of studies ( n  = 3) rated as ‘some concerns’, and 14.5% ( n  = 3) at low risk of bias b. Multiple diagnostic methods accepted, various population settings (e.g. neurology clinic), not all studies excluded endometriosis in the control group (e.g. negative laparoscopy) c. 67% ( n  = 2) rated as ‘some concerns’ and 33% ( n  = 1) at very high risk of bias d. One study found similar rates in both groups e. 95% CI 0.53–22.89 Sensitivity analyses excluding studies at very high or high risk of bias by the ROBINS-E assessment confirmed that the risk of migraine was higher in endometriosis compared to controls (OR 2.64; 95% CI = 1.62–4.31; n  = 4 studies; I 2  = 77%) (Figure S2 ). The Egger’s test for this analysis showed some publication bias ( p  = 0.03). Sensitivity analysis including only studies which adjusted for clinically relevant confounders (full list of confounders adjusted for in each study is provided in Table  1 ) also confirmed an association between the two conditions (OR 2.35; 95% CI = 1.77–3.13; n  = 6 studies; I 2  = 88%), despite some publication bias (Egger’s, p  = 0.04) (Figure S3 ). Sensitivity analysis including only studies which adjusted for hormonal therapy use also confirmed this association (OR 1.95; 95% CI = 1.42–2.66; n  = 3 studies; I 2  = 92%), with non-significant publication bias (Egger’s, p  = 0.72) (Figure S4 ). Subgroup analysis based on endometriosis diagnostic method showed a greater association in the studies that diagnosed endometriosis via laparoscopy in comparison to those with a non-surgical diagnosis: OR 3.48 (95% CI = 2.50–4.83; n  = 7; I 2  = 43%) vs. OR 1.65 (95% CI = 1.42–1.92; n  = 6; I 2  = 73%) (Figure S5 ). Non-significant publication bias was found for both studies based on surgical or non-surgical diagnosis (Egger’s, p  = 0.28 and p  = 0.66, respectively). When considering migraine diagnostic method, pooled risk estimates were higher when migraine was diagnosed by a neurologist (OR 3.38; 95% CI = 2.42–4.73; n  = 4; I 2  = 0%), with non-significant publication bias (Egger’s, p  = 0.71) (Figure S6 ). Three studies could be included for this specific comparison [ 7 , 10 , 43 ], accounting for a total of 970 individuals (482 with endometriosis vs. 488 controls). Maitrot et al., 2020 [ 10 ] and Wu et al., 2022 [ 43 ] were rated as having “some concerns”, and Ferrero et al., 2004 [ 7 ] was rated at “very high” risk of bias (Figure S1 ). Meta-analysis of the included studies revealed a greater risk of migraine without aura in individuals with endometriosis compared to controls (OR 2.64, 95% CI 1.89–3.69; n  = 3 studies; I 2  = 0%) (Fig.  3 ). There was no significant publication bias (Egger’s, p  = 0.67). Fig. 3 Forest plot for the association between endometriosis and migraine without aura diagnosis Forest plot for the association between endometriosis and migraine without aura diagnosis The certainty of the evidence according to GRADE was graded as Very Low (Table  2 ): downgraded two levels due to very high risk of bias, downgraded one level due to indirectness (multiple diagnostic methods were accepted, the population settings varied, and not all studies excluded endometriosis in the control group), downgraded one level due to small number of studies/population, and upgraded one level for strong association (OR > 2). Only three studies could be included in this comparison [ 7 , 10 , 43 ], accounting for a total of 970 individuals (482 with endometriosis vs. 488 controls). Maitrot et al., 2020 [ 10 ] and Wu et al., 2022 [ 43 ] were rated as having “some concerns”, and Ferrero et al., 2004 [ 7 ] was rated at “very high” risk of bias (Figure S1 ). Meta-analysis of the included studies showed a substantial uncertainty in the association between endometriosis and migraine with aura, with a very wide confidence interval (OR 3.47, 95% CI = 0.53–22.89; n  = 3) and high heterogeneity (I 2  = 73%). There was no significant publication bias (Egger’s, p  = 0.66) (Fig.  4 ). Fig. 4 Forest plot for the association between endometriosis and migraine with aura diagnosis Forest plot for the association between endometriosis and migraine with aura diagnosis The certainty of the evidence according to GRADE was graded as Very Low (Table  2 ): downgraded two levels due to very high risk of bias, downgraded one level due to indirectness (multiple diagnostic methods were accepted, the population settings varied, and not all studies excluded endometriosis in the control group), and downgraded two levels due to imprecision (wide confidence interval crossing 0), and downgraded one level due to small number of studies/population. Only one study reported on the risk of migraine according to endometriosis severity, and one according to lesion localization. Therefore, meta-analysis could not be performed. Wu et al., 2022 [ 43 ] reported rASRM stage and found that a diagnosis of migraine was more than four times more likely in individuals with moderate-severe endometriosis (rASRM stages III-IV) compared to those without endometriosis (OR 4.52; 95% CI = 2.49–8.20), whereas the association between migraine and minimal-mild endometriosis (rASRM stages I-II) was not statistically significant (OR 1.4; 95% CI = 0.64–3.05). Maitrot et al., 2020 [ 10 ] reported disease phenotypes: superficial peritoneal endometriosis, ovarian endometrioma, and deep infiltrating endometriosis. The risk of migraine was significantly increased in individuals with endometriomas (OR 2.78; 95% CI = 1.11–6.98) and deep infiltrating endometriosis (OR 2.51; 95% CI 1.25–5.07) in comparison to controls without endometriosis. The risk of migraine was not increased with superficial peritoneal endometriosis (OR 1.97; 95% CI = 0.88–4.40) in comparison to controls without endometriosis.

This systematic review and meta-analysis of 331,655 individuals across 13 studies confirms the association between endometriosis and migraine (Fig.  2 ), although the certainty of evidence was low (Table  2 ). When considering the secondary outcomes, meta-analysis revealed an increased risk of migraine without aura in individuals with endometriosis (Fig.  3 ). However, the association between endometriosis and the risk of migraine with aura was uncertain (Fig.  4 ). Further secondary outcomes explored included the association between migraine and endometriosis by severity or localization. In the single study that reported rASRM stage [ 43 ], migraine was associated with moderate-severe endometriosis. Similarly, in the single study addressing disease localization [ 10 ], ovarian endometriosis and deep lesions were associated with migraine, whereas superficial endometriosis was not. Given the importance of accounting for confounding factors, we conducted sensitivity analyses including only studies that reported adjusted effect estimates for clinically relevant confounders (Figure S3 ), including only studies rated as “some concerns” or “low” risk of bias (Figure S2 ), and pooling adjusted risk estimates for hormonal therapy use (Figure S4 ). All sensitivity analyses demonstrated a robust and consistent association between endometriosis and migraine. Furthermore, subgroup analysis confirmed that this association was stronger when endometriosis was diagnosed via laparoscopy (Figure S5 ) and migraine was confirmed through neurological assessment (Figure S6 ). The potential biological mechanisms that could underlie the association between endometriosis and migraine include hormonal milieu, inflammation, and genetic predisposition. The hormonal milieu of the menstrual cycle may influence migraine through the ability of estrogen and progesterone to modify neurotransmitter systems linked with migraine pathogenesis [ 44 – 47 ]. Estrogen withdrawal has been suggested to trigger migraine [ 22 ]. Indeed, women with migraine experience a more rapid decline in estrogen levels during the late luteal phase of the menstrual cycle compared to controls, leading to an increased frequency of migraine attacks in the menstrual period [ 48 ]. Notably, the frequency of migraine attacks is higher during perimenopause, a time characterized by more frequent anovulatory cycles and fluctuating, lower estrogen levels [ 47 , 49 , 50 ]. While the precise interaction between female sex hormones and migraine remains to be fully defined, available data suggest an association. In this context, it is well-established that estrogens play a crucial role in endometriosis [ 51 ], promoting the growth of the ectopic endometrium that defines the disease. Therefore, the alterations in estrogen signaling observed in endometriosis may facilitate mechanisms underlying migraine [ 51 ]. When considering potential therapeutic implications, amenorrhea induced by first-line hormonal treatments has been shown to reduce menstrual-related migraine attacks triggered by hormonal withdrawal [ 52 ]. Moreover, in individuals with endometriosis and migraine without aura, continuous regimens with progestin-only pills (POPs) have been suggested to be highly effective in reducing migraine attacks [ 53 ], further supporting the key role of achieving amenorrhea. On the other hand, in the sensitivity analysis including only studies which adjusted for hormonal therapy the association between endometriosis and migraine was maintained. In addition to the role of estrogens, the altered peritoneal environment in women with endometriosis triggers a robust inflammatory response, characterized by elevated levels of prostaglandins, cytokines, and chemokines [ 51 , 54 ]. This inflammatory milieu appears to operate not only within the peritoneal cavity, contributing to the pathophysiology of pelvic endometriosis lesions, but also at a systemic level, partially underpinning its association with systemic immune dysregulatory and inflammatory conditions [ 55 , 56 ]. Emerging evidence highlights the importance of neuro-inflammation in endometriosis, driven by interactions between nerve fibers and immune cells that trigger peripheral and central pain pathways [ 54 ]. Migraine shares similar mechanisms, with local inflammation triggering meningeal nociceptors and activating the trigeminovascular system, processes that are thought to drive migraine pain [ 57 ]. Furthermore, systemic inflammation appears to play a key role in migraine as in endometriosis, with elevated levels of serum inflammatory markers in individuals with migraine compared to controls [ 58 ]. In the context of endometriosis and migraine, the inflammatory and neuro-inflammatory processes associated with both conditions are likely to sensitize central pain pathways, predisposing individuals to the development of nociplastic pain—a type of pain arising from altered nociception, characterized by heightened central nervous system responsiveness to sensory input [ 59 ]. This heightened sensitivity amplifies pain perception, even in the absence of ongoing noxious stimuli, and contributes to widespread pain originating from multiple body sites. The concept of chronic overlapping pain conditions provides a useful framework for understanding this association. Chronic overlapping pain conditions refer to the coexistence of multiple pain disorders, such as migraine and endometriosis, in the same individuals, underpinned by shared mechanisms such as central sensitization, inflammation, and altered nociceptive processing [ 60 ]. Finally, there is growing evidence of a genetic link between endometriosis and migraine. An extensive investigation of genome-wide association study (GWAS) data identified multiple lead single nucleotide polymorphisms associated to both conditions [ 14 ]. Further support to this genetic overlap was confirmed in another GWAS analysis, which identified multiple loci linked to both conditions, though no causal relationship was established [ 13 ]. At the individual level, a study conducted in a twin population found a significant additive genetic correlation between endometriosis and migraine [ 15 ]. These findings suggest the existence of shared genetically controlled biological mechanisms between endometriosis and migraine. Overall, it is quite challenging to infer which is exactly the biological mechanism underlying the association between the two diseases based on the available epidemiological evidence, primarily due to the scarcity of data pertaining to specific subgroups or affected populations. However, some plausible biological explanations supporting causality do exist. This meta-analysis adopted strict inclusion and exclusion criteria and was conducted with a rigorous methodology. Compared to the previous meta-analysis on this topic [ 3 ], it included a more extensive population, with 14 studies included in comparison to 9 in the previous review. This review performed multiple subgroup and sensitivity analyses to thoroughly investigate the association between endometriosis and migraine, whereas the previous review only conducted one subgroup analysis according to study design (i.e. cohort and case-control vs. cross-sectional). Additionally, although there was some overlap of studies included in both reviews, the previous review found that the records included were high quality, whereas a systematic assessment in this review found a high risk of bias, indicating stricter criteria and robust methodology in this review compared to the previous one [ 3 ]. However, several important limitations should be considered when interpreting the results. Notably, the majority of the included studies were rated as having a high or very high risk of bias. To address this limitation, we conducted a sensitivity analysis including only studies rated as having ‘some concerns’ or ‘low risk’ in the risk of bias assessment. The association between endometriosis and migraine remained significant. The principal cause of the high risk of bias was confounding, as few studies performed analyses adjusting for potential confounders. We tackled this issue through a separate sensitivity analysis including only studies that adjusted for confounders, and again the association persisted. Results not adjusted for confounding of potential migraine risk factors are less robust, as they could contribute to the increased risk of migraine if the exposure and control groups have baseline differences. The findings of this review may have been influenced by this high risk of bias and lack of adjustment for confounders, but it is unclear whether the direction of the bias would make the association between endometriosis and migraine stronger or less strong. Another key limitation was the high heterogeneity of pooled estimates, reflecting substantial variability across studies and emphasizing the need for caution in interpreting these findings. Additionally, in both the sensitivity analyses excluding studies at high risk of bias and including only studies adjusting for confounders, the Egger’s test revealed significant publication bias. This suggests that studies reporting stronger associations may have been more likely to be published, potentially leading to an overestimation of the association between the two conditions. Among the studies that included adjustment factors, only two considered hormonal therapy in their analyses, and just one accounted for surgical treatment of endometriosis. Both factors significantly influence the clinical manifestation of endometriosis and could therefore impact its association with migraine, making them important potential confounders. Another limitation relates to diagnostic methods. As outlined in the PROSPERO protocol (CRD42023449492), all diagnostic methods for endometriosis and migraine were accepted in order to gain the broadest possible overview of the available literature. Some of the included studies relied on self-reported diagnoses, which, while accounted for in the risk of bias assessment, may have introduced misclassification bias. To address this, a subgroup analysis based on endometriosis diagnostic method was conducted, which showed a greater association between endometriosis diagnosed via laparoscopy and migraine than other endometriosis diagnostic methods. However, the finding of a stronger association with migraine in individuals with endometriosis diagnosed via laparoscopy could be due to selection bias, another potential limitation to consider. While informative clinical insights could potentially be gained from stratifying migraine risk according to endometriosis severity, this analysis was not feasible, as only one study [ 43 ] reported data on endometriosis stage according to the ASRM classification. Therefore, conclusions on this aspect should be drawn with caution. The variations in study designs, condition frequencies, diagnostic methods or unaccounted confounding factors across comparisons may introduce significant clinical heterogeneity. Importantly, although the results of this review suggest a link between the two conditions and a biological plausibility can be proposed, no causal relationship between endometriosis and migraine has been established, and indeed the direction of this association is unclear. We cannot exclude that shared underlying risk factors might simply explain the association. Whilst this review suggests an association between endometriosis and migraine, further research is needed to establish this link with greater certainty. Future research should prioritize large, high-quality prospective studies with standardized diagnostic criteria and adjustment for confounding factors. Additionally, exploring whether the association changes based on endometriosis severity or migraine severity would provide new insights. Apart from the severity of the condition, the types of migraine, particularly migraine with and without aura, should investigated in future studies. The finding of publication bias in two sensitivity analyses indicates the possibility that not all research is being published, and studies which find no association between the conditions are also essential to understand this link.

This systematic review and meta-analysis suggest an association between endometriosis and migraine, although given risk of bias, substantial heterogeneity among studies, residual confounding, and evidence of publication bias, these results should be interpreted with caution. Recognizing the overlap between these conditions may have implications for clinical practice, particularly in screening, diagnosis, treatment selection, and patient education. Despite the limitations, our findings underscore a noteworthy link that warrants further investigation. Future research should focus on larger, methodologically rigorous studies designed to account for clinically relevant confounders, ensuring a more precise understanding of the relationship between endometriosis and migraine.

Endometriosis is a multifaceted disease characterized by complex epidemiological, etiological, diagnostic, phenotypic and prognostic features [ 1 ]. It predominantly affects young women of reproductive age, with a peak incidence rate of 6 per 1,000 person-years in the 30–34 age range [ 2 ]. Growing evidence suggests that considering endometriosis exclusively as a pelvic gynecological disorder fails to capture its complex nature and heterogenous clinical manifestations, including its association with multiple comorbidities [ 3 ]. Among these comorbidities, the association between endometriosis and migraine, supported by various studies [ 3 – 7 ], warrants special attention. Individuals with endometriosis are more likely to have migraine, and similarly, the prevalence of endometriosis is higher in patients with migraine compared to the general population [ 3 , 4 , 6 – 8 ]. Migraine disproportionately affects women, with an incidence rate more than double the rate for men (23 per 1000 person-years compared to 10 per 1000 person-years) [ 12 – 14 ]. Moreover, it occurs most commonly in young women of reproductive age, with peak incidence between 25 and 34 years old [ 9 ]. Notably, pain is a central feature of both endometriosis and migraine, significantly worsening quality of life and interfering with daily activities [ 10 – 12 ]. Recent studies have also uncovered a strong and significant genetic overlap between endometriosis and migraine [ 13 – 15 ], further supporting this association [ 3 – 5 ]. Finally, both endometriosis and migraine have been linked to similar comorbidities, including autoimmune-related traits, fibromyalgia, and chronic fatigue syndrome [ 16 – 20 ]. However, an epidemiological association of endometriosis with specific migraine subtypes such as with or without aura, as well as the effect of clinical covariates, including hormonal factors, has not yet been fully elucidated. Considering that menstruation is an important migraine trigger [ 21 ] and estrogen withdrawal has been suggested to trigger migraine [ 22 ], hormonal factors are indeed an important aspect when investigating this link. In this systematic review and meta-analysis, we aimed to quantify the risk of migraine in individuals with endometriosis. Additionally, we aimed to gain understanding of this link by exploring whether endometriosis is associated with various migraine subtypes, and whether this association was maintained when considering only high quality studies. Although this has been previously explored in a meta-analysis [ 6 ], our study provides an updated perspective by incorporating the latest available literature with a greater number of records included in the analysis and offering new insights into the association between these conditions by conducting multiple subgroup and sensitivity analyses.

Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1