Initial data from the prospectively randomized G-MEMBRANE trial and systematic review on the embolization of the middle meningeal artery in the treatment of chronic subdural hematomas

Abstract

Middle meningeal artery embolization (MMAE) is a promising treatment, as surgical treatment of chronic subdural hematoma (cSDH) still exhibits high recurrence rates. G-MEMBRANE is a single-center, investigator-initiated, prospectively randomized controlled trial (RCT) evaluating the efficacy of adjuvant MMAE. Patients with cSDH requiring surgery were randomly assigned to either surgery plus MMAE or surgery alone. The primary endpoint was cSDH recurrence within 3 months of follow-up. Assuming a risk difference of 20%, an interim analysis was planned after the enrollment of 77 patients. A systematic literature review identified all completed RCTs published between 2015 and 2025. The interim analysis included 93 randomized patients, 64 of whom were available for primary endpoint analysis. The drop-out and lost-to-follow-up rate of 19.1% in the control group and 43.5% in the intervention group were high. Mean age was 74.6 ± 13.3 years, 66.7% patients were male. In the intervention group, 5/26 patients (19.2%) developed recurrent cSDH, compared to 9/38 (23.7%) in the control group (p = 0.672). Revision surgeries were performed in 0% in the intervention group and in 4 (10.5%) of the control group (p = 0.140). No adverse events were related to MMAE. Eight RCTs were included in the systematic review with significant results reported in two large multicenter studies and two small single-center studies regarding “recurrences/reoperation.” The small risk difference between the study groups at the time of interim analysis led to the premature termination of the study which remains inconclusive. Considering the limitations and international study results, G-MEMBRANE provides preliminary information on safety and possible recurrence prevention of cSDH using MMAE for routine clinical practice.

Trial registration: Prospectively registered on Nov. 11th, 2021 German Clinical Trials Registry (DRKS00020465).

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-57303-w.

Introduction

Chronic subdural hematomas (cSDH) are symptomatic fluid accumulations in the natural space between the brain surface and the dura mater¹. Their consistency varies from old blood to watery. The long-dominant theory that venous bleeding caused by minor trauma would transform into serous fluid as part of a chronic process has been superseded by more complex ideas about the interaction of inflammatory processes and the formation of neovascularized membranes with hyperpermeable capillaries². Incidence data vary enormously, ranging from 1.72 to 91 cases per 100,000 inhabitants per year^3,4. Demographic change with the associated increase in the prevalence of comorbidities and the use of anticoagulants are associated with increases in incidence⁵. The standard treatment for symptomatic cSDH is evacuation and, if necessary, drainage via burr hole trepanation, which is associated with recurrence rates between 10 and 40%^6,7. Adjunctive and alternative drug therapy, including anti-inflammatory, anticoagulant, and angiogenesis-modulating substances, is the subject of systematic research^8–10.

Middle meningeal artery embolization (MMAE) in three patients with recurrent cSDH was first described by Takahashi in 2002¹¹ and has since been used in individual cases. In 2018, Ban et al. reported a recurrence rate of 1.4% for 27 patients who had undergone MMAE alone or in addition to surgical evacuation, which they compared to a recurrence rate of 27.5% in a historical control group¹². The publication attracted international interest from the research community and led to the initiation of multiple studies.

The G-MEMBRANE study (Middle Meningeal Artery Embolization Minimizes Burdensome Recurrence Rates After Newly Diagnosed Chronic Subdural Hematoma Evacuation) is a randomized controlled trial to investigate the additional benefit of MMAE in patients who had previously received standard therapy for their cSDH by means of burr hole trepanation. The results will be compared with the results of all previously published randomized controlled trials (RCTs) on MMAE for cSDH in the form of a systematic literature review.

Methods

MEMBRANE

G-MEMBRANE is a science-initiated, randomized controlled, monocentric study conducted at a tertiary care hospital in Berlin, Germany. Prior to the start of the study, the study protocol was published¹³, and the research project was registered in the German Clinical Trials Registry (DRKS00020465) on 11/11/2021. The study was approved by the Charité Ethics Committee (EA1_119_19) and carried out in accordance with the Declaration of Helsinki and the German Radiation Protection Act. The candidates’ informed consent was obtained by the study physicians (neuroradiologists and neurosurgeons) prior to their inclusion in the study. The study was funded as research project FR341 by the German Social Accident Insurance.

Screening and inclusion

Patients aged 18 years and older with a diagnosis of cSDH requiring surgery were included preoperatively after providing their informed consent. Indications for surgery were based on cSDH volume (midline shift ≥ 5 mm) and clinical symptoms such as headache, seizures, paresis, sensory or speech disorders. In a 1:1 randomization with variable block length, patients were then assigned to the intervention group (burr hole trepanation with drainage plus MMAE) or the control group (burr hole trepanation with drainage only). SecuTrial (interActive Systems) was used by the researchers for randomization and data collection.

Individuals with acute intracranial hemorrhage, no indication for surgical treatment or who had undergone surgical treatment by craniotomy/craniectomy, homozygous factor 13 activity < 10%, and pregnant women were excluded. Individuals in whom previous examinations of the cervical vessels had identified a high procedural risk or technical impossibility of embolization were also excluded. In an addendum to the advisory ethics committee, inclusion was made possible in cases of existing care status and after consent had been obtained from authorized persons. Participants were excluded from the study if compliance with follow-up appointments was not expected (see study flow in Figure S1 in the supplement).

Performing embolization after surgical evacuation

The procedure was performed in the radiology department using a biplane angiography system (Philips Azurion) by medical staff specializing in neuroradiology. First, the ipsilateral internal carotid artery was probed with a 5 F vertebral catheter via a 5 F sheath in the common femoral artery in order to identify any external-internal collaterals or anatomical variants. After placing the guide catheter in the ipsilateral external carotid artery, a 1.5 F flow-controlled microcatheter (Marathon, Medtronic) was navigated into the MMA. The frontal and parietal branches of the MMA were located superselectively and embolized with polyvinyl alcohol (PVA) particles measuring 40–150 μm or liquid embolic agent (Onyx 18/20, Medtronic or Squid 12/18, Balt) (see Fig. 1).

Fig. 1.

MMA embolization. Lateral angiogram of the MMA before embolization (A) with frontal (1) and parietal (2) branches, and after embolization with PVA particles in both branches (B). The white arrowheads image a “leafless tree” with occlusion of the distal vascular territories. Anterior-posterior (C) and lateral (D) angiogram of the MMA with frontal (1) and parietal (2) branches before embolization. Distribution of the radiopaque liquid embolic (*) parietal and posterior in the anterior-posterior (E) and lateral (F) images. “Dangerous” external-internal collaterals in the anterior-posterior (G) and lateral (H) angiogram of the external carotid artery with retrograde contrast enhancement of the internal carotid artery (*) including branches of the middle carotid artery and the ophthalmic artery (black arrowhead) in case of proximal occlusion of the internal carotid artery.

Endpoints

The primary endpoint was defined by at least one criterion: computed tomography evidence of recurrence of at least the same volume (± 10%) compared to initial postoperative imaging, recurrence with a midline shift ≥ 5 mm, or recurrence requiring surgery (due to clinical symptoms such as headache, seizures, paresis, sensory or speech disorders) within three months after surgery. Volume measurements were performed using Brainlab software (The Brainlab Companies) by a local radiologist independent of the intervention team. The radiologist determined the cSDH volume in cm^³ on axial 3 mm CT slices from the inner membrane to the skull in a semi-automatic process after defining the cSDH as a region of interest (see Figure S2 in the supplement). Postoperative intracranial air and drainage-systems were excluded from the cSDH volume. Blinding was not possible due to the visibility of the embolization material.

Secondary endpoints were the occurrence of neurological deficits as quantified using the modified Rankin Scale (mRS), as well as recurrence and intervention-associated complications during the three-month follow-up period.

Statistical planning

Based on the results of published studies (in particular¹², we expected a difference in the recurrence rate between the control group and the intervention group of 20% during the 3-month observation period. With a power of 80% and a significance level of 5%, a total of 138 evaluable subjects (i.e., 69 per treatment arm) were required to demonstrate the assumed risk difference of 20%, resulting in 154 individuals to be included with an estimated dropout rate of 10%. After the recruitment of half of the target sample, one interim analysis was planned using a group sequential design with equally spaced information in increments. The boundary structure comprised two-sided symmetric efficacy-only boundaries based on the O’Brien-Fleming alpha-spending function. No continuity correction or zero-cell adjustment was applied. The trial was to be stopped early for efficacy if the interim analysis demonstrated a statistically significant reduction in the recurrence rate in the experimental group (chi-square test: p < 0.0031). Conversely, the trial was to be terminated prematurely if the observed treatment effect proved considerably smaller than expected and a subsequent sample size re-estimation indicated that completion of the study with a feasible sample size was unlikely. The trial would also be terminated early, if major novel scientific evidence demonstrated any substantial favor of one approach over the other.

The primary analysis was performed on the modified intent-to-treat (mITT) set including all randomized patients and based on the treatment arm they were randomized to. The modification of the ITT set implied that the primary outcome must be fully documented at the follow-up visit at 3 months. The primary endpoint was analyzed using Pearson’s chi-square test for independent samples. A sensitivity analysis additionally included patients in whom the primary endpoint had been documented before the 3-month follow-up time point. The proportion of recurrences requiring surgery was compared using Fisher’s exact test. Since this test was not planned a priori and was only performed to better understand the implications, the result should be considered purely exploratory. The analysis of secondary endpoint 1, impairment after neurological deficits in comparison of the two treatment groups, was conducted using the Wilcoxon rank sum test. The statistical evaluation was performed using STATA (StataCorp LLC, College Station, TX) version 16.1.

Systematic literature search

In January 2026, two radiologists with 11 and 7 years of professional experience, respectively, conducted a systematic literature search in EMBASE, Web of Science, PubMed/Medline, and Cochrane in accordance with PRISMA guidelines¹⁴. A first literature review took place in October 2025 within the drafting process of the manuscript but was repeated in January 2026 to ensure the actuality of the review results within the manuscript before manuscript submission.

The titles and abstracts of the identified articles were independently reviewed for inclusion and exclusion criteria. The databases were searched for RCTs on the therapeutic effect of MMAE in cSDH. Full texts written in English and published in peer-reviewed journals between 2015 and 2025 were included (see search strategy in the Supplement Table S1) and examined for differences, similarities, and specific characteristics in terms of study populations, implementation of interventions, observation periods, endpoints and results. Due to the heterogeneity of the study results, a conscious decision was made not to attempt quantitative data aggregation in a meta-analysis. The risk of bias assessment was performed by both radiologists according to six categories in a consensus fashion (Table 1).

Table 1.

Baseline characteristics.

Variable	Intervention group (n = 26)	Control group (n = 38)	Total (n = 64)
Gender, n (%)	n = 24	n = 36	n = 60
Female	5 (20.8)	15 (41.7)	20 (33.3)
Male	19 (79.2)	21 (58.3)	40 (66.7)
Age, Mean ± SD [range]	73.7 ± 12.6 [43–89]	75.3 ± 13.9 [30–94]	74.6 ± 13.3 [30–94]
Symptoms on admission, n (%)	n = 20	n = 30	n = 50
Amnestic aphasia	5 (25.0)	8 (26.7)	13 (26.0)
Paresis (Janda grades)
No discernible muscle contraction (0)	0	1 (3.3)	1 (2.0)
Noticeable reaction, insufficient for movement (1)	0	0	0
Full range of motion, not against gravity (2)	0	1 (3.3)	1 (2.0)
Full range of motion against gravity, without resistance (3)	1 (5.0)	4 (13.3)	5 (10.0)
Full range of motion against light/moderate resistance (4)	9 (45.0)	13 (43.3)	22 (44.0)
Full range of motion against strong external resistance (5)	10 (50.0)	11 (36.7)	21 (42.0)
Alertness	20 (100.0)	30 (100.0)	50 (100.0)
Anisocoria	0	2 (6.7)	2 (4.0)
Modified Rankin Scale
0	5 (25.0)	9 (30.0)	14 (28.0)
1	9 (45.0)	12 (40.0)	21 (42.0)
2	5 (25.0)	4 (13.3)	9
3	0	4 (13.3)	4 (8.0)
4	1 (5.0)	1 (3.4)	2 (4.0)
5	0	0	0
6	0	0	0
BMI, n (%)	n = 6	n = 11	n = 17
Underweight (< 18.5)	0	0	0
Normal weight (18.5–24.9)	2 (33.3)	4 (36.4)	6 (35.3)
Pre-obesity (25–29.9)	2 (33.3)	3 (27.3)	5 (29.4)
Obesity (> = 30)	2 (33.3)	4 (36.4)	6 (35.3)
Type of pre-existing condition, n (%)	n = 25	n = 37	n = 62
Alcoholism	1 (4.0)	0	1 (1.6)
Arterial hypertension	17 (68.0)	26 (70.3)	43 (69.4)
Autoimmune disease	0	2 (5.5)	2 (3.2)
Diabetes	10 (40.0)	9 (24.3)	19
Coagulopathy	0	0	0 (30.6)
Liver disease	1 (4.0)	0	1 (1.6)
Kidney disease	5 (20.0)	2 (5.5)	7 (11.3)
Tumor disease	0	1 (2.7)	1 (1.6)
Use of anticoagulants/platelet function inhibitors, n (%)	n = 26	n = 38	n = 64
Yes	10 (38.5)	16 (42.1)	26 (40.6)
ASS (100 mg)	6 (23.1)	9 (23.7)	15 (23.4)
Xarelto (20 mg)	0	1 (2.6)	1 (1.6)
Xarelto (15 mg)	0	1 (2.6)	1 (1.6)
Eliquis (10 mg)	1 (3.8)	1 (2.6)	2 (3.1)
Eliquis (5 mg)	1 (3.8)	1 (2.6)	2 (3.1)
Heparin (40 mg)	1 (3.8)	1 (2.6)	2 (3.1)
Lixiana (600 mg)	1 (3.8)	0	1 (1.6)
Marcumar	0	2 (5.3)	1 (1.6)
Genesis of cSDH, n (%)	n = 25	n = 37	n = 62
History of acute SDH	1 (4.0)	5 (13.5)	6 (9.7)
No history of acute SDH	15 (60.0)	23 (62.2)	38 (61.3)
Unclear	9 (36.0)	9 (24.3)	18 (29.0)
Embolization performed	n = 26
General anesthesia, n (%)	26 (100.0)
Number of embolized branches, n (%)
One branch	5 (19.2)
Two branches	21 (80.8)
Main trunk embolized, n (%)	3 (11.5)
Embolization material, n (%)
PVA particles	10 (38.5)
Liquid embolization material	16 (61.5)
Orbital collaterals	7 (26.9)
Other “dangerous” collaterals	1 (3.8)

n/N – number of subjects, % – percentage, mean – arithmetic mean, SD – standard deviation, cSDH – chronic subdural hematoma, PVA – polyvinyl alcohol.

Results

G-MEMBRANE interim analysis

Study population

Of the 93 subjects enrolled between May 2022 and February 2025, 29 were excluded from the analysis population, corresponding to a dropout rate of 31.2%. The main reasons for dropouts in the intervention group were failure to perform MMAE either due to withdrawn consent or surpassing the timeframe of 72 h post-surgery (n = 9) and lack of follow-up CT in the control group (n = 7). In detail, the research team experienced severe difficulties in convincing the elderly participants to adhere to the study protocol and actually undergo additional MMAE after initial surgery. Furthermore, collaboration with anesthesiology was a serious bottleneck during the study due to the elective nature of postoperative MMAE, resulting in missed time windows of 72 h in some cases. Missing follow-up scans might partially be attributed to reduced compliance related to comorbidities. Primary analysis was based on a modified ITT population with completed 3-month follow-up records. G-MEMBRANE was planned in accordance with CONSORT guidelines; thus, the participant flow is presented accordingly, yet it should be perceived and considered within the scope of an interim analysis (see Figure S1 ‘CONSORT Flow diagram’ in the supplement).

Two-thirds (n = 40; 67%) of the primary analysis set (N = 64) consisted of men with an average age of 75 years. There were no relevant differences between the two groups in terms of age distribution, comorbidities, and BMI, but the proportion of women was higher in the control group than in the intervention group (see Table 2). The interval between the initial examination and the follow-up examination was 252 ± 257 days on average in the intervention group compared to 196 ± 190 days in the control group. All MMAEs were performed under general anesthesia, with liquid embolization material used in 16 cases (62%) and PVA particles in 10 cases (38%). During 21 of 26 (81%) MMAEs, two branches were embolized, while during the remaining 5 (19%) MMAEs, only one branch was occluded. Embolization of the main trunk was performed in three cases (12%).

Table 2.

RCTs included in the systematic literature search.

Study	Period	Country	Number of subjects IG/CG	Inclusion/exclusion - Criteria	Control group(s)	MMAE Time	MMAE technique	Endpoints	Results	Sponsor
Ng et al. − 2020	04/2018-10/2018	France Single-center	Initial 21 Final 19/22 Median age 77 and 75 years, respectively, 56% ♂	Inclusion: > 18 years, cSDH requiring surgery Exclusion: previous cSDH surgery, cSDH width < 10 mm or > 50% acute component, underlying pathology (cyst, vascular), < 6 months life expectancy	Surgery (drill hole/craniotomy)	Additive during the same hospital stay	PVA 150–250 μm (coils if necessary) 100% under sedation n = 1 technically unsuccessful due to tortuous vessels	1.cSDH resorption within 90 days on CT	Lower cSDH volume in the IG (mean difference 17.5 mL, 95% CI 3.87–31.16, p = 0.015)	No funding declared
								2. Recurrence (persistent cSDH with remaining or new symptoms)	One recurrence in each group
								3. Safety	n = 1 death in the CG due to pneumonia, no MMAE-associated complications
Davies et al. 2024 – “EMBOLISE"	12/2020-08/2023	USA Multicenter in 39 centers	Initial 197/203 Final 167/180 Median age 73 and 71 years 73% ♂	Inclusion: 18–90 years, subacute/chronic SDH requiring surgery Exclusion: Life expectancy < 1 year, mRS ≥ 4 before cSDH, Markwalder Grading Scale ≥ 3	Surgery (drill hole/craniotomy)	Additive pre-surgery (58%) or post-surgery (42%) within 48 h after randomization	Open-label Onyx 79% under general anesthesia 93.9% technically successful n = 10 Anastomotic failures n = 1 coiling n = 53 (13.2%) lost to follow-up n = 1 refused MMAE	1. recurrence or progression with indication for surgery within 90 days	Fewer recurrence surgeries in the IG n = 8 (4.1%) in IG n = 23 (11.3%) in CG (relative risk 0.36; 95% CI 0.11–0.80, p = 0.008; NNT 14)	Industrial (Medtronic)
								2. Worsening mRS within 90 days	11.9% neurological deterioration in the IG, 9.8% in the CG
								3. Mortality within 90 days	n = 10 (5.1%) in the IG n = 6 (3.0%) in the CG
								4. SAE within 180 days	n = 4 (2.0%) in the IG after 30 days (n = 2 strokes in MMAE, not caused by embolism)
								5. Rehospitalization, change in hematoma volume and width, midline shift within 90 days	Similar in both groups
Fiorella et al. − 2025 STEM	11/2020-05/2023	USA, France, Sweden Multicenter in 32 centers	Initial 149 Final 120/129 Median age 73 70% ♂	Inclusion: symptomatic cSDH > 10 mm width Exclusion: mRS > 1 before cSDH, craniotomy	surgery (burr hole/port evacuation) n = 189 or Conservative therapy n = 121	Additive pre-surgery or additive to conservative therapy within 24 h after randomization	Open-label Squid 100% under general anesthesia 99% technically successful n = 1 No embolization due to anatomical causes n = 1 partial embolization plus coiling	1. recurrence or residual findings > 10 mm after 180 days; Reoperation or rescue surgery within 180 days; Disabling stroke/myocardial infarction/death due to neurological causes within 180 days	Fewer primary endpoints in the IG n = 19 (16%) in IG, n = 47 (36%) in CG (OR 0.36; 95% CI 0.20 to 0.66; p = 0.001) Especially in the conservatively treated subgroup benefit of additive MMAE (19% in the IG and 56% in the CG) Most frequent primary event Recurrence/rescue surgery, n = 51 (20%)	Industrial (BALT)
								2. Disabling stroke or death within 30 days	No difference n = 4 (3%) in the IG n = 5 (3%) in the CG
								3. mRS after 180 days	No significant difference
								4. AE/SAE due to MMAE	None
Debs et al. − 2024	03/2019-11/2022	USA Mono-centric	Initial 19 Final 17/18 Median age 66 and 71 years, respectively 66% ♂	Inclusion: 18–90 years, cSDH requiring surgery, mRS < 5 Exclusion: mRS > 4, Creatinine > 1.8, Coagulation disorder, Coagulopathy, Pregnancy	Surgery (drill hole or mini-craniotomy)	Additive within 72 h post-surgery	Onyx 100% under general anesthesia	1. Neurological outcome 4–6 weeks after discharge (or earlier in case of deterioration)	No difference at discharge, but deterioration in the control group at follow-up (33% vs. 71%, RR 0.44, p = 0.03, NNT 2.6).	No funding declared
								2. Medical resource utilization until complete resolution of cSDH	More repeat surgeries in the CG n = 1 (6%) in IG n = 7 (39%) in CG (p = 0.02, NNT 3) Longer hospital stay in the CG (p = 0.02) More imaging in the CG (p = 0.01) Faster resolution in IG (⌀ 39 d in IG and 205 d in CG, p = 0.02)
Liu et al. − 2024 – “MAGIC-MT"	03/2021-05/2023	China Multicenter in 31 centers	Initial 360/362 Final 346/345 Median age 69 years, 82.5% ♂	Inclusion: > 18 years, symptomatic non-acute SDH with space-occupying effect, mRS < 3 Exclusion: Craniotomy, bilateral SDH with unclear etiology of symptoms, life expectancy < 1 year, mRS > 2 before onset of symptoms, SDH recurrence	Surgery (drill hole) in 78% or Conservative therapy in 22%	Additive pre-surgery or additive to conservative therapy	Open-label Onyx Local anesthesia or general anesthesia In 99.6% pre-op	1. Symptomatic SDH recurrence (width > 10 mm) or progression (increase > 3 mm from baseline/post-op) or re-op within 90 days	Fewer primary endpoints in the IG n = 24 (6.7%) in the IG n = 36 (9.9%) in the CG (− 3.3%; 95% confidence interval, − 7.4–0.8%; p = 0.10) Especially in the conservatively treated subgroup benefit of additive MMAE	Partially industrial (Covidien/ Medtronic)
								2. Late recurrence after 1 year, successful MMAE in DSA, change in hematoma thickness or volume after 90 days, change in midline after 90 days, change in mRS between 90 days and 1 year, quality of life after 90 days and 1 year, length of hospital stay, re-hospitalizations, hospital costs, deaths in 1 year	No difference between the groups
								3. SAE within 90 days and 1 year, intervention-associated complications within 30 days	SAE less frequent in the IG than in the CG (6.7% vs. 11.6%) Mortality lower in the IG (0.6% vs. 2.2%) Intervention-associated complications: n = 1 facial nerve palsy and n = 2 KM allergy
Lam et al. − 2023	04/2021-09/2022	Australia Multicenter in 2 centers	Initial 17 Final 16/16 Median age 64 and 72 years, respectively 75% ♂ in IG, 58% ♂ in KG	Inclusion: > 18 years, cSDH requiring surgery > 10 mm width Exclusion: underlying cause (malformation, cyst)	Surgery (drill hole/craniotomy)	Additive post-op during the same hospital stay	Qnyx-18, Squid-12, Phil 25%, NBCA with 25% Lipiodol 100% under general anesthesia	1. recurrence requiring surgery (persistent or new cSDH with persistent or new symptoms)	No recurrence in the IG with n = 3 (15.8%) recurrences in the CG (p = 0.234)	Not industrial (Life Health Care Group)
								2. Residual hematoma width after 6 weeks or 3 months	No significant difference between the groups
								3. mRS after 6 weeks or 3 months	Significantly lower mRS (0–1) after 3 months in the IG (100% vs. 53%, p = 0.018).
								4. AE/SAE	No MMAE-associated complications
Shotar et al. − 2025 – “EMPROTECT"	07/2021-03/2023	France Multicenter in 12 centers	Initial 171/171 Final 162/157 Median age 77 80% ♂	Inclusion: > 18 years with recurrent cSDH or high risk of recurrence Exclusion: mRS > 4 prior to cSDH, life expectancy < 6 months, renal failure, pregnancy, contrast medium allergy	Surgery (drill hole)	Additive post-op within 7 days	Embosphere-TAGM 300–500 μm n = 13 No embolization at anastomoses n = 4 Frustrated catheterization in cases of difficult access n = 2 No embolization in severe atheromatosis n = 5 refused MMAE	1. cSDH recurrence after 6 months (> 10 mm width, > 5 mm midline shift or symptoms, reoperation, rehospitalization)	Fewer cSDH recurrences in IG n = 24 (14.8%) in the IG n = 33 (21%) in the CG (OR 0.64, 95% CI 0.36–1.14, p = 0.13)	No funding declared
								2. Recurrence surgery rate after 6 months, mRS ≥ 4 after 1 and 6 months, mortality after 1 and 6 months	Fewer repeat surgeries in the IG n = 7 (4.3%) in the IG n = 13 (8.3%) in the CG (p = 0.14) No significant differences in mRS or mortality
								3. AE/SAE due to MMAE	n = 1 (0.6%) MCA occlusion during MMAE n = 2 (1.2%) transient hemiparesis or aphasia n = 1 (0.6%) Headache
Manzato et al. − 2025	08/2021-08/2024	Brazil Single-center	Initial 33/43 Final 27/38 Median age 73 and 74 years 84% ♂ in IG 75% ♂ in CG	Inclusion: > 18 years, cSDH requiring surgery, Exclusion: acute SDH (> 50% fresh components), previous intervention for cSDH, underlying disease	Surgery (drill hole)	Additive post-op within 7 days	NBCA with 14% lipiodol 100% general anesthesia n = 1 technically not possible	1. Recurrence rate within 90 days (increase in cSDH with symptoms and need for surgery)	Fewer recurrences in the IG n = 1 (4.2%) in the IG n = 4 (12.5%) in the CG (p = 0.379)	No funding declared
								2. 90-day mortality	Lower mortality in the IG n = 2 (6.1%) in the GI n = 5 (11.6%) in the CG (p = 0.434)
								3. AE/SAE, length of hospital stay	More frequent complications during hospitalization in CG (p = 0.060). n = 1 SAE due to MMAE (pseudoaneurysm of the AFC) Similar length of hospitalisation in the IG and CG.

IG – intervention group, CG – control group, AE/SAE – (serious) adverse events (complications), SDH – subdural hematoma, AFC – common femoral artery, MCA – middle cerebral artery, CM – contrast agent, NBCA – n-butyl cyanoacrylate.

Analysis of the primary endpoint

The primary endpoint was based on the modified ITT population: During the observation period, 5 of 26 patients (19.2%; 95% confidence interval [CI]: 0.09–0.38) in the intervention group experienced cSDH recurrence, while the recurrence rate in the control group was 9 of 38 patients (23.7%; 95% CI: 0.13–0.39). The relative risk of recurrence for the intervention group compared to the control group was 0.81 (95% CI: 0.31–2.15) and did not differ significantly between groups, as indicated by Pearson’s chi-square test, χ²(1, N = 64) (p = 0.672). The sensitivity analysis, additionally accounting for patients whose primary endpoint had been documented prior to the 3-month follow-up assessment, showed similar recurrence rates with 7 of 34 patients (20.6%; 95% confidence interval [CI]: 0.10–0.37) in the intervention group experiencing cSDH recurrence, compared with 9 of 40 patients (22.5%; 95% CI: 0.12–0.38) in the control group. Consequently, the risk of recurrence was also similar between groups, with a relative risk of 0.92 (95% CI: 0.38–2.20) for the intervention group versus control group.

None of the patients in the intervention group (0 of 26, 0.0%; 95% CI: 0.000–0.129) experienced recurrence requiring surgery, while 4 of 38 patients (10.5%; 95% CI: 0.042–0.241) required surgery due to recurrence in the control group. The relative risk of recurrence requiring surgery for the intervention group was 0.16 (95% CI: 0.01–2.86; p = 0.14) compared to the control group (see Fig. 2).

Fig. 2.

G-MEMBRANE Primary Analysis. (A) The incidence of CT-morphological and surgically required recurrences was not significantly lower in the embolization group. (B) Relative risk reduction for the occurrence of recurrence according to volume measurement and surgical intervention.

Analysis of secondary endpoints

The neurological impairments assessed using mRS were similarly distributed across the groups at baseline, with 25% of subjects in the intervention group and 30% of subjects in the control group experiencing no neurological symptoms. At the follow-up visit, however, the proportion of subjects without neurological symptoms was higher in the intervention group compared to the control group (75% vs. 50%), and the proportion of subjects with neurological impairment was correspondingly lower (z=−1.70, p = 0.089) (see Fig. 3).

Fig. 3.

Neurological symptoms according to the modified Rankin Scale.

No complications associated with burr-hole drainage or the intervention were observed during the course of the study. An epileptic seizure that occurred six weeks after MMAE in the intervention group could not be attributed to recurrence or therapy by imaging or neurophysiological testing.

Premature termination of the study

The planned interim analysis commenced in March 2025 with volumetry of the cSDH on follow-up imaging. It was completed in August 2025 and revealed that the difference in recurrence rates between the two study groups was considerably smaller than anticipated. Based on the prespecified statistical assumptions for the primary endpoint, an unblinded sample size re-estimation was conducted using PASS 2020 Power Analysis and Sample Size software (NCSS, LLC. Kaysville, Utah, USA). Including a conditional power of 80% and a nominal alpha level of 0.0469 for the final analysis and taking into account the numbers of patients in both groups and the observed recurrence rates at interim analysis, the required sample size to achieve statistically significant results was recalculated to be approximately 2800 participants. Given the observed enrollment rate, reaching this sample size within a reasonable time frame was deemed infeasible. Consequently, the study was terminated early, as continuation was unlikely to demonstrate a statistically significant treatment effect.

Systematic literature search

Of the 204 hits identified, 111 publications remained after removing duplicates (n = 81) and ongoing/unpublished studies (n = 12) such as the American MEMBRANE trial¹⁵. After screening the titles and abstracts, eleven full texts were extracted, of which eight RCTs were ultimately included in the systematic literature analysis (see Figure S3 in the supplement and Table 3).

Table 3.

Risk of Bias Assessment (RoB 2)³¹.

Between 2018 and 2025, a total of 1,961 [35–722] individuals were included in five multicenter^16–20 and three monocentric^21–23 studies from 119 centers, China¹⁸, the US^17,19,22, Brazil²³, Australia¹⁶, Sweden¹⁹, and France^19–21. All studies were published between 2020 and 2025, two publications contained data from an interim analysis^16,22 and another contained data from a pilot study²¹. The drop-out and lost-to-follow-up rates ranged between 13% and 21%^17,19,23. Three studies were industry-funded^17–19.

Regarding the endpoint “recurrence/recurrence surgery,” significant results were reported in two large multicenter studies with liquid embolization^17,19 and in two small monocentric studies with liquid embolization and particle embolization^21,22.

An overview of the included studies and a detailed summary of the inclusion and exclusion criteria, study populations, implementation of the interventions, observation periods, endpoints, and results can be found in the supplement. Table 1 assesses the risk of bias of each trial according to six crucial categories. The included RCTs show some or high concern for bias, mainly due to inherently missing blinding and dropouts in the intervention groups. Quantitative data aggregation in a meta-analysis was beyond the scope of this systematic review.

Discussion

MMAE has been used to treat cSDH for approximately 25 years²⁴. Serious complications have rarely been reported, while there is growing evidence of the effectiveness of the procedure in the primary prevention of recurrence¹². This development prompted a large number of studies. G-MEMBRANE is a science-initiated RCT without industrial sponsorship from Germany on the effectiveness of MMAE in the prevention of recurrence after surgical decompression of cSDH. The study was prematurely terminated after the planned interim analysis because no significant difference in the observed radiological recurrence rates (19.2% vs. 23.7%) was to be expected within a reasonable period of time based on the recalculated number of cases. One reason could be the very high drop-out and lost-to-follow-up rates of 19.1% in the control group and 43.5% in the intervention group. The proportion of recurrence operations (0% vs. 10.5%) and of individuals without neurological symptoms (75% vs. 50%) indicated an advantage of MMAE in each case.

The limitations of the G-MEMBRANE study lie primarily in its monocentricity and the associated small number of subjects. In addition, high dropout and lost-to-follow-up rates posed a risk for attrition bias due to more than twice as many dropouts in the intervention group. International comparison emphasized, that G-MEMBRANE had a higher dropout rate than other RCTs and that the results need to be analyzed and interpreted with the utmost caution. Missing follow-up CTs in both groups reflect the difficulties of studies with elderly participants who are challenged to reach follow-up examinations. To minimize dropouts through postoperative withdrawals of consent for MMAE, future researcher could perform MMAE preoperatively, as done in three multicenter RCTs^17–19. Implementation of additional reminders before the due date of follow-up examinations (phone calls, emails, text messages) could help decrease the lost-to-follow rate in this patient group. Finally, dropouts and lost-to-follow up cases were analyzed through a modified ITT analysis as planned per protocol; however, this weakened the randomized comparison of G-MEMBRANE.

Within the last seven years, eight working groups have published RCTs with similar results. Regarding the endpoint “recurrence/recurrence surgery,” significant results were achieved in two large multicenter studies with liquid embolics^17,19 as well as in two small single-center studies with liquid and particle embolization^21,22. Of note, recurrence surgeries were necessary in none of the participants of the intervention group in G-MEMBRANE, which concurred well with reported rates of 0–16% in international RCTs^{16,17,19,20,22,23}. Four RCTs were also limited by monocentricity and small sample sizes^16,21–23 and not all studies performed alpha adjustment despite multiple testing and increased risks of false-positive results^16,22,23.

G-MEMBRANE differs from international RCTs in various aspects: G-MEMBRANE included adults with cSDH requiring surgery while other groups included an age limit of 90 years^17,22 and excluded patients with reduced neurological status based on the modified Rankin Scale (mRS)^17–20. One study required the presence of a recurrence or a high risk of recurrence for study inclusion²⁰. These criteria pose significant risk for systematic bias. MMAE was performed within 72 h postoperatively in G-MEMBRANE. Some RCTs selected a treatment period of up to 7 days postoperatively^20,23, others performed MMAE preoperatively^18,19. Both aspects can potentially reduce dropouts compared to G-MEMBRANE. Additionally, in G-MEMBRANE MMAE was performed upon the neuroradiologists preference with PVA or liquid embolics. Larger industry-sponsored trials limited treatments to liquid embolics, possibly resulting in more homogeneous but potentially biased treatment results^17–19,22. Moreover, some groups used shorter observation periods of 4–6 weeks^16,22 which might have reduced the lost-to-follow-up rate compared to G-MEMBRANE. At last, the primary and largely radiological endpoint of G-MEMBRANE is opposed by positive trials including clinical symptoms in their primary endpoints^17,19,22. Only one positive trial focused on radiological cSDH resorption within 90 days as primary outcome²¹. Radiological criteria are certainly highly objective, but existing literature suggests that outcome of cSDH patients might be insufficiently reflected by hematoma volume and midline shift.

Neither the inclusion and exclusion criteria, the timing and technique of embolization, nor the study periods by itself sufficiently explain why some studies were able to confirm the efficacy of MMAE, whereas G-MEMBRANE and four other mono- and multicenter studies were not. Rather, a multifactorial influence of these parameters is to be assumed. However, significantly better treatment outcomes in terms of mRS^16,22, hematoma resolution²², the occurrence of severe complications¹⁸, length of hospital stay, and amount of imaging required²² suggest that the effects of MMAE go beyond simple measurements such as hematoma volume. Future studies focussing on combined endpoints including recurrence surgeries and long-term clinical follow-up rather than radiological recurrence might describe significant advantages of MMAE superiorly.

For future evaluation of therapeutic effects in subpopulations, the studies by Fiorella et al. and Liu et al^18,19. might be valuable for research groups, as both studies showed evidence of a greater therapeutic effect in individuals with cSDH without compelling indication for surgery. Strict regulations with regards to ethical principles and the use of ionizing radiation for medical research purposes had previously exempted individuals with cSDH undergoing conservative therapy from participating in MMAE studies.

It is expected that MMAE will play a major role in the treatment of cSDH in the future despite a relatively high risk of bias within published RCTs. The positive outcome of some recent studies and meta-analysis support that MMAE has the potential to play a significant role as a treatment option for certain patient groups in the future. Despite many similarities between the RCTs analyzed, available data is currently insufficient to define which group of individuals will benefit most from MMAE and should therefore be recommended for MMAE as part of standard therapy. This conclusion is supported by recent meta-analyses, which describe MMAE as a safe and effective method^25,26, but call for further research, as data on functional outcomes, long-term progression, and effects on both quality of life and mortality in general and in subgroups are currently insufficient^27–29. This particularly applies to women with cSDH and young people, as their proportion within published RCTs was low due to the natural distribution of cSDH in the population³⁰.

In summary, G-MEMBRANE was inconclusive at the time of interim analysis. In any case, a possible effect of MMAE on the recurrence rate of cSDH would be much smaller than expected. Considering the above limitations and the background of international study results, G-MEMBRANE still provides information on MMAE in clinical routine and offers the possibility for improved external validation now and in the future.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

cSDH

Chronic subdural hematomas

MMAE

Middle meningeal artery embolization

G-MEMBRANE

Middle Meningeal Artery Embolization Minimizes Burdensome Recurrence Rates After Newly Diagnosed Chronic Subdural Hematoma Evacuation

RCTs

Randomized controlled trials

PVA

Polyvinyl alcohol

mRS

Modified Rankin Scale

mITT

Modified intent-to-treat

Author contributions

L.G., A.H., J.L. wrote the main manuscript text A.H., S.S. prepared the figures S.M., P.S. reviewed the literature, A.H., D.S. statistical analysis, S.S., M.S. measurements, S.M., J.L., D.S. study protoco, lM.S., P.S. patient examinations. All authors reviewed the manuscript.

Funding

G-MEMBRANE was funded by the German Social Accident Insurance (DGUV) with €130,000 (funding project FR 341). We would like to express our gratitude for this support.

Data availability

Deidentified participant data are available upon reasonable request through the corresponding author (leonie.goelz@ukb.de).

Declarations

Competing interests

The authors declare no competing interests.

Human ethics and consent to participate declarations

The study was approved by ethics committee of the Charité University Medicine, Berlin, Germany (EA1_119_19). The candidates’ informed consent was obtained prior to their inclusion in the study.