← Back
NeuroTrials.ai
Neurology Clinical Trial Database

PREDICT-AEDH

Safety and efficacy of decompressive craniectomy versus standard craniotomy for large acute epidural haematoma with tentorial herniation in China (PREDICT-AEDH): a nationwide, multicentre, open-label, parallel-group, randomised controlled trial

Share Share Copied! Compare

Year of Publication: 2026

Authors: Feng J, Yang C, Xie L, ..., for the PREDICT-AEDH investigators

Journal: Lancet Neurology

Citation: Lancet Neurol 2026; 25: 645–53

Link: https://doi.org/10.1016/S1474-4422(26)00143-2

Clinical Question

In patients with large acute epidural haematoma and tentorial herniation, does primary decompressive craniectomy improve 6-month functional outcomes compared with standard craniotomy with bone-flap replacement?

Bottom Line

Primary decompressive craniectomy did not improve 6-month functional outcomes compared with standard craniotomy with bone-flap replacement in patients with large acute epidural haematoma and tentorial herniation, and significantly increased the risk of delayed intracranial haemorrhage. Routine prophylactic decompressive craniectomy is not supported in this population.

Major Points

  • Largest randomised trial to date directly comparing decompressive craniectomy vs standard craniotomy specifically in large acute epidural haematoma with tentorial herniation
  • No significant difference in 6-month functional outcome on GOSE ordinal analysis (common OR 0.79, 95% CI 0.41–1.58; p=0.51)
  • Decompressive craniectomy nearly quadrupled the odds of delayed intracranial haemorrhage (36% vs 13%; OR 3.79, p=0.0049)
  • 30-day mortality and postoperative cerebral infarction rates were similar between groups
  • Findings refute the prevailing assumption that prophylactic decompressive craniectomy is beneficial in this high-risk population

Design

Study Type: Nationwide, multicentre, open-label, parallel-group, randomised controlled trial

Randomization: 1

Blinding: Open-label (patients and surgeons not masked due to surgical nature and skull defect); outcome assessors and statisticians masked to treatment allocation

Allocation: 1:1, central web-based randomisation with block sizes of 4, allocation concealed until assignment

Enrollment Period: Sept 7, 2020 to March 14, 2025

Follow-up Duration: 6 months

Centers: 28

Countries: China

Sample Size: 120

Analyzed: 120

Analysis: Intention-to-treat; primary outcome analysed with proportional-odds (ordinal) model on GOSE; no imputation for missing data (no participants lost to follow-up); safety outcomes assessed in ITT population

Power Calculation: Sample size based on retrospective data showing mean GOSE scores of 6.63 (SD 1.27) with decompressive craniectomy vs 5.78 (SD 1.58) with standard craniotomy; two-sided α=0.05, 80% power, 10% anticipated loss to follow-up → total 120 patients

Registration: ClinicalTrials.gov NCT04261673

Inclusion Criteria

  • Adults aged 18–65 years
  • Traumatic large acute epidural haematoma (volume >30 mL, thickness >1.5 cm, midline shift >5 mm)
  • Obliterated ambient cistern confirmed by CT
  • Clinical signs of transtentorial herniation (unilateral or bilateral pupillary dilation with or without motor posturing)
  • Required urgent surgical evacuation within 12 h of injury

Exclusion Criteria

  • Extensive intraparenchymal contusions judged to be the primary contributor to mass effect
  • Bilateral epidural haematomas
  • Unsurvivable injury (GCS 3 with bilateral fixed pupils and absence of brainstem function)
  • Absence of preoperative mydriasis
  • Injury-to-surgery interval exceeding 12 h
  • Any condition precluding safe randomisation (e.g., profound haemodynamic instability)

Baseline Characteristics

0:

  • Characteristic: Age, years, mean (SD)
  • Decompressive craniectomy (n=58): 44.24 (11.14)
  • Standard craniotomy (n=62): 42.16 (11.25)

1:

  • Characteristic: Male
  • Decompressive craniectomy (n=58): 50 (86%)
  • Standard craniotomy (n=62): 55 (89%)

2:

  • Characteristic: Road-traffic incident
  • Decompressive craniectomy (n=58): 29 (50%)
  • Standard craniotomy (n=62): 33 (53%)

3:

  • Characteristic: Incidental fall
  • Decompressive craniectomy (n=58): 24 (41%)
  • Standard craniotomy (n=62): 27 (44%)

4:

  • Characteristic: Violence
  • Decompressive craniectomy (n=58): 5 (9%)
  • Standard craniotomy (n=62): 2 (3%)

5:

  • Characteristic: Major extracranial injury
  • Decompressive craniectomy (n=58): 21 (36%)
  • Standard craniotomy (n=62): 25 (40%)

6:

  • Characteristic: Pre-injury antithrombotic medication
  • Decompressive craniectomy (n=58): 0 (0%)
  • Standard craniotomy (n=62): 1 (2%)

7:

  • Characteristic: GCS at enrolment, median (IQR)
  • Decompressive craniectomy (n=58): 6 (4–7)
  • Standard craniotomy (n=62): 6 (4–7)

8:

  • Characteristic: Bilateral mydriasis
  • Decompressive craniectomy (n=58): 53 (91%)
  • Standard craniotomy (n=62): 52 (84%)

9:

  • Characteristic: Skull fracture on admission CT
  • Decompressive craniectomy (n=58): 54 (93%)
  • Standard craniotomy (n=62): 60 (97%)

10:

  • Characteristic: Haematoma volume, mL, median (IQR)
  • Decompressive craniectomy (n=58): 105.16 (84.47–133.80)
  • Standard craniotomy (n=62): 116.57 (89.67–140.79)

11:

  • Characteristic: Midline shift, mm, mean (SD)
  • Decompressive craniectomy (n=58): 10.82 (3.90)
  • Standard craniotomy (n=62): 10.48 (4.16)

Arms

FieldDecompressive craniectomyControl
N5862
InterventionEvacuation of epidural haematoma with removal of a large frontotemporoparietal bone flap to allow decompressionEvacuation of haematoma through a bone flap of sufficient size followed by bone-flap replacement
DurationSingle intraoperative interventionSingle intraoperative intervention

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
Functional status assessed using the Glasgow Outcome Scale-Extended (GOSE), analysed with a proportional-odds (ordinal) modelPrimaryStandard craniotomy: 52/62 (84%) achieved favourable outcome (GOSE ≥5)Decompressive craniectomy: 46/58 (79%) achieved favourable outcome (GOSE ≥5)0.790.51
Favourable functional outcome (GOSE ≥5) at 6 monthsSecondary52/62 (84%)46/58 (79%)0.74
Further cranial surgery within 6 months post-injurySecondary4/62 (6%)5/58 (9%)1.360.74
Health-related quality of life — total EQ-5D-5L score at 6 months, median (IQR)Secondary5 (5–9)6 (5–9)0.13
Cognitive impairment (MMSE <24) at 6 months (survivors only)Secondary7/56 (13%)7/52 (13%)0.921.0
Death within 30 daysSafety3/62 (5%)5/58 (9%)1.850.48
Postoperative cerebral infarction within 6 months post-injury (surgical treatment: 0/11 craniectomy vs 2/11 [18%] craniotomy)Safety11/62 (18%)11/58 (19%)1.081.0
Delayed intracranial haemorrhage within 6 months post-injury (surgical treatment: 5/21 [24%] craniectomy vs 2/8 [25%] craniotomy; non-surgical: 16/21 [76%] vs 6/8 [75%])Safety8/62 (13%)21/58 (36%)3.790.0049
Subdural effusion within 6 months post-injurySafety11/62 (18%)13/58 (22%)1.340.65
Hydrocephalus within 6 months post-injurySafety5/62 (8%)5/58 (9%)1.071.0
Surgical wound infection within 6 months post-injurySafety1/62 (2%)1/58 (2%)1.071.0
Non-cranial adverse events (severe pneumonia and deep venous thrombosis)Adverse5/62 (8%)1/58 (2%)0.20.21

Subgroup Analysis

Prespecified subgroup analyses of favourable outcome (GOSE ≥5) at 6 months (by sex, age, GCS, pupillary reactivity, midline shift, haematoma volume) showed no evidence of heterogeneity across subgroups; all interaction p-values >0.05

Criticisms

  • Open-label design (masking of surgeons and patients not feasible due to visible skull defect)
  • Crossover: 1 patient assigned to craniectomy received craniotomy; 10 patients assigned to craniotomy received craniectomy — may dilute treatment effect in ITT analysis
  • Conducted entirely in China — generalisability to other healthcare systems and surgical practices uncertain
  • Modest sample size (n=120) may limit power for subgroup analyses and detection of smaller effects
  • Narrow age range (18–65 years) excludes older adults who often present with acute epidural haematoma
  • Follow-up limited to 6 months — longer-term functional and quality-of-life outcomes unknown

Funding

None

Based on: PREDICT-AEDH (Lancet Neurology, 2026)

Authors: Feng J, Yang C, Xie L, ..., for the PREDICT-AEDH investigators

Citation: Lancet Neurol 2026; 25: 645–53

Content summarized and formatted by NeuroTrials.ai.