Skip to main content

Foundational Trials

DESTINY II

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Hemicraniectomy in Older Patients with Extensive Middle-Cerebral-Artery Stroke
  • Acronym: DESTINY II
  • Year: 2014
  • Journal published in: The New England Journal of Medicine
  • Citation: Jüttler E, Unterberg A, Woitzik J, et al. Hemicraniectomy in older patients with extensive middle-cerebral-artery stroke. N Engl J Med. 2014;370(12):1091-1100.

Context & Rationale

  • Background
    • “Malignant” middle cerebral artery (MCA) infarction causes rapidly progressive cerebral oedema with life-threatening mass effect and historically very high case fatality under best medical management.
    • Earlier randomised trials in younger adults showed decompressive hemicraniectomy can reduce mortality, but most survivors remain moderately-to-severely disabled.
    • Older patients (a large proportion of malignant MCA cases) were largely excluded from earlier trials, leaving major uncertainty about benefit–harm balance, acceptable outcomes, and shared decision-making.
  • Research Question/Hypothesis
    • In patients aged ≥61 years with malignant MCA infarction, does early decompressive hemicraniectomy plus intensive care increase survival without “severe disability” (pre-specified as modified Rankin Scale [mRS] 0–4 at 6 months) compared with intensive care alone?
  • Why This Matters
    • Defines whether the mortality benefit of surgery extends to older adults and quantifies the disability profile among survivors.
    • Directly informs time-critical neurosurgical referral pathways and ethically fraught discussions about survival with dependency.
    • Provides trial data for guideline recommendations, counselling, and resource planning for neurocritical care and rehabilitation.

Design & Methods

  • Research Question: Among patients ≥61 years with malignant MCA infarction, does early decompressive hemicraniectomy (within 48 hours of symptom onset) plus maximum conservative intensive care improve the proportion with mRS 0–4 at 6 months versus maximum conservative intensive care alone?
  • Study Type: Prospective, randomised, controlled, open-label, multicentre trial (Germany); 1:1 allocation; sequential design with interim analyses and a pre-specified stopping boundary.
  • Population:
    • Setting: Neurocritical care units at participating hospitals.
    • Key inclusion criteria:
      • Age ≥61 years.
      • Acute unilateral MCA infarction involving ≥2/3 of the MCA territory (CT or MRI), including basal ganglia involvement on imaging criteria used for “malignant” infarction prediction.
      • Severe stroke with high NIHSS (dominant hemisphere: NIHSS ≥20; non-dominant hemisphere: NIHSS ≥15).
      • Reduced level of consciousness (NIHSS item 1a ≥1).
      • Ability to start assigned treatment within 6 hours after randomisation and within 48 hours after symptom onset.
    • Key exclusion criteria:
      • Pre-stroke disability (mRS >1).
      • Contraindication to surgery/anaesthesia or anticipated poor life expectancy due to severe comorbidity.
      • Other causes of impaired consciousness or major competing neurological pathology (e.g., intracranial haemorrhage, tumour).
      • Known coagulopathy or ongoing anticoagulation not correctable within the required timeframe (as per trial criteria).
  • Intervention:
    • Early decompressive hemicraniectomy plus maximum conservative intensive care.
    • Surgery performed as soon as possible after randomisation (target within 6 hours) and within 48 hours of symptom onset.
    • Standardised surgical approach: large fronto-temporo-parietal craniectomy (typically ≥12 cm diameter) with dural opening and duraplasty; bone flap removed and stored; subsequent cranioplasty later.
  • Comparison:
    • Maximum conservative intensive care without early decompressive surgery.
    • Conservative management included protocolised neurocritical care (airway/ventilation as needed, haemodynamic targets, temperature and glucose management, osmotherapy/ICP-directed measures, and management of complications), with no planned hemicraniectomy unless protocol deviation/rescue.
  • Blinding: Open-label treatment; outcome assessment was performed by investigators not involved in acute care, but full blinding was not feasible (risk of performance and detection bias for subjective endpoints).
  • Statistics: Sequential (triangular) design; assumed an increase in the primary endpoint (mRS 0–4 at 6 months) from 8.6% (control) to 31.0% (surgery); alpha 5% (two-sided) and 90% power; maximum sample size 160; primary analysis by intention-to-treat with bias-corrected estimates appropriate for sequential monitoring.
  • Follow-Up Period: Primary outcome at 6 months (±14 days); key secondary outcomes at 12 months (±14 days).

Key Results

This trial was stopped early. Recruitment was halted after a pre-specified interim analysis (sequential monitoring boundary crossed for efficacy once sufficient patients had reached the 6‑month endpoint); a total of 112 patients were randomised (including “overrun” patients already enrolled).

Outcome Decompressive hemicraniectomy + ICU (n=49) Conservative ICU care (n=63) Effect p value / 95% CI Notes
Primary endpoint: Survival without severe disability (mRS 0–4) at 6 months 20/49 (38.5%) 10/63 (17.7%) OR 2.91 95% CI 1.06 to 7.49; P=0.039 Bias-corrected estimates accounting for sequential monitoring (primary analysis).
Death (mRS 6) at 6 months 16/49 (32.7%) 46/63 (73.0%) Not reported P<0.001 Marked survival advantage drove most of the primary endpoint difference.
mRS distribution at 6 months (selected strata) mRS 3: 3/49; mRS 4: 15/49; mRS 5: 15/49 mRS 3: 2/63; mRS 4: 7/63; mRS 5: 8/63 Not reported Not reported mRS 0–2: 0 in both groups at 6 months (trial population had very severe strokes).
Survival at 12 months 27/47 (57.4%) 15/62 (24.2%) Not reported 95% CI (surgery) 42.2 to 70.0; 95% CI (control) 14.4 to 35.8; P<0.001 Denominators reflect available 12‑month follow-up data.
mRS 0–4 at 12 months 27/47 (57.4%) 15/62 (24.2%) OR 4.42 95% CI 1.75 to 10.95; P=0.001 Secondary endpoint; survivors remained predominantly dependent.
Health-related quality of life at 12 months (EQ‑5D index, mean ± SD) 0.50 ± 0.31 0.26 ± 0.34 Not reported P=0.03 Available-case analysis (n=38 vs n=45).
Depressive symptoms at 12 months (HDRS, mean ± SD) 9.2 ± 7.5 14.0 ± 7.8 Not reported P=0.02 Available-case analysis (n=37 vs n=45).
Serious adverse events (total events) 88 events 84 events Not reported Not reported Infections and infestations: 18 vs 8 events; nervous system disorders: 23 vs 46 events (including malignant oedema/herniation more frequent under conservative care).
Retrospective consent at 12 months (would agree again to assigned treatment) 17/27 (63.0%) 8/15 (53.3%) Not reported Not reported Among survivors able to provide a response; “no” responses: 5/27 vs 4/15.
  • The primary endpoint improvement reflected a large reduction in mortality, with most survivors living with substantial dependency (mRS 4–5 predominated).
  • At 12 months, surgery remained associated with higher survival and better (but still impaired) health-related quality of life; depressive symptom scores were lower in the surgical arm.
  • Serious infections were more frequent after surgery, whereas neurological deterioration/herniation events were more frequent with conservative care.

Internal Validity

  • Randomisation and allocation:
    • Central randomisation stratified by centre; sealed envelopes were available as back-up (allocation concealment likely adequate up to assignment).
    • Sequential monitoring increases risk of overestimating treatment effects; trial used bias-corrected estimates for the primary endpoint.
  • Dropout or exclusions (post-randomisation):
    • Primary endpoint data were available for the full analysis set at 6 months.
    • At 12 months, outcome denominators were reduced (mRS available: 47/49 vs 62/63), introducing potential attrition bias for longer-term endpoints.
  • Performance/Detection bias:
    • Open-label treatment could influence co-interventions (e.g., intensity of supportive care, timing of limitation of therapy) and subjective outcomes.
    • The primary outcome (mRS) is semi-structured but still subject to assessment bias; assessors were not part of acute care, which may reduce but not eliminate bias.
  • Protocol adherence:
    • Protocol compliance was high: 48/49 (98.0%) received assigned surgery; 62/63 (98.4%) in the conservative arm remained without early surgery.
    • Per-protocol set excluded protocol deviations (n=5 surgery; n=6 control); per-protocol primary endpoint OR 3.61; 95% CI 1.20 to 9.80; P=0.024 (directionally consistent).
  • Baseline characteristics:
    • Groups were broadly similar in age (median 70 vs 71 years), sex distribution, pre-stroke function (mRS 0–1), stroke severity (NIHSS median 18 vs 19), and imaging infarct extent.
  • Heterogeneity:
    • Multicentre design improves robustness; however, surgical technique and ICU practices may vary across sites (mitigated by protocolised surgical standards and ICU care guidance).
  • Timing:
    • Randomisation occurred early (median 25 hours from symptom onset in both groups).
    • Time from randomisation to hemicraniectomy was short (median 1.3 hours; IQR 0.7–2.2), supporting biological plausibility for oedema prevention rather than rescue only.
  • Dose (intervention fidelity):
    • Surgical “dose” (large hemicraniectomy with duraplasty) was standardised; separation from control was strong (minimal crossover).
  • Separation of the variable of interest:
    • Assigned treatment delivered: 48/49 received surgery vs 1/63 in the conservative arm receiving hemicraniectomy (protocol deviation/rescue).
    • Time to randomisation: median 25 hours (IQR 18–30) vs 26 hours (IQR 20–30).
  • Crossover:
    • Crossovers were rare (1 patient in each arm received the opposite strategy), limiting dilution of effect.
  • Outcome assessment:
    • Primary endpoint (mRS dichotomised 0–4 vs 5–6) is clinically interpretable but embeds a value judgement about what constitutes “severe disability”.
    • Secondary endpoints included survival, mRS distribution, and patient-centred measures (EQ‑5D, SF‑36, HDRS), improving interpretability beyond mortality alone.
  • Statistical rigour:
    • Sequential design with pre-specified stopping boundaries was appropriate for an invasive intervention and high-mortality condition, but early stopping can exaggerate effect sizes and widen uncertainty.
    • Confidence intervals were wide for the primary OR (reflecting modest sample size and stopping).

Conclusion on Internal Validity: Moderate-to-strong. Randomisation, high protocol adherence, and strong treatment separation support internal validity, but open-label delivery and early stopping (with modest sample size and wide CIs) limit certainty about the precise magnitude of benefit and the disability trade-off.

External Validity

  • Population representativeness:
    • Enrolled a typical “malignant MCA infarction” phenotype with very severe deficits (high NIHSS, depressed consciousness, large infarct burden) and low pre-morbid disability.
    • Applies primarily to older patients who are independent pre-stroke and can undergo major neurosurgery within a narrow time window.
  • Applicability:
    • Findings generalise best to well-resourced neurocritical care settings with rapid neurosurgical access and rehabilitation capacity.
    • Conducted largely before widespread modern thrombectomy pathways; malignant oedema still occurs despite reperfusion therapy, but incidence and case-mix may differ today.
    • Patients with substantial pre-stroke disability, major comorbidity, or delayed presentation were excluded; extrapolation to these groups is limited.

Conclusion on External Validity: Good for independent older adults with true malignant MCA infarction managed in neurocritical care systems capable of rapid surgery; limited for frail patients, delayed presenters, and health systems without consistent neurosurgical capacity.

Strengths & Limitations

  • Strengths:
    • Addresses a major evidence gap in older adults, a group under-represented in earlier decompressive surgery trials.
    • Clear physiological rationale with tight timing and high protocol adherence; strong separation between strategies.
    • Patient-centred secondary outcomes (quality of life, mood) and retrospective consent data enrich clinical interpretation.
    • Sequential monitoring ethically appropriate for high-risk surgery with high baseline mortality.
  • Limitations:
    • Open-label design; potential influence on co-interventions and treatment limitation decisions.
    • Early stopping and modest sample size yield wide confidence intervals and potential overestimation of effect.
    • Primary endpoint uses a disability threshold (mRS 0–4) that may not align with all patients’ values, particularly given high rates of mRS 4–5 among survivors.
    • Missing data for some 12‑month outcomes and available-case analyses for quality-of-life measures.

Interpretation & Why It Matters

  • Clinical decision-making
    Early decompressive hemicraniectomy in older adults substantially increases survival, but most additional survivors live with dependency (commonly mRS 4–5); this must be explicitly discussed with patients’ families using clear prognostic framing.
  • Practical pathway implication
    Because time to randomisation and surgery was short, systems need pre-defined triggers (imaging + NIHSS + reduced consciousness) to facilitate urgent neurosurgical consultation within the first day after symptom onset.
  • Outcome nuance
    Survival benefit is clear; quality-of-life and mood data at 12 months suggest survivors are not uniformly “worse off”, but the disability burden remains high and preferences vary widely.

Controversies & Subsequent Evidence

  • Disability threshold and value judgements:
    • The pre-specified primary outcome (mRS 0–4) treats mRS 5 as an “unfavourable” outcome, yet many surgical survivors were mRS 4–5; critics emphasised that the clinical meaning of “benefit” depends on patient values and acceptable dependency levels.1
  • Early stopping and effect-size inflation:
    • Stopping at an interim analysis can inflate apparent benefit and increase uncertainty around the true treatment effect size, particularly for disability outcomes where CIs were wide.
  • Open-label care and withdrawal/limitation decisions:
    • Letters highlighted that differential expectations and treatment limitation practices could influence mortality and disability outcomes in unblinded surgical trials; this remains a critical interpretive lens for any surgery-versus-medical management study.2
  • Subsequent evidence base:
    • Individual patient data meta-analysis of randomised trials (including older cohorts) supports a mortality benefit of decompressive surgery with outcome trade-offs that vary by age and disability thresholds, reinforcing the need for preference-sensitive decisions.3
    • European Stroke Organisation guidance on space-occupying brain infarction incorporates DESTINY II and recommends decompressive surgery in carefully selected patients, with explicit attention to expected disability and shared decision-making.4
    • AHA/ASA acute ischaemic stroke guidance recognises decompressive hemicraniectomy as a key option for malignant cerebral oedema in appropriately selected patients, framing it as time-sensitive and preference-dependent.5

Summary

  • DESTINY II tested early decompressive hemicraniectomy versus conservative ICU care in patients aged ≥61 years with malignant MCA infarction.
  • The trial stopped early for efficacy; hemicraniectomy increased survival without “severe disability” (mRS 0–4) at 6 months (38.5% vs 17.7%; OR 2.91; 95% CI 1.06 to 7.49).
  • Mortality was markedly lower with surgery at 6 months (32.7% vs 73.0%) and survival remained higher at 12 months (57.4% vs 24.2%).
  • Most survivors in the surgical arm were dependent (mRS 4–5), underscoring the need for nuanced counselling rather than “mortality-only” framing.
  • Serious infections were more frequent after surgery, while neurological deterioration/herniation events were more frequent with conservative management; QoL and depressive symptoms at 12 months favoured surgery among assessed survivors.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • DESTINY II’s primary endpoint (mRS 0–4) is intentionally less stringent than “independence” thresholds (mRS 0–2), reflecting the older population and ethical focus on survival versus very severe disability.
  • When counselling families, presenting both absolute survival and the distribution of disability (mRS 3–6) is essential for informed preference-sensitive decisions.

Overall Takeaway

DESTINY II established that early decompressive hemicraniectomy in older adults with malignant MCA infarction substantially improves survival, with a corresponding increase in survivors living with significant dependency. Its landmark contribution is not merely demonstrating a mortality signal, but quantifying the disability profile and patient-centred outcomes that underpin modern shared decision-making and guideline recommendations in neurocritical care.

Overall Summary

  • Early hemicraniectomy in patients ≥61 years with malignant MCA infarction increases 6‑month survival without very severe disability (mRS 0–4), largely via reduced mortality.
  • Most additional survivors remain dependent (often mRS 4–5), making value-based counselling central to applying the evidence.
  • Later evidence syntheses and guidelines support surgery in selected patients, emphasising time sensitivity and shared decision-making.

Bibliography