PDF: ARUBA
(2014)Objective
To compare medical management alone with medical management plus interventional therapy (neurosurgery, endovascular embolisation, and/or stereotactic radiotherapy) for preventing death or symptomatic stroke in adults with an unruptured brain arteriovenous malformation.
Study Summary
• Death or mRS ≥2 at 30 months lower with medical: 8/53 (15.1%) vs 24/52 (46.2%); RR 0.33 (95% CI 0.16–0.66).
• Intervention caused more total strokes (45 vs 12, p<0.0001) and more non-stroke neurological deficits (14 vs 1, p=0.0008); no unexpected harms.
Intervention
Neurosurgery, endovascular embolisation, and/or stereotactic radiotherapy (alone or combined at local investigator discretion) plus medical management, vs medical management alone (pharmacotherapy for symptoms and vascular risk factors).
Inclusion Criteria
Adults ≥18 y with an unruptured brain AVM diagnosed by neurovascular imaging, mRS ≤1, and a lesion deemed suitable for eradication by the local centre.
Study Design
Arms: Interventional therapy + medical management (n=114) vs Medical management alone (n=109)
Patients per Arm: 114 vs 109 (226 randomised; 3 excluded because randomisation occurred after dataset lock, leaving 223 analysed)
Outcome
• Death or mRS ≥2 at 30 mo: 15.1% (8/53) vs 46.2% (24/52); RR 0.33 (0.16–0.66). At 36 mo: 14.0% (6/43) vs 38.6% (17/44); RR 0.36 (0.16–0.83).
• Interventional therapy composition: neurosurgery alone n=5, embolisation alone n=30, radiotherapy alone n=31, multimodal n=28; median time to first intervention 76 d (IQR 45–119); 20 had not yet initiated therapy at analysis.
• Safety: total strokes 45 vs 12 (p<0.0001); non-stroke neurological deficits 14 vs 1 (p=0.0008). Spontaneous AVM rupture in medical arm 2.2%/year (95% CI 0.9–4.5). Withdrawal 3% overall.
Bottom Line
In adults with an unruptured brain AVM, medical management alone was superior to medical management with interventional therapy for preventing the composite of death or symptomatic stroke over ~33 months (10.1% vs 30.7%; HR 0.27, 95% CI 0.14–0.54, p<0.0001), and it also reduced death or mRS ≥2. The trial was halted early for benefit of medical management; upfront intervention roughly tripled the near-term risk of stroke or death.
Major Points
- First RCT to compare medical management alone vs interventional therapy (surgery, embolisation, and/or radiotherapy) for adults with unruptured brain AVMs; 226 randomised at 39 sites in 9 countries.
- Halted early at the second planned interim analysis on April 15, 2013 by the NINDS/NIH-appointed DSMB — log-rank statistic 4.10 exceeded the prespecified stopping boundary of 2.87.
- Primary composite (death or symptomatic stroke): 11/109 (10.1%) medical vs 35/114 (30.7%) interventional; HR 0.27 (95% CI 0.14–0.54), p<0.0001; as-treated HR 0.19 (0.09–0.38), p<0.0001.
- Death or mRS ≥2 at 30 months: 15.1% (8/53) medical vs 46.2% (24/52) interventional (RR 0.33, 95% CI 0.16–0.66); at 36 months: 14.0% vs 38.6% (RR 0.36, 95% CI 0.16–0.83).
- Neurological adverse events: total strokes 45 (interventional) vs 12 (medical), p<0.0001; non-stroke neurological deficits 14 vs 1, p=0.0008.
- Spontaneous AVM rupture rate in the medical arm was low: 2.2% per year (95% CI 0.9–4.5), consistent with modern population-based cohorts.
- Interventional composition: neurosurgery alone n=5, embolisation alone n=30, radiotherapy alone n=31, embolisation+neurosurgery n=12, embolisation+radiotherapy n=15, all three n=1; median time to first intervention 76 days (IQR 45–119); 20 patients had not initiated therapy at analysis.
- ARUBA enrolled a favourable interventional cohort (62% Spetzler-Martin grade ≤2, none >4; 138/223 AVMs <30 mm; all baseline mRS ≤1), yet intervention still worsened outcomes.
- Limitations: non-blinded, mean follow-up only 33 months, early stopping may inflate effect estimate, and many interventional patients were still awaiting or undergoing treatment at analysis.
Study Design
- Study Type
- Prospective, multicentre, parallel-design, non-blinded, randomised controlled superiority trial (with a non-inferiority secondary aim)
- Randomization
- Yes
- Blinding
- Non-blinded (open-label) for patients, treating clinicians, and site investigators. A senior study neurologist not involved in interventional procedures performed the clinical outcome assessment; all primary and secondary events and imaging were reviewed by an independent multidisciplinary international adjudication committee.
- Sample Size
- 226
- Follow-up
- Mean 33.3 months (SD 19.7; median 33.2, IQR 16.3–49.8) at primary analysis; observational phase planned to continue for an additional 5 years
- Centers
- 39
- Countries
- 9 countries (site list not enumerated in the primary manuscript)
Primary Outcome
Definition: Time to composite of death from any cause or symptomatic stroke (haemorrhage on CT/MRI/CSF, or new ischaemic lesion on CT or DWI/T2/FLAIR MRI, with any new focal deficit, seizure, or new-onset headache)
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| 11/109 (10.1%) | 35/114 (30.7%) | 0.27 (0.14–0.54) | <0.0001 |
Limitations & Criticisms
- Non-blinded (open-label) design — although outcome events were objective and centrally adjudicated.
- Early stopping at the second interim (only 53% of planned outcome information) may inflate the effect estimate for medical management.
- Short follow-up (~33 months) relative to a lifelong disease; any delayed benefit of eradication (e.g., after successful radiosurgery latency) could accrue only later.
- 20/116 interventional patients had not initiated therapy and 53 had ongoing treatment plans at analysis — because procedural harm clusters near treatment, the near-term comparison may not reflect steady-state outcomes.
- Small final sample (226 vs original 800 target) — underpowered to compare interventional modalities (surgery vs embolisation vs radiotherapy) or to detect modality-specific safety.
- Enrolled a highly selected 'optimum' interventional cohort (62% Spetzler-Martin ≤2, none >4, all baseline mRS ≤1); generalisability to larger/higher-grade AVMs is uncertain.
- Modest enrolment yield: 226 randomised of 1740 screened (13%); 63% of eligible patients agreed to randomisation — selection bias possible.
- Definition of symptomatic stroke included new-onset headache with imaging changes, which could differentially capture events in interventional patients undergoing more surveillance imaging.
- No standardisation of interventional technique across sites — investigators used usual practice, so results reflect real-world heterogeneity rather than a single defined intervention.
Citation
Lancet 2014;383(9917):614–621. DOI: 10.1016/S0140-6736(13)62302-8