RAPID IMPROVE is a seamless design which integrates dose selection (Phase IIb) and confirmatory testing (Phase III) into a single continuous trial. This approach preserves study power, controls the overall type I error rate, and accelerates the evaluation of rapid local ischemic postconditioning (RL-IPostC) in patients with acute anterior circulation large-core ischemic stroke who achieve successful recanalization defined as modified Thrombolysis In Cerebral Infarction (mTICI) 2b/3 after mechanical thrombectomy.
Randomization is performed centrally using a web-based system immediately after confirmation of successful recanalization. In Phase IIb, eligible patients are allocated 1:1:1 to high-dose RL-IPostC, low-dose RL-IPostC, or control (no postconditioning). After the dose selection decision at the end of Phase IIb, Phase III continues with a 1:1 randomization between the selected dose and control. Randomization is stratified by stroke onset to imaging time (<6 hours vs. 6-24 hours) and baseline NIHSS score (<17 vs. ≥17) to ensure balance of prognostic factors.
The treating interventionalist cannot be blinded due to the nature of the postconditioning procedure. However, all outcome assessors - including the investigators who perform follow-up mRS assessments, the imaging core laboratory personnel, and the clinical events committee - are blinded to treatment assignment. The Data Safety Monitoring Board (DSMB) has access to unblinded data for safety monitoring. An imaging core laboratory centrally evaluates all baseline and follow-up CT, CTA, CTP, MRI, and DSA images. Separate analysts interpret baseline eligibility images (ASPECTS, core volume) and post-procedure images (midline shift, infarct volume, net water uptake, CSF/ICV ratio), all blinded to treatment allocation.
RL-IPostC is performed using a balloon guide catheter (BGC) positioned at the C1-C2 segment of the ipsilateral internal carotid artery, carefully avoiding the carotid sinus. The intervention should be initiated as soon as possible after confirmation of successful recanalization, and no later than 5 minutes. All patients receive standard post-thrombectomy care in a stroke unit, including a mandatory repeat CT scan at 24 hours. Antithrombotic or antiplatelet therapy is given according to local clinical practice and is recorded.
The trial uses a 3-arm, 2-stage multi-arm multi-stage (MAMS) design. The primary endpoint for stage 1 (Phase IIb) is the absence of midline shift (>3 mm at the level of the septum pellucidum) on 24 hours imaging (non-contrast computed tomography or magnetic resonance imaging). Based on the research team's cohort study, the control arm response rate (absence of midline shift) is assumed to be 40%, with an absolute improvement of 30% for an effective dose. One-sided α = 0.10, power = 95%.
The primary endpoint for stage 2 (Phase III) is the proportion of patients with mRS 0-3 at 90 days post-procedure. The control arm response rate is assumed to be 40% (derived from SELECT2 trial results), with an absolute improvement of 20% for the selected dose. One-sided α = 0.025, power = 85%.
Overall family-wise error rate (FWER) is controlled at one-sided 0.045. The overall pairwise α is 0.025 (one-sided) and the overall pairwise power is 80.7%. A positive predictive value (probability of mRS 0-3 given absence of midline shift) is prespecified as 0.36 for control and 0.60 for treatment, based on the same cohort study, to appropriately link the two stages.
Sample size was calculated using the nstagebin function (version 1.0.2) in Stata. The maximum planned sample size is 288 patients assuming 10% dropout across both stages. Stage 1 (Phase IIb) requires 48 patients per arm (144 total). Stage 2 (Phase III) adds 72 patients per arm (144 total for two arms), accumulating to 120 evaluable patients per arm for the selected dose and control.
An adaptive sample size increase is permitted in Phase III. At Interim Analysis #3 (84 patients enrolled per arm in Phase III, i.e., 168 patients), a blinded statistician will calculate the conditional power for the primary endpoint (mRS 0-3) using the method of Mehta and Pocock. If the conditional power falls within a prespecified "promising" range, the sample size may be increased up to a maximum of 200 patients per arm in Phase III, with the goal of raising conditional power to 0.80. The total sample size including this adaptation is capped at 448 patients (200 per arm in Phase III plus the non-selected arm patients from Phase IIb).
Three protocol-specified interim analyses are scheduled:
IA1 (72 patients, 24 per arm): early safety review. IA2 (144 patients, 48 per arm): end of Phase IIb - dose selection and safety assessment.
IA3 (216 patients: 84 control, 84 selected intervention, 48 from the non-selected arm): conditional power for sample size re-estimation.
Safety stopping rules follow a two-step frequentist confidence approach for symptomatic intracranial hemorrhage (sICH) and 3-month mortality.
For sICH:
Step 1 - If the confidence that the sICH rate in an intervention arm exceeds 20% is >90%, proceed to Step 2.
Step 2 - If the confidence that the excess sICH rate over the control arm is ≥5% exceeds 90%, enrollment is placed on hold and the DSMB convenes to determine continuation.
For mortality:
Step 1 - Threshold 40% (confidence >90%). Step 2 - Excess over control ≥10% (confidence >90%).
Dose selection rules at end of Phase IIb (efficacy boundary p < 0.1, one-sided, for the imaging endpoint):
Only low dose crosses boundary → select low dose. Only high dose crosses boundary → select high dose. Both doses cross → select the dose with lower 3-month mortality. Neither cross → pause enrollment; DSMB meets to discuss possible trial termination.
A clinical events committee of three independent experts adjudicates all complications, including sICH (using the SITS-MOST definition: parenchymal hematoma type 2 with an NIHSS worsening of ≥ 4 points compared to the pre-deterioration assessment) and other serious adverse events.
