Sequential decision-making in endovascular stroke therapy: benefit and harm of the next retrieval attempt after...

After failed reperfusion in endovascular thrombectomy (EVT), interventionalists must outweigh the expected benefit of an additional retrieval attempt with the outcome of the current angiographic result. Framing EVT as an infinite-horizon optimal stopping problem, the present study provides a per-pass evaluation of the expected benefit of EVT continuation versus procedural termination after unsuccessful reperfusion, based on large-scale real-world registry data.

Patients from the German Stroke Registry - Endovascular Treatment (GSR-ET) between 2015 and 2023 were screened. Inclusion criteria were (1) age ≥ 18 years, (2) anterior circulation LVO, (3) 1-15 retrieval attempts and (4) availability of relevant clinical data. Primary endpoint was the rate of good functional outcome, defined as modified Rankin Scale (mRS) score 0-2 at 90 days. For each retrieval attempt, patients were stratified into successful reperfusion (mTICI 2b-3) and persistent occlusion (mTICI 0-2a). The net expected benefit of EVT continuation, Δ(k), was defined as the difference between the mRS 0-2 rate at pass k + 1, and the mRS 0-2 rate for unsuccessful reperfusion at pass k. A positive Δ(k) indicates that continuation is favored at that decision node k. The pass at which Δ(k) first becomes non-significant identifies the empirical, optimal stopping threshold.

Among 9,495 included patients, successful reperfusion (mTICI 2b-3) was achieved in 8,405 (89%), and 3,428 (36%) achieved good functional outcome at 90 days. Up to six retrieval attempts, Δ(k) was positive and statistically significant (p < 0.05), indicating that continuation of EVT was associated with superior expected outcome at each of these decision points. At pass seven, Δ(k) became negative and lost statistical significance (p = 0.53), identifying this as the empirical optimal stopping threshold.

Continuation of EVT up to six retrieval attempts confers a significant functional benefit over termination after unsuccessful reperfusion. These findings provide a decision-theoretic basis for individualized intraprocedural decision-making and suggest that functional independence can be achieved beyond conventionally assumed pass thresholds.