Introduction/Purpose The treatment effect of mechanical thrombectomy (MT) for acute basilar occlusion (ABO) on patient outcomes is uncertain. We used Bayesian multilevel models to synthesize the evidence from the four randomized clinical trials (RCTs) of MT for ABO to estimate the aggregate treatment effect (ATE) of this intervention and predict the consequences for the general patient population.

Materials and Methods A Bayesian multilevel generalized logistic model with trials as clusters and treatment (MT) versus control (best medical management) as correlated cluster-level features was fit to individual patient-level data from the four RCTs. Posterior distributions for model parameters were used to directly estimate distributions for individual probabilities of specific 90-day outcomes, and posterior predictive simulation using these distributions was performed to determine the predicted difference in individual outcomes per 100 patients treated. Models were fit in STAN using the R package rethinking. ATEs are presented as odds ratios (OR) with 95% compatibility intervals (CI). Ambulation was interpreted as modified Rankin score (mRS) 0-3, functional independence as mRS 0-2, and total dependence as mRS 5, all at 90 days.

Results The mean ATE of MT for ABO was 1.90 (95% CI 1.16-2.86) for ambulation, 1.99 (95% CI 0.95-3.90) for functional independence, and 1.01 (95% CI 0.61-1.80) for total dependence. The predicted difference in ambulation was 15 individuals in favor of MT per 100 treated (95% CI -2 to 31). The predicted difference in functional independence was 14 individuals in favor of MT per 100 treated (95% CI -5 to 34). The predicted difference in total dependence was 0 individuals per 100 treated (95% CI -11 to 12). The predicted probability that MT results in at least one additional ambulatory patient per 100 treated compared to medical management alone was 95%. The predicted probability that MT results in at least one additional functionally independent patient per 100 treated, when compared to medical management alone, was 92%.

Conclusion Aggregating evidence from 4 randomized trials of MT for ABO using Bayesian techniques demonstrates that MT is likely to have a clinically substantial effect on ambulation and functional independence at 90 days after ABO, without incurring increased risk of total dependence.

Disclosures B. Kolb: None. S. Munich: None.