Introduction: Although intravenous alteplase(tPA) in acute ischemic stroke(AIS) has been proved as the standard treatment, there remain concerns about increased risk of symptomatic intracranial hemorrhage. Lower dosage of tPA has been used in Japan and other Asian countries. We aim to find which AIS patients benefit from the lower dosage of intravenous tPA.
Methods: We used the TIMS-CHINA database of 1440 stroke patients from 67 centers and the 4S database of 2779 patients from 11 centers. Both databases include AIS stroke patients treated with tPA, recorded with similar medical databases structures. The variables include the severity of stroke at baseline (NIHSS score), onset to needle time, dosage of tPA, symptomatic intracranial hemorrhage(sICH), modified Rankin Scale(mRS) at 3 months. We used a machine-learning method of subgroup analysis using decision trees with recursive portioning.
Results: In this cluster analysis, the appropriate dosage in subsets of acute ischemic stroke patients has been explored. Based on NIHSS score to optimize the interaction effect between t-PA dosage (standard/lower), three subgroups were defined as NIHSS (1) ≤4, (2) 5-14, and (3)≥15. Their risk of sICH and benefit (as mRS0-1) are identified. The estimated difference in log odds of bleeding between standard and lower dosage is: -0.85 (p=0.15) for Group 1, 1.95 (p=0.02) for Group 2, and 0.39 (p=0.56) for Group 3, after effect adjustment of other factors and bias reduction. We are especially careful about bias reduction due to rare events: the number of bleeding cases can be small in certain subgroups. The estimated difference in log odds of efficacy between standard and lower dosage is: 0.16 (p=0.49) for Group 1, -0.16 (p=0.57) for Group 2, and -0.08 (p=0.80) for Group 3, after effect adjustment of other factors.
Conclusion: Our results advocate using a lower dosage of tPA in moderate stroke, which significantly reduce the bleeding risk while achieving non-inferior performance in efficacy