Rescue therapy for vasospasm following aneurysmal subarachnoid hemorrhage: a propensity score-matched analysis with...

Area of Science:

• Neurosurgery
• Neurology
• Data Science

Background:

• Subarachnoid hemorrhage (SAH) can lead to angiographic vasospasm (aVSP) and delayed cerebral ischemia (DCI).
• Rescue therapies are recommended but lack robust evidence from randomized clinical trials regarding safety and efficacy.

Purpose of the Study:

• To determine if rescue therapy improves 3-month outcomes in post-SAH aVSP/DCI patients using explainable machine learning (ML) and propensity score matching.
• To identify patient subgroups more likely to receive rescue therapy and factors influencing outcomes after therapy.

Main Methods:

• Utilized data from 8 clinical trials and 1 observational study within the Subarachnoid Hemorrhage International Trialists repository.
• Developed gradient boosting ML models to predict rescue therapy probability and 3-month Glasgow Outcome Scale (GOS) scores.
• Employed Shapley Additive Explanation (SHAP) values for feature importance and conducted propensity score-matched analysis.

Main Results:

• Identified 1532 patients with aVSP or DCI.
• Aneurysm characteristics and neurological complications were key predictors for rescue therapy administration.
• Rescue therapy was associated with increased odds of favorable 3-month GOS scores (OR 1.63, 95% CI 1.22-2.17).

Conclusions:

• Rescue therapy may enhance the likelihood of good outcomes for patients experiencing aVSP or DCI post-SAH.
• Future trials should focus on interventions targeting cerebral ischemia/infarction to demonstrate clinical improvements.
• Explainable ML insights can refine patient selection and clinical trial design.