Predicting futile recanalization after mechanical thrombectomy in patients with ischemic stroke and cancer: the CACTUS nomogram.
TL;DR
The CACTUS prediction model showed adequate predictive performance for FR, and the derived nomogram may serve as a practical tool for individualized risk assessment in patients with LVO who are candidates for MT and have a history of cancer.
Abstract
Background
Patients with large vessel occlusion (LVO) and a history of cancer have worse long-term outcomes after mechanical thrombectomy (MT) despite similar reperfusion rates. We aimed to develop a clinical prognostic score to predict futile recanalization (FR) in this population.
Methods
We retrospectively analyzed a single-center prospective registry (ARTISTA) of patients with LVO and active or previous cancer who underwent MT between January 2021 and December 2023. The primary outcome was FR, defined as successful angiographic reperfusion (expanded Thrombolysis in Cerebral Infarction (eTICI) ≥2b) with poor functional outcome (modified Rankin Scale (mRS) ≥3) at 90 days. A clinically prespecified logistic regression model was developed to predict FR. Model performance was assessed, and internal validation was performed using bootstrap resampling and repeated cross-validation. An expanded penalized least absolute shrinkage and selection operator (LASSO) model was evaluated as a secondary analysis and a nomogram was derived from the final model.
Results
Of 1584 patients who underwent MT during the study period, 167 patients with a history of cancer met the eligibility criteria for analysis. FR occurred in 50.9% of patients. A clinically prespecified prediction model included age, pre-stroke mRS score, admission National Institutes of Health Stroke Scale (NIHSS) score, and metastatic disease. A higher pre-stroke mRS score (adjusted OR (aOR) 1.84, 95% CI 1.09 to 3.21; P=0.025), higher admission NIHSS score (aOR 1.11 per point, 95% CI 1.05 to 1.17; P<0.001), and metastatic disease (aOR 8.38, 95% CI 2.77 to 31.93; P<0.001) were independently associated with FR. Internal bootstrap validation yielded an optimism-corrected AUC of 0.758 and a Brier score of 0.206. An optimism-corrected nomogram was derived for clinical use.
Conclusion
The CACTUS prediction model showed adequate predictive performance for FR, and the derived nomogram may serve as a practical tool for individualized risk assessment in patients with LVO who are candidates for MT and have a history of cancer.