Nomogram Predicts Parenchymal Hematoma After EVT

Key Takeaways

• In a single-center Chinese cohort of adults with large-core anterior circulation stroke treated with EVT, parenchymal hematoma was observed in 19.5% of patients.
• The final nomogram incorporated baseline NIHSS, monocytes, platelets, post-thrombectomy fasting blood glucose, and number of thrombectomy passes.
• The five-variable nomogram showed good discrimination and acceptable calibration after internal validation.
• More thrombectomy passes were associated with higher PH risk only in patients who did not receive bridging IVT in an unadjusted exploratory subgroup analysis, and the interaction was statistically significant.
• PH was associated with lower adjusted odds of 90-day functional independence, while adjusted links with independent ambulation and mortality were not significant.

Hemorrhagic transformation remains a major concern after endovascular thrombectomy (EVT) in patients with large-core anterior circulation infarction, and parenchymal hematoma (PH) is the severe subtype examined here. The unresolved question is which markers of baseline stroke severity, early laboratory status, and procedural burden during thrombectomy tracked most closely with PH after reperfusion. A retrospective cohort from a Chinese stroke center examined those candidate markers.

In a retrospective single-center cohort at Xuanwu Hospital, Capital Medical University, investigators enrolled 133 consecutive adults from March 2018 to December 2024, as described by Li et al. in CNS Neuroscience & Therapeutics. Eligible patients had anterior-circulation acute ischemic stroke with internal carotid artery or middle cerebral artery M1/M2 occlusion and large-core infarction on baseline non-contrast computed tomography (NCCT), defined by an Alberta Stroke Program Early CT Score (ASPECTS) of 0 to 5, and all underwent EVT. Patients were excluded for missing follow-up computed tomography or magnetic resonance imaging within 48 hours or at neurologic deterioration, or for a pre-event modified Rankin Scale (mRS) score greater than 2. Follow-up imaging at 24 to 48 hours after EVT, or earlier with neurologic deterioration, was classified by European Cooperative Acute Stroke Study II (ECASS II) criteria, with PH comprising PH1 and PH2. Predictor selection used univariable screening, least absolute shrinkage and selection operator with 5-fold cross-validation, multivariable logistic regression, and 1000-bootstrap internal validation, and 90-day outcomes came from structured telephone follow-up. This Chinese single-center cohort bounds the model's geographic and practice-setting generalizability.

Parenchymal hematoma developed in 26 of 133 patients. The final model retained baseline National Institutes of Health Stroke Scale (NIHSS) score, monocyte count, platelet count, post-thrombectomy fasting blood glucose, and number of passes. Among the most clinically prominent predictors, post-thrombectomy fasting blood glucose had an odds ratio of 1.326 (95% CI 1.121-1.616; p=0.002) and number of passes had an odds ratio of 2.312 (95% CI 1.443-3.990; p=0.001), while higher baseline NIHSS and lower monocyte and platelet counts also entered the nomogram.

The nomogram showed good discrimination, with an area under the curve of 0.876 and a bootstrap-adjusted area under the curve of 0.844, and calibration was described as acceptable after internal validation. Investigators also reported a significant interaction between number of passes and intravenous thrombolysis (IVT) (p for interaction=0.020), with the higher-pass association seen only in the subgroup that did not receive bridging IVT. In adjusted outcome analyses, PH remained associated with lower 90-day functional independence, while independent ambulation and mortality were not significantly associated.

Because the data came from a retrospective single-center cohort, local practice patterns and selection effects may have influenced the observed associations. The sample size and PH event total were modest, and the pass-by-IVT finding was explored in unadjusted subgroup models because subgroup events were limited. The authors interpreted the interaction as a signal that procedural burden may matter differently in direct thrombectomy, but the model had internal bootstrap validation only and no external validation cohort.

The authors concluded that a five-variable nomogram was established and internally validated for PH risk estimation after EVT in large-core anterior circulation infarction, with good performance within the derivation cohort. They called for external validation in multicenter prospective cohorts before broader testing across other stroke populations.

Clinician Questions

Which stroke patients does this PH nomogram apply to after EVT?

The nomogram applies to adults aged 18 years or older with anterior-circulation acute ischemic stroke, internal carotid artery or middle cerebral artery M1/M2 occlusion, and large-core infarction defined on baseline NCCT by ASPECTS 0 to 5 who underwent EVT at a single Chinese stroke center. Patients with a pre-event mRS score greater than 2 or without follow-up imaging for hemorrhage assessment were excluded, which narrows how broadly the model can be generalized.

Why did the number of thrombectomy passes appear to matter more without prior IVT?

Investigators observed a significant interaction between number of passes and IVT, with higher pass burden associated with PH only in the non-IVT subgroup. In the authors' discussion, IVT may soften thrombi and potentially lessen endothelial trauma during repeated retrieval attempts, although that interaction analysis was unadjusted and exploratory because subgroup event counts were limited.

What still needs to happen before this PH nomogram is tested more broadly?

The model underwent internal bootstrap validation only and was not assessed in an external validation cohort. Its retrospective single-center design and modest PH event count may limit precision and transportability, and the authors called for validation in independent multicenter prospective cohorts.