Integrating clinical predictors and glial fibrillary acidic protein in prediction models for the prehospital identification of intracerebral haemorrhage in suspected stroke.

Abstract:

Introduction: Accurate and timely differentiation of intracerebral haemorrhage (ICH) from other suspected stroke cases is crucial in prehospital settings, where early blood pressure reduction in the ambulance can improve outcomes. This study aims to assess whether combining clinical predictors and glial fibrillary acidic protein (GFAP) in prediction models can effectively distinguish ICH from other suspected stroke cases.

Methods: Data were derived from the Testing for Identification Markers of Stroke trial, a prospective diagnostic accuracy study. Suspected stroke patients within 6 hours of symptom onset were included. Clinical predictors were selected based on known associations with ICH, and a predefined GFAP cut-off of 290 pg/mL was applied. Logistic regression was used to assess the performance of clinical predictors and GFAP, individually and in combination. Internal validation and optimism correction were performed via bootstrapping, and comparisons of the area under the curve (AUC) were conducted using DeLong's test.

Results: We included 209 suspected stroke cases, of which 5% were finally diagnosed with ICH. Clinical predictors alone achieved an optimism-corrected AUC of 0.74 (95% CI 0.60 to 0.88), while GFAP alone resulted in an optimism-corrected AUC of 0.83 (95% CI 0.69 to 0.99). Combining clinical predictors with GFAP significantly enhanced the AUC, yielding an optimism-corrected value of 0.90 (95% CI 0.79 to 0.98). This combined model also demonstrated high predictive accuracy, with an optimism-corrected sensitivity of 60% (95% CI 29.0% to 90.0%) and a specificity of 98% (95% CI 96.1% to 100.0%).

Conclusions: Combining clinical predictors with GFAP shows promise for the prehospital identification of ICH to support transport decision-making and potentially initiate treatment while en route for these patients. Prospective validation using portable point-of-care devices is required to confirm the utility of this approach in the prehospital setting.