Development and validation of a prognostic prediction model for antithrombotic-related chronic subdural hematoma in patients with recent acute myocardial infarction.

Abstract

Antithrombotic therapy is a cornerstone of acute myocardial infarction (AMI) treatment and is thought to be associated with an increased risk of chronic subdural hematoma (CSDH). However, no well-established model exists to predict subsequent antithrombotic treatment outcomes after CSDH in patients with recent AMI. We aimed to identify a prognostic model to predict the 6-month outcome of treatment with antithrombotic therapy. This multicenter retrospective analysis involved 553 patients with recent AMI with antithrombotic-related CSDH. Several candidate clinical variables and biomarkers were examined in the training cohort (Chengdu training cohort; n=368). Patients with unfavorable outcomes had experienced at least 1 of the following: major adverse cardiovascular events (MACE), recurrence, or a modified Rankin scale (mRS) score of 2 to 6. To develop a 6-month outcome prediction model, three approaches were used: (I) a demographic variable model, (II) a clinical marker model and (III) a decision-driven model. A clinical outcome prediction model based on the superior predictors was assessed by logistic regression analysis. The nomogram for the final model was internally validated using a bootstrap procedure and externally validated in an independent cohort (Anhui cohort; n=185). Model A produced 7 predictors of unfavorable outcomes, while models B and C yielded 2 and 1 predictors, respectively. The areas under the curve (AUC) increased from 0.743 [model A; 95% confidence interval (CI): 0.680-0.782] to 0.889 (model A + B + C; 95% CI: 0.851-0.916). The final prediction model included age, systolic blood pressure (SBP), body mass index (BMI), the Glasgow Coma Scale (GCS), the estimated glomerular filtration rate (eGFR), the early resumption of antithrombotic therapy, hematoma thickness and the presence of abdominal obesity, frailty and previous bleeding. Internal and external validation of the selected final model revealed adequate C-statistics and calibration slope values (internal validation: 0.81 and 0.78; external validation: 0.80 and 0.76, respectively). This model provided a risk stratification tool to predict unfavorable outcomes in patients with recent AMI with antithrombotic-related CSDH. Because the study was based on ten readily practical and available variables, it may be widely applicable to guide management and complement clinical assessment.