Hyperlipidemia is a critical element in the etiology of acute ischemic stroke (AIS). However, the potential link connecting the atherogenic index of plasma (AIP) with spontaneous hemorrhagic transformation (sHT) remains obscure, especially in patients managed without reperfusion. We explored the relationship between AIP and sHT in AIS patients and utilized machine learning algorithms to evaluate predictive risk factors for sHT. We reviewed 3,116 AIS patients not receiving intravenous thrombolysis or endovascular thrombectomy. AIP was computed as log10(TG/HDL-C). The outcome was sHT. Associations were analyzed using multivariable logistic regression and restricted cubic spline (RCS) models. Furthermore, we applied 5 machine learning models and used Shapley Additive Explanations (SHAP) to interpret predictor contributions. sHT occurred in 200 patients (6.4%). In the fully adjusted model, each 1 SD decrease in AIP indicated higher odds of sHT (OR 1.212, 95% CI 1.037–1.420). The RCS regression demonstrated a linear relationship between decreasing AIP and sHT, and the association was significant in patients aged < 65 years. The association differed by age (P for interaction = 0.001). Machine learning models, particularly the Random Forest (RF) model (AUC: 0.813), provided robust prediction for these endpoints and suggested that AIP contributes to the risk of sHT. Our findings suggest that a lower AIP may serve as a potential independent marker for a higher risk of sHT in AIS patients, particularly those younger than 65 years.
Xuan et al. (Sat,) studied this question.