Appraising non-HDL-C, systolic pressure, and a nomogram-based diagnostic model as auxiliary biomarkers in confirming acute ischemic stroke and transient ischemic attack

Abstract

Acute ischemic stroke (AIS) is characterized by the abrupt onset of neurological dysfunction stemming from focal brain ischemia, confirmed through imaging evidence of infarction. In contrast, transient ischemic attack (TIA) manifests with neurological deficits in the absence of infarction, with imaging serving as the definitive diagnostic criterion. This study aims to assess the diagnostic value of combining non-high-density lipoprotein cholesterol (non-HDL-C) and blood pressure (BP) in differentiating AIS from TIA. We recruited 207 untreated AIS patients diagnosed within 72 h and 99 age- and gender-matched TIA patients. Upon admission, serum non-HDL-C levels, other lipid profiles, and BP measurements were obtained. Binary logistic regression was employed to identify risk factors, while receiver operator characteristic (ROC) curves were used to evaluate diagnostic performance. Furthermore, least absolute shrinkage and selection operator (LASSO) regression coupled with multivariate logistic regression was utilized to develop a nomogram model. The AIS group exhibited higher prevalence rates of hypertension, diabetes, family history of diabetes, and smoking (P < 0.05). Notably, non-HDL-C, systolic BP, diastolic BP, and other lipid markers were significantly elevated in the AIS group (P < 0.05). Multivariate analysis pinpointed non-HDL-C (OR [odds ratio] = 1.663, 95% CI [confidence interval]: 1.239-2.234, P < 0.01) and systolic BP (OR = 1.035, 95% CI: 1.012-1.057, P < 0.01) as independent risk factors. ROC analysis revealed that systolic BP alone achieved an AUC of 0.686 (sensitivity: 78.7%, specificity: 51.5%), whereas the combination of systolic BP and non-HDL-C enhanced diagnostic accuracy (AUC [area under the ROC curve] = 0.736, sensitivity: 75.4%, specificity: 64.6%). A nomogram incorporating low-density lipoprotein cholesterol (LDL-C), glucose (GLU), homocysteine, and smoking demonstrated high predictive accuracy, with training and validation AUCs of 0.769 and 0.704, respectively. Non-HDL-C and systolic BP emerge as independent risk factors for AIS, and their combined use augments diagnostic precision in differentiating AIS from TIA. A nomogram model presents a practical differentiation tool, particularly in settings with limited resources.

Keywords: Acute ischemic stroke; Non-high density lipoprotein cholesterol; Systolic pressure; Transient ischemic attack.

Fig. 1

Comparison of 10 indicators including (A) TG, (B) TC, (C) HDL-C, (D) LDL-C, (E) GLU, (F) IMA, (G) Hcy, (H) non-HDL-C, (I) systolic pressure, and (J) diastolic pressure between AIS and TIA groups. ns: not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 2

Comparison of indices involved (A) TG, (B) TC, (C) HDL-C, (D) LDL-C, (E) GLU, (F) IMA, (G) Hcy, (H) non-HDL-C, (I) systolic pressure, and (J) diastolic pressure between two groups based on different onset time of AIS. ns: not significant.

Fig. 3

The diagnostic efficacy of various indices, either single or in combination, in differentiating AIS from TIA.

Fig. 4

Predictive model distinguishing AIS from TIA based on the LASSO algorithm. (A, B) Variable Selection. (C, D) Outcome of LASSO regression for vital parameter. Using a tenfold cross-validation approach, the coefficient lambda is minimized based on the criterion of the smallest standard deviation, ultimately selecting clinical indicators with non-zero coefficients. In the cross-validation process, the function of the binomial deviance values is represented by log(lambda), with the Y-axis depicting binomial deviance values. The lower X-axis represents log (lambda), while the upper X-axis shows the average number of parameters. (E) Columnar graph of the predictive model. (F) Diagnostic AUCs for the training and validation sets. (G) Calibration Curves for the training and validation Sets. (H) Clinical Decision Curves for the training and validation sets.