Investigating the Prognostic Potential of Plasma ST2 in Patients with Peripheral Artery Disease: Identification and Evaluation

Abstract

Soluble interleukin 1 receptor-like 1 (ST2) is a circulating protein demonstrated to be associated with cardiovascular diseases; however, it has not been studied as a biomarker for peripheral artery disease (PAD). Using a prospectively recruited cohort of 476 patients (312 with PAD and 164 without PAD), we conducted a prognostic study of PAD using clinical/biomarker data. Plasma concentrations of three circulating proteins [ST2, cytokine-responsive gene-2 (CRG-2), vascular endothelial growth factor (VEGF)] were measured at baseline and the cohort was followed for 2 years. The outcome of interest was a 2-year major adverse limb event (MALE; composite of major amputation, vascular intervention, or acute limb ischemia). Using 10-fold cross-validation, a random forest model was trained using clinical characteristics and plasma ST2 levels. The primary model evaluation metric was the F1 score. Out of the three circulating proteins analyzed, ST2 was the only one that was statistically significantly higher in individuals with PAD compared to patients without PAD (mean concentration in plasma of 9.57 [SD 5.86] vs. 11.39 [SD 6.43] pg/mL, p < 0.001). Over a 2-year period, 28 (9%) patients with PAD experienced MALE. Our predictive model, incorporating clinical features and plasma ST2 levels, achieved an F1 score of 0.713 for forecasting 2-year MALE outcomes. Patients identified as high-risk by this model showed a significantly increased likelihood of developing MALE (HR 1.06, 95% CI 1.02-1.13, p = 0.003). By combining clinical characteristics and plasma ST2 levels, our proposed predictive model offers accurate risk assessment for 2-year MALE in PAD patients. This algorithm supports risk stratification in PAD, guiding clinical decisions regarding further vascular evaluation, specialist referrals, and appropriate medical or surgical interventions, thereby potentially enhancing patient outcomes.

Keywords: biomarkers; peripheral artery disease; prognosis; soluble interleukin 1 receptor-like 1.

Figure 1

Variable importance scores (gain) for the random forest model input features for prognosis of 2-year major adverse limb events in individuals with peripheral artery disease. Abbreviations: soluble interleukin 1 receptor-like 1 (ST2), congestive heart failure (CHF), coronary artery disease (CAD), diabetes mellitus (DM), hypercholesterolemia (HC), hypertension (HTN).

Figure 2

Kaplan–Meier analysis of the freedom from major adverse limb events among patients categorized as low versus high risk by the random forest model. Classification into these risk groups was based on a threshold of 0.60 derived from the Youden Index, optimizing the prediction model’s sensitivity and specificity through receiver operating characteristic curve analysis. Cox proportional hazards analysis adjusted for sex, age, dyslipidemia, hypertension, diabetes, smoking history, coronary artery disease, congestive heart failure, previous stroke, use of statins, acetylsalicylic acid, angiotensin II receptor blocker or angiotensin-converting enzyme inhibitor, calcium channel blocker, beta blocker, furosemide or hydrochlorothiazide, oral antihyperglycemic agent, and insulin. Abbreviations: CI (confidence interval), HR (hazard ratio).