sST2与NT-proBNP乘积及比值指标与心力衰竭患者NYHA心功能分级和LVEF ≥ 50%相关表型的关联分析

Association Analysis of Product and Ratio Indices Derived from sST2 and NT-proBNP with NYHA Functional Class and Phenotypes Related to LVEF ≥ 50% in Patients with Heart Failure

摘要: 目的：探讨sST2与NT-proBNP乘积指标对心力衰竭患者NYHA III~IV级心功能状态的识别价值，并探索sST2/NT-proBNP比值与LVEF ≥ 50%相关表型的关联。方法：回顾性纳入心力衰竭住院患者111例，按NYHA心功能分级分为NYHA II级和NYHA III~IV级，以后者作为主要心功能严重程度结局。LVEF ≥ 50%定义为保留射血分数相关表型，hs-CRP ≥ 10 mg/L定义为hs-CRP升高的探索性界值。构建sST2 × NT-proBNP、ln[sST2 × NT-proBNP]及ln[sST2/NT-proBNP]指标。采用ROC曲线、logistic回归、限制性立方样条、分层及敏感性分析评价其与NYHA III~IV级及相关表型的关联；模型调整年龄、性别和肌酐，并以CKD-EPI 2021公式计算的eGFR替代肌酐进行肾功能敏感性分析。结果：111例患者中，NYHA II级30例、NYHA III~IV级81例。sST2、NT-proBNP及sST2 × NT-proBNP识别NYHA III~IV级心功能状态的AUC分别为0.776 (95% CI: 0.688~0.864)、0.874 (95% CI: 0.811~0.938)和0.895 (95% CI: 0.838~0.953)。DeLong检验显示，乘积指标AUC高于sST2 (P < 0.001)，但与NT-proBNP比较差异无统计学意义(P = 0.430)。调整年龄、性别和肌酐后，ln[sST2 × NT-proBNP]每升高1个标准差与NYHA III~IV级相关(OR = 10.83, 95% CI: 3.981~29.44, P < 0.001)；以eGFR替代肌酐后结果稳定(OR = 10.68, 95% CI: 3.948~28.87, P < 0.001)。Bootstrap内部验证显示调整模型校正后AUC为0.869。ln[sST2/NT-proBNP]升高与LVEF ≥ 50%相关(OR = 1.839, 95% CI: 1.142~2.962, P = 0.012)，但与NYHA III~IV级呈反向关联。在107例完整病例中，临床基础模型加入ln[sST2 × NT-proBNP]的连续型NRI为1.176 (95% CI: 0.802~1.683)，IDI为0.403 (95% CI: 0.232~0.600)；在含NT-proBNP的基础模型中再加入sST2时，连续型NRI为0.025 (95% CI: −0.543~1.238)，IDI为0.019 (95% CI: −0.001~0.123)。结论：sST2 × NT-proBNP对识别心力衰竭患者NYHA III~IV级心功能状态具有一定探索性判别价值；其判别效能优于单独sST2，但未显示出相较NT-proBNP的显著AUC增益。sST2/NT-proBNP比值不宜作为NYHA心功能严重程度的正向判别指标，但其与LVEF ≥ 50%相关表型的关联提示，该比值可能反映“sST2相对NT-proBNP突出”的生物标志物模式。NRI/IDI提示乘积相对于临床基础模型具有增量判别价值，但sST2在NT-proBNP基础上的额外增益有限。上述结果仍需在更大样本及外部队列中验证。

Abstract: Objective: To evaluate the discriminatory value of the sST2 × NT-proBNP product index for identifying NYHA class III~IV functional status in patients hospitalized with heart failure, and to explore the association between the sST2/NT-proBNP ratio and LVEF ≥ 50% related phenotype. Methods: We retrospectively included 111 hospitalized patients with heart failure. According to NYHA functional class, the patients were divided into NYHA class II and class III~IV groups, with the latter serving as the primary outcome for cardiac functional severity. LVEF ≥ 50% was used as an operational definition of a preserved-ejection-fraction related phenotype, and hs-CRP ≥ 10 mg/L as an exploratory inflammatory phenotype. The sST2 × NT-proBNP product, ln[sST2 × NT-proBNP], and ln[sST2/NT-proBNP] were constructed. Associations with NYHA class III~IV and related phenotypes were evaluated using ROC curve analysis, logistic regression, restricted cubic spline analyses, stratified analyses, and sensitivity analyses. Models were adjusted for age, sex, and creatinine. In the renal-function sensitivity analysis, creatinine was replaced by eGFR calculated using the 2021 CKD-EPI equation. Results: Of 111 patients, 30 had NYHA class II and 81 had NYHA class III~IV. The AUCs of sST2, NT-proBNP, and sST2 × NT-proBNP for identifying NYHA class III~IV were 0.776 (95% CI: 0.688~0.864), 0.874 (95% CI: 0.811~0.938), and 0.895 (95% CI: 0.838~0.953), respectively. The product index outperformed sST2 (DeLong P < 0.001), but not NT-proBNP (P = 0.430). After adjustment for age, sex, and creatinine, each SD increase in ln[sST2 × NT-proBNP] was associated with NYHA class III~IV (odds ratio, 10.83; 95% CI: 3.981~29.44; P < 0.001); the association remained stable after substitution of eGFR for creatinine (odds ratio, 10.68; 95% CI: 3.948~28.87; P < 0.001). The bootstrap-corrected AUC was 0.869. Higher ln[sST2/NT-proBNP] was associated with the LVEF ≥ 50% related phenotype (odds ratio, 1.839; 95% CI: 1.142~2.962; P = 0.012) but was inversely associated with NYHA class III~IV. In 107 complete cases, adding ln[sST2 × NT-proBNP] to the clinical model yielded a continuous NRI of 1.176 (95% CI: 0.802~1.683) and IDI of 0.403 (95% CI: 0.232~0.600); adding sST2 to a model already containing NT-proBNP yielded a continuous NRI of 0.025 (95% CI: −0.543~1.238) and IDI of 0.019 (95% CI: −0.001~0.123). Conclusion: The sST2 × NT-proBNP product showed exploratory discriminatory value for NYHA class III~IV functional status and outperformed sST2, but did not provide a significant AUC gain over NT-proBNP. The sST2/NT-proBNP ratio was not a positive marker of functional severity, but may reflect a biomarker pattern characterized by relatively higher sST2 in patients with an LVEF ≥ 50% related phenotype. NRI/IDI supported incremental discrimination of the product beyond the clinical model, but showed limited additional improvement of sST2 beyond NT-proBNP. Larger external cohorts are required for validation.