目的： 通过分析新的血清标志物与钙化性主动脉瓣膜病（calcific aortic valve disease,CAVD）关系,建立新的生物标志物预测模型,用于预测CAVD。方法： 收集446例心内科住院患者外周血及临床资料,通过SPSS20.0将患者按简单随机法分为筛查队列（n=202）和验证队列（n=244）,酶联免疫吸附法检测胰高血糖素样肽1（glucagon-like peptide-1,GLP-1）和骨桥蛋白(osteopontin, OPN)血清浓度。通过分析筛查队列中各因子及临床变量与CAVD的关系,其后在验证队列中加以验证,构建CAVD预测模型。结果： 二元回归分析显示,筛查队列中多个因素与CAVD显著相关,最终确定高密度脂蛋白胆固醇（high density lipoprotein cholesterol,HDL-C）、GLP-1及OPN纳入预测模型,并绘制列线图。通过C统计量[筛查队列：0.73（95% 置信区间：0.66~0.80）,验证队列0.70：（95% 置信区间：0.64~0.77）]及Hosmer-Lemeshow检验（筛查队列：P=0.14,验证队列：P=0.23）发现在筛查和验证队列中模型的区分度和一致性均表现良好。决策曲线分析显示,生物标志物模型比临床因素模型具有更高的净效益。此外,该模型在不同性别及年龄组仍具良好鉴别性。结论： GLP-1、OPN及HDL-C水平与CAVD具明显相关性。基于此,成功构建并验证了一种CAVD预测模型,该模型鉴别能力和准确性均表现良好,为预测CAVD提供可能。

Objective To establish a new biomarker predictive model for calcific aortic valve disease (CAVD) by analyzing the relationship between new serum markers and CAVD. Methods The peripheral blood and clinical data of 446 in-patients in the department of cardiology were collected. The patients were randomly divided into a screening cohort (n=202) and a verification cohort(n=244) by SPSS20.0. The enzyme-linked immunosorbent assay was used to detect the serum concentrations of glucagon-like peptide 1(GLP-1) and osteopontin (OPN). By analyzing the relationship between multiple factors and clinical variables in the screening cohort and CAVD, and verifying it in the verification cohort, the CAVD predictive model was constructed. Results The binary regression analysis showed that multiple factors in the screening cohort were significantly related to CAVD. Based on it, it was determined that high density lipoprotein cholesterol (HDL-C), GLP-1 and OPN could be included in the predictive model, and a nomogram was drawn. Through Pass C statistics [screening cohort: 0.73(95% CI: 0.66-0.80), verification cohort 0.70 (95% CI: 0.64-0.77)] and Hosmer-Lemeshow test (screening cohort: P=0.14, verification cohort: P=0.23), it was found that the discrimination and consistency of the model in the screening and validation cohorts performed well. Decision curve analysis indicated that the biomarker model had a higher net benefit than that in the clinical factor model. In addition, the model showed nice discriminating ability in different gender and age groups. Conclusions GLP-1, OPN and HDL-C levels are significantly correlated with CAVD. Based on it, a CAVD predictive model which performed well in discriminating ability and accuracy was successfully constructed and verified, and provided the possibility for CAVD prediction.