步进法筛选心绞痛患者血清炎症因子建立判别函数的研究

The research of establishing discriminant function for patients with angina pectoris by stepwise analysis based on serum inflammatory factors

目的采用步进法（stepwise）判别分析（discriminantanalysis）建立基于稳定型心绞痛（SAP）及不稳定型心绞痛（UAP）患者血清超敏C-反应蛋白（hs-CRP）、巨噬细胞移动抑制因子（MIF）、白细胞介素（IL-4、IL-10）水平的判别函数（discriminantfunctions），在控制成本的前提下协助对心绞痛患者进行分型。方法选择39例SAP患者、47例UAP患者及39例健康志愿者以构成训练样本（trainingsamples），采用酶联免疫吸附试验（ELISA）检测血清hs-CRP、MIF、IL-4及IL-10水平，分别采用传统的纳入全变量方式及步进法对结果进行判别分析，根据函数考核结果对两种方法建立的判别函数的判别效率进行评估。结果根据纳入全变量的方式对训练样本的年龄以及hs-CRP、MIF、IL-4、IL-10等生物标志物进行判别分析获得判别函数：健康对照组=-129．858＋2．869×年龄-2．451×hs-CRP＋1．393×MIF＋6．001×IL-4＋4．848×IL-10；SAP组=-161．037＋2．896×年龄-2．022Xhs-CRP＋1．662×MIF＋6．703XIL-4＋6．287×IL-10；UAP组=-199．087＋2．468×年龄-1．440×hs-CRP＋3．404×MIF-13．875×IL-4＋7．752×IL-10。对上述函数进行回顾性考核的总错判率为4．8％，交互验证考核的总错判率为5．6％。再采用步进法筛选上述指标，程序建议纳入年龄、MIF及IL-10并建立判别函数：健康对照组=-125．218＋2．659×年龄＋0．599×MIF＋5．040×IL-10；SAP组=-157．864＋2．721×年齿令＋1．008×MIF＋6．468×IL-10；UAP组=-197．327＋2．360×年齿令＋2．932XMIF＋7．640×IL-10。对步进法建立函数进行回顾性考核及交叉验证考核的总错判率均为6．4％。两种方法所建立的判别函数对SAP及UAP患者的判别正确率相当，均达100％。结论步进法判别分析筛选后建立的基于年龄、血清MIF、IL-10的判别函数可协助判别不典型的SAP与UAP患者，具有较优的成本效益。

Objective To improve cost-efficiency, discriminant functions in stepwise method was founded for the differential diagnosis of angina pectoris by detecting the serum level of high-sensitivity C-reactive protein （hs-CRP）, macrophage migration inhibitory factor （MIF）, interleukin-4 （IL-4） and interleukin-10 （IL-IO） in patients with stable angina pectoris （SAP） and unstable angina pectoris （UAP）. Methods Thirty-nine SAP patients and 47 UAP patients were enrolled into the study, while 39 healthy volunteers were enrolled into the controlled group forming the entire set of training samples. The serum levels of hs-CRP, MIF, IL-4 and IL-10 were measured by enzyme linked immunosorbent assay （ELISA）. Data was analyzed by software to define discriminant functions in the ways of ＂entering＂ and ＂stepwise＂. Both functions were evaluated by the results of validation. Results By the way of ＂enter independent together＂, the following discriminant functions were defined based on the data of training samples＇ age, hs-CRP, MIF, IL-4, IL-IO： healthy control group =-129.858 ＋ 2.869 x age -2.451 x hs-CRP ＋ 1.393 x MIF ＋ 6.001 x IL-4 ＋ 4.848 x 1L-IO; SAP group =-161.037 4- 2.896 x age-2.022 x hs-CRP ＋ 1.662 x MIF ＋ 6.703 x IL-4 ＋ 6.287 x IL-10; UAP group =-199.087 ＋ 2.468 x age-1.440 x hs-CRP ＋ 3.404 x MIF-13.875 x IL-4 ＋ 7.752 x 1L-IO. Retrospective validation showed 4.8% of total miss-grouping, while cross-validation showed 5.6% of total miss-grouping. By the way of ＂stepwise＂, the above data was screened by software and training samples＇ age, MIF and IL-IO were suggested to define the following functions： healthy control group = - 125.218 ＋ 2.659 x age ＋ 0.599 x MIF ＋ 5.040 x IL-10; SAP group=-157.864 ＋ 2.721 x age ＋ 1.008 x MIF ＋ 6.468 x IL-10; UAP group= - 197.327 ＋ 2.360 x age ＋ 2.932 x MIF ＋ 7.640 x IL-10. Both retrospective and cross validation showed 6.4% of total miss-grouping. Both sets of discriminant functions had the same efficiency （ 100% ） for differential diagnosis of SAP and UAP. Conclusion The discriminant functions based on samples＇ age, MIF and IL-10, which were screened and suggested by stepwise method, may contribute to the differential diagnosis of atypical SAP and UAP, and therefore demonstrate better cost-efficiency.