A Diagnostic Study To Set Up NT-proBNP Diagnostic Model For Chronic Systolic Heart Failure Diagnose

BACKGROUND:Heart failure is one of the most important cardiovascular disease in this age and BNP/NT-proBNP is a popular and convenient diagositc method for HF with high sensitivity. While the diagnostic value of BNP/NT-proBNP still not be fully explored as current diagnostic studies for it are only focus on simple influence factor, but the serum BNP/NT-proBNP level is influenced by multiple factors. A well designed diagnostic study for BNP/NT-proBNP and multiple influence factors can contribute a lot for the HF diagnosis in clinical practice.OBJECTIVE:This study is a diagnostic trial which focus on NT-proBNP and the chronic systolic heart failure. It test serum NT-proBNP level and other clinical factors of suspected chronic systolic HF subjects to set up a chronic systolic HF diagnostic model, and do external validation for the diagnostic value of the model.METHOD:Suspected chronic systolic heart failure patients were chosen as the subjects of this study, and Left ventricular ejection fraction (EF)<50% is used as the Golden Standard of chronic systolic HF diagnose. All subjects were divided into 2 study group (Disease Group & Non-Disease Group) according to their EF level (the Golden Standard). Test serum NT-proBNP level and other clinical factors such as eGFR, BMI, HbA1c etc of study subjects. All study subjects data randomizely divided into 2 data sets (MODEL data set & VALIDAYION data set). Use logistic regression to set up chronic systolic HF diagnostic model which include NT-proBNP and other clinical factors in MODEL data set. Use the set diagnostic model to diagnose chronic systolic HF in VALIDAYION data set to do external validation of the model. Report Sensitivity (Se) in Disease Group, and report Specificity (Sp) in Non-Disease Group according to the Golden Standard. Report Positive Predictive Value (PPV), and Negative Predictive Value (NPV) in different prevalence of chronic systolic HF. Create ROC curve of the model in VALIDAYION data set, and report AUC for the diagnostic value of the model. The study also compared the AUC of diagnostic model and non-stratified NT-proBNP in the VALIDAYION data set.RESULT:The diagnostic model which set up in MODEL data set is:Logit(P)=-9.768+ 0.014NT-proBNP+0.116BMI+0.054eGFR-0.032age, cut-off point= 0.46525 (Se= 0.932; Sp= 0.929), subject with the level above 0.46525 can be diagnosed as chronic systolic HF. When use diagnostic model to do chronic systolic HF in VALIDATION data set, Sensitivity of the model in Disease Group is 0.954 (95%CI:0.9854,0.9236), and the Specificity of the model in Non-Disease Group is 0.917 (95%CI:0.9718,0.8615). AUC of the model in VALIDATION data set is 0.988 (95%CI= 0.972,0.999), when compare to the AUC of non-stratified NT-proBNP [0.947 (95%CI=0.896,0.996)] in same data set, P=0.0473.CONCLUSION:The study result indicated that the NT-proBNP diagnostic model which set by logistic regression and with NT-proBNP and multiple clinical factors can effectively diagnose chronic systolic HF among the suspected chronic systolic HF subjects.