Exploring the relation between changes in NT-proBNP and renal function in patients with suspected heart failure using structural equation modelling

Background: The relation between changes in NT-proBNP and renal function has commonly been studied using multiple regressions, which may ignore the complexity of relations between related variables. Methods and results: Data were collected from patients referred with suspected heart failure (HF) to a community service. Structural equation modelling (SEM) was used to assess the association between changes in NT-proBNP at 1 year, and other pre-specified variables including age, sex, BMI, eGFR, loop diuretics and ACE inhibitor. Of 1006 patients with a follow-up NT-proBNP at 1 year, 882 (88%) had HF. The baseline median age was 72 (IQR: 63–78) years, 732 (73%) were men, 668 (66%) had left ventricular systolic dysfunction and 769 (76%) had NT-proBNP > 400 pg/ml. For all patients at 1 year, 243 (24%) patients had at least a 50% reduction in NT-proBNP, and 199 (20%) had at least a 50% increase, only 40 (3%) had < 3% change. Change in NT-proBNP was strongly associated with baseline NT-proBNP (the standardized coefficient (r) = 0.73, p < 0.001). The change in NT-proBNP was not associated with changes in eGFR, and was indirectly related with age, BMI, eGFR and loop diuretics (p < 0.01 for all). Conclusions: Baseline NT-proBNP was the main determinant of change in NT-proBNP at one year.