A combined clinical and biomarker approach to predict diuretic response in acute heart failure

BACKGROUND: Poor diuretic response in acute heart failure is related to poor clinical outcome. The underlying mechanisms and pathophysiology behind diuretic resistance are incompletely understood. We evaluated a combined approach using clinical characteristics and biomarkers to predict diuretic response in acute heart failure (AHF). METHODS AND RESULTS: We investigated explanatory and predictive models for diuretic response--weight loss at day 4 per 40 mg of furosemide--in 974 patients with AHF included in the PROTECT trial. Biomarkers, addressing multiple pathophysiological pathways, were determined at baseline and after 24 h. An explanatory baseline biomarker model of a poor diuretic response included low potassium, chloride, hemoglobin, myeloperoxidase, and high blood urea nitrogen, albumin, triglycerides, ST2 and neutrophil gelatinase-associated lipocalin (r(2) = 0.086). Diuretic response after 24 h (early diuretic response) was a strong predictor of diuretic response (β = 0.467, P < 0.001; r(2) = 0.523). Addition of diuretic response after 24 h to biomarkers and clinical characteristics significantly improved the predictive model (r(2) = 0.586, P < 0.001). CONCLUSIONS: Biomarkers indicate that diuretic unresponsiveness is associated with an atherosclerotic profile with abnormal renal function and electrolytes. However, predicting diuretic response is difficult and biomarkers have limited additive value. Patients at risk of poor diuretic response can be identified by measuring early diuretic response after 24 h.