OBJECTIVES: Heart failure (HF) is a clinical syndrome with a global burden. In Austria, 291,400 people affected a HF, of which 49,007 are newly detected. The symptoms of HF are non-specific and sometimes difficult to distinguish from those of other diseases. NT-proBNP serves as useful biomarker for the diagnosis of HF not only in patients with acute symptoms but also in outpatients with an ambiguous clinical presentation. The aim of the analysis is to evaluate the cost-effectiveness of screening patients with or without type 2 diabetes-mellitus (T2DM) with NT-proBNP to detect HF compared to no-screening, i.e. diagnosis based on clinical signs from the perspective of the Austrian health-care system.

METHODS: A time-discrete Markov model was developed to simulate the effect/improvement (lifetime costs, quality-adjusted life-years/QALYs and life-years/LYs) due to a NT-proBNP screening in the undetected subgroup of the total HF patients. All patients with HF (un-/detected) run into the model according to a distribution of NYHA-classes. The model considers disease-progression by transition of NYHA-classes. Undetected patients may remain undetected or be detected with the help of NT-proBNP or symptoms. Patients with known HF exhibit a slower disease-progression than undiagnosed patients. The probability of dying is influenced by the respective NYHA-class. Direct costs (2020€) were derived from published sources. QALYs, LYs and costs were discounted (3%p.a).

RESULTS: In the per-patient analysis (at age 60 over lifetime), the ICER/QALY of NT-proBNP vs. no-screening was 5,160€ for HF patients; 5,236€ for the subgroup with T2DM and 5,004€ without T2DM. Considering the total cohort (n=58,388) with the corresponding age structure over lifetime, ICERs improve for the total HF patients (4,265€) and considerably for the subgroup with T2DM (n=20.267) to 1,878€.

CONCLUSIONS: Overall, the analysis concludes that screening with NT-proBNP is a highly cost-effective to cost-saving diagnostic option for patients with HF. Sensitivity analysis confirmed the robustness of these findings.