A prediction model for estimating NT‐proBNP in a general Japanese population: the Toon Health Study

Abstract

Abstract Aims As part of the Toon Health Study, which is an ongoing population‐based cohort study, we aimed to develop a prediction model for N‐terminal pro‐brain natriuretic peptide (NT‐proBNP) in a general Japanese population. We sought to explore the influence of various demographic and clinical factors on NT‐proBNP levels and assess the model's performance. In addition, our objectives included internal validation and investigation of the diagnostic potential of the observed‐to‐predicted NT‐proBNP ratio (OPR) at baseline for predicting the risk of heart failure with preserved ejection fraction (HFpEF). Methods and results In this prospective cohort study, participants were recruited from Toon City, Japan, as part of the larger Toon Health Study, focusing on cardiovascular risk factors. We measured the NT‐proBNP levels and used linear regression with penalization (ridge regression) to develop the model. The model incorporated 10 prespecified predictors (age, gender, body mass index, diastolic blood pressure, heart rate, haemoglobin, albumin, total cholesterol, haemoglobin A1c, and estimated glomerular filtration rate) and underwent assessment using R2 and root mean squared error (RMSE). Internal validation was conducted through bootstrapping. In a post hoc analysis, we explored the OPR's diagnostic potential using 5 year follow‐up data (n = 636) to predict the elevation of NT‐proBNP > 125 pg/mL at the 5 year follow‐up as the risk of HFpEF. A total of 2505 participants (age: 60.4 ± 12.9 years, men: 35%) were enrolled in this study. There was a linear relationship between the observed and predicted values of NT‐proBNP in which the logarithm of observed NT‐proBNP was 125 pg/mL at the 5 year follow‐up with an area under the curve of 0.793. Conclusions This study introduces the first prediction model for NT‐proBNP in a general Japanese population. Although the model has acceptable performance, ongoing refinement is essential. Our transparent approach to model development, alongside a web‐based interactive tool, lays the groundwork for further improvements and external validation. The OPR holds potential for predicting the future risk of HFpEF. This research contributes to understanding the nuanced influence of patient backgrounds on levels of NT‐proBNP in asymptomatic individuals within the context of a broader population‐based cohort study.

Keywords

• Heart failure
• Heart failure with preserved ejection fraction
• NT‐proBNP
• Prediction model