NT-proBNP and left ventricular ejection fraction correlation: Methodological considerations and clinical heterogeneity

To the Editor,

We read with interest the recent article by Lundgren and colleagues examining the correlation between N-terminal pro-B-type natriuretic peptide and left ventricular ejection fraction in the Swedish DEMONSTRATE database. ¹ The authors report a negative correlation of r= -0.40 between NT-proBNP and LVEF in nearly 15,000 patients, which decreased to r= -0.20 after adjusting for age and kidney function estimates. While the large sample size represents a significant strength, several methodological considerations warrant discussion regarding the interpretation of these findings and their applicability to clinical practice.

The study design introduces potential selection bias that may limit generalizability. All included patients underwent echocardiography, which typically indicates clinical suspicion of cardiovascular disease beyond isolated heart failure. As the authors acknowledge, echocardiography is indicated for multiple conditions including valvular heart disease and acute coronary syndrome. ² This selection criterion likely enriched the cohort with patients having lower LVEF or higher NT-proBNP values compared with the general population, potentially inflating the observed correlation coefficients. Furthermore, the DEMONSTRATE database inclusion criteria requiring at least one potassium measurement and diagnosis of heart failure, hyperkalaemia, chronic kidney disease, diabetes, or hypertension introduces additional selection bias toward patients with complex comorbidity profiles.

The handling of LVEF data presents interpretive challenges. LVEF was frequently reported as intervals or qualitative ranges rather than precise numerical values, requiring conversion assumptions that introduce measurement uncertainty. The authors’ decision to interpret “greater than 55%” uniformly as 55%, despite this being the most common result, systematically truncates higher normal values and may artificially compress the observed range of LVEF measurements. This approach could attenuate correlation estimates and affects the clinical interpretation of preserved ejection fraction cases.

The substantial inter-individual variability in NT-proBNP levels within each LVEF category, clearly demonstrated in the authors’ box plots, underscores a critical limitation for individual patient management. While group-level correlations inform population trends, the broad distributions suggest limited predictive value for individual clinical decision-making. The authors appropriately note that sequential NT-proBNP measurements within individual patients may prove more clinically useful than cross-sectional comparisons. ³ This distinction is particularly relevant given known within-week intra-individual variability of 21% in heart failure patients. ⁴

The methodological differences between the two NT-proBNP assay platforms used in the study merit consideration. Although the authors report no statistically significant difference between Abbott Architect and Roche Cobas methods, previous external quality assessment data from Swedish laboratories demonstrate systematic bias, with Abbott methods yielding results 6%-8% higher than Roche methods. ⁵ The lack of assay standardization across manufacturers represents a persistent challenge in natriuretic peptide measurement that complicates establishment of universal clinical thresholds.

The authors present valuable data on the ability of NT-proBNP to discriminate between LVEF categories and predict heart failure diagnosis. However, the sensitivity of 78% for NT-proBNP above 635ng/L in diagnosing heart failure falls below the approximately 90% sensitivity reported in established diagnostic algorithms. ⁶ This discrepancy may reflect the retrospective design and inability to distinguish acute from chronic heart failure presentations, as the authors acknowledge. The heterogeneity of clinical presentations within the database limits direct comparison with prospective diagnostic studies.

Despite these methodological considerations, the study makes an important contribution by demonstrating persistent correlation between NT-proBNP and LVEF even after adjusting for major confounders in a real-world clinical population. The findings support the complementary roles of these measurements in heart failure assessment while highlighting the complexity of their relationship. Future prospective studies specifically designed to address the limitations identified here, particularly regarding patient selection, assay standardization, and distinction between acute and chronic presentations, would strengthen clinical interpretation and application of these important biomarkers.