Advantages of the lognormal approach to determining reference change values for N-terminal propeptide B-type natriuretic peptide

Serial measurement of NT-proBNP is performed routinely in the monitoring and assessment of the effectiveness of therapy in patients being treated for chronic heart failure (CHF). Intra-individual changes in NT-proBNP levels over time are compared typically to a reference change value (RCV) determined using either a standard [i.e., nested analysis of variance (nANOVA)] or a lognormal approach. The RCV defines the minimum percent change in serial analyte values that exceeds the percent change expected due to biological variation alone. Currently, there is no consensus on which approach (nANOVA or lognormal) to determining RCV is better. AIMS: Based on these considerations, we aimed to illustrate the impact of data transformation on the calculation of the biological variation of NT-proBNP and discuss the utility of logarithmic transformation in monitoring patients with heart failure. METHODS: 15 healthy subjects were enrolled after informed consent; 5 blood specimens were collected twice a week. Nested ANOVA from replicate analyses was applied to obtain components of biological variation, on the raw data and after data transformation. RESULTS: NT-proBNP distribution being highly skewed required data transformation. Natural log transformation yielded normalization. An example demonstrates that for untransformed values the RCV was overestimated for low concentrations of NT-proBNP and underestimated for higher concentrations. CONCLUSIONS: Log-transformed data are often used in the establishment of reference intervals for evaluating laboratory tests results in clinical practice, especially when the reference interval data are not Gaussian distributed. As log-normal approach is the best approach to determining RCV values we encourage its use assessing the clinical utility of NT-proBNP serial testing. We propose that the log-normal approach becomes the standard approach to determining RCV and replaces the use of nANOVA