Looking beyond linear regression and Bland-Altman plots: a comparison of the clinical performance of 25-hydroxyvitamin D tests.

Abstract

BACKGROUND: The systematic evaluation of the clinical concordance of various 25-hydroxyvitamin D (25OHD) testing methods is presented. The need for this approach is raised by the discrepancies in the analytical performance of the available assays. METHODS: The analytical and clinical performance of six automated 25OHD assays and an in-house liquid chromatography-tandem mass spectrometry (LC-MS/MS) method was investigated. Leftover serum samples (n=162, SA: n=114) were analyzed and all 21 assay combinations were evaluated. The utility of Cohen's kappa values was assessed by transforming them into minimum percentage agreement (MPA). McNemar's hypothesis test was employed for testing the symmetry of the disagreeing classification outcomes within each method pair. RESULTS: Depending on the assay method, the ratio of results classified as positive (<20 ng/mL) was 13.5%-40.0%. The percentage agreement (PA) was 74.1%-92.6%. Compared to other methods, significantly more hypovitaminosis cases were delivered by DiaSorin Liaison(R) 25 OH vitamin D Total (DL) and significantly fewer by IDS-iSYS 25-Hydroxy Vitamin DS (II). The strongest clinical concordance was exerted by II vs. LC-MS/MS. The kappa-derived MPA showed close similarity to the PA scores. McNemar's tests confirmed the asymmetry of the disagreement in the classification in 14 method combinations. CONCLUSIONS: The presented approach allows the prediction of the clinical consequences of a 25OHD method transfer. Differences in the clinical classification of assay results are likely encountered when transferring to a new method, even between assays standardized according to the Vitamin D Standardization Program (VDSP) Reference Method Procedure (RMP).