DefinePK

Abstract

The aim of the study was to evaluate the diagnostic accuracy of Gem Premier 3000 (GEM) and Edan i15 Vet (EDAN) for hypoglycemia and hyperglycemia by the analysis of blood glucose concentration (BGC) compared to the reference method (RM) with conventional statistics and using a new method (reversed multiple cut-off points: RMC). BGC was analysed with whole blood at GEM, EDAN and serum at RM in 123 Holsteins in different stages of lactation (DSL). Increasing cut-off points (ICP) (BGC ?40, ?45, ?50, ?55, ?60 and ?65mg/dL) and decreasing cut-off points (DCP) (BGC <40, <45, <50, <55, <60 and <65mg/dL) were defined for the RMC method. Excellent to low limits of agreement with RM (±1 to ±15% deviations) were determined for BGC analysis, differences of absolute and true positive prevalence (APP, TPP) and area under the curve (AUC). Passing-Bablok regression and Bland-Altman plots showed inconsistencies and large deviations for GEM/EDAN in DSL. Sensitivities and specificities were low and different for GEM/EDAN in hypoglycemia/hyperglycemia in receiver operating characteristic. RMC showed that APP and AUC differences were outside the limits for DCP and ICP for GEM and EDAN, also for DSL. The differences in BGC and TPP were outside the limits in DCP for GEM and in DCP and ICP for EDAN. To summarise, BGC in DSL influenced the results of conventional statistics. GEM/EDAN were not accurate for BGC analysis in Holsteins according to conventional statistics, which was also confirmed by the new RMC method in high and low BGC and in DSL. A higher probability of false-identification of hypoglycemia by EDAN and hyperglycemia by GEM according to RMC was observed.