Premium
Objective evaluation of analyzer performance based on a retrospective meta‐analysis of instrument validation studies: point‐of‐care hematology analyzers
Author(s) -
Cook Andrea M.,
Moritz Andreas,
Freeman Kathleen P.,
Bauer Natali
Publication year - 2017
Publication title -
veterinary clinical pathology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.537
H-Index - 51
eISSN - 1939-165X
pISSN - 0275-6382
DOI - 10.1111/vcp.12492
Subject(s) - hematology analyzer , spectrum analyzer , medicine , statistics , quality (philosophy) , mathematics , computer science , telecommunications , philosophy , epistemology
Background Information on quality requirements and objective evaluation of performance of veterinary point‐of‐care analyzers (POCAs) is scarce. Objectives The study was aimed at assessing observed total errors (TE obs s) for veterinary hematology POCAs via meta‐analysis and comparing TE obs to allowable total error (TE a ) specifications based on experts′ opinions. Methods The TE obs for POCAs (impedance and laser‐based) was calculated based on data from instrument validation studies published between 2006 and 2013 as follows: TE obs = 2 × CV [%] + bias [%]. The CV was taken from published studies; the bias was estimated from the regression equation at 2 different concentration levels of measurands. To fulfill quality requirements, TE obs should be < TE a . Measurands were considered as globally acceptable if > 60% of analyzers showed TE obs < TE a . Results Six studies evaluating 11 analyzers and 5 studies evaluating 5 analyzers were included for canine and feline hematology variables, respectively. For the CBC, TE obs was < 15% for canine and < 13% for feline measurands, except for HGB and platelet counts. Measurands of the CBC, excluding differential WBC and platelet counts, and HGB concentration were considered globally acceptable. For most of the cell types in the WBC differential count, TE obs was > TE a (data from 3 analyzers). Conclusion This meta‐analysis is considered a pilot study. Experts′ requirements (TE obs < TE a ) were fulfilled for most measurands except HGB (due to instrument‐related bias for the ADVIA 2120) and platelet counts. Available data on the WBC differential count suggest an analytic bias, so nonstatistical quality control is recommended.