Automated Clinical Chemistry Platform Performance Across Two Academic Medical Centers

Abstract

Abstract Our objective was to define the performance of automated chemistry platforms at two large academic medical centers by calculating and comparing sigma metrics for 28 analytes. Performance characteristics of chemistry assays on two Roche Cobas analyzers and four Abbott Architect analyzers were estimated using 12 months of Bio-Rad quality control (QC) data at two concentrations. Method imprecision was calculated as the cumulative QC coefficient of variation (CV) across analyzers at each QC concentration and percent bias was calculated by comparison of analyzer QC mean to peer group means. Sigma values were calculated for each method as [(TEa – Bias%)/CV%] using allowable total error (TEa) from two sources: the CLIA evaluation limits and desirable biological variation. Average sigma values were generated for each site and graded as optimal, >6 sigma; good, 5 to 6 sigma; marginal, 3 to 5 sigma; or poor, <3 sigma. Analysis of NIST SRM1950 standards for a subset of analytes allowed an estimation of absolute bias. Sigma metrics were highly comparable across both study sites. Considering CLIA TEa, just over half (Roche 57%; Abbott 54%) of the 28 analytes met the six-sigma standard of performance. The highest performing groups to meet or exceed the six-sigma standard were lipids (cholesterol, LDL, HDL, and triglycerides) and enzymes (ALP, ALT, AST, CK, GGT, LDH, and lipase). Electrolytes (Na, K, Cl, Mg) and metabolites (total bilirubin, BUN, CO2) failed to meet six-sigma. Notably, there were dramatic differences in sigma values calculated using CLIA and Ricos TEa criteria. Almost 40% of the analytes had at least one QC that performed poorly using Ricos criteria. Only 4 of the 28 assays (CK, GGT, lipase, and triglycerides) demonstrated optimal performance at both study sites using Ricos and CLIA criteria. Overall, 11 analytes at each institution exhibited unacceptable performance using Ricos criteria as opposed to only two analytes with the CLIA TEa limits. Analysis of NIST SRM1950 at both study sites gave comparable sigmas for all analytes except total bilirubin, cholesterol, Mg, and total protein. Neither Abbott nor Roche analyzers met six-sigma quality standards for all analytes tested. Overall assay performance across these two platforms at two major academic medical centers was almost identical. CLIA TEa and RICOS TEa criteria are significantly different, with wider limits of acceptability for CLIA. Variations between individual analyzers and manufacturers and limitations in automation would make tailored QC rules based on sigma metrics difficult to implement in a high-volume laboratory.