Evaluation of an in-clinic dry chemistry analyzer for canine, equine, and feline plasma samples

Abstract

Method validation studies characterize the performance of new laboratory methods relative to established methods using quality guidelines in order to define the new method's performance characteristics and to identify differences that could influence data interpretation. We investigated the performance of an in-clinic dry chemistry analyzer (Catalyst One, IDEXX) for measuring 19 routine plasma biochemistry analytes in dogs, cats, and horses. We analyzed 2 levels of quality control material (QCM) in duplicate twice daily for 5 d to determine the coefficient of variation (CV), percent bias, observed total error (TE_obs), and sigma metric (σ) for each analyte at each level of QCM. We analyzed 82 canine, equine, and feline plasma samples with the in-clinic dry chemistry analyzer and a reference wet chemistry analyzer, and results were compared using correlation coefficients, Deming regression, and Bland-Altman analyses. CVs were <5% for 16 analytes and ⩾5% for 3 analytes. TE_obs was less than allowable total error (TE_a) for 9 analytes, and exceeded TE_a for 10 analytes. Sigma metrics were >4 at both levels of QCM for 5 analytes, and at one level of QCM for 5 analytes; sigma metrics were <3 or could not be calculated at the remaining analyte concentrations. All analytes, except glucose, showed various magnitudes of bias compared to the wet chemistry analyzer. Based on these results, we recommend statistical (5 analytes) and non-statistical (14 analytes) QC measures and analyzer-specific reference intervals.

Keywords: Catalyst One; clinical chemistry tests; point-of-care systems; quality control.

Figure 1.

Deming regression and Bland–Altman plots of albumin (Alb), alkaline phosphatase (ALP), alanine aminotransferase (ALT), blood urea nitrogen (BUN), and calcium (Ca) for the in-clinic dry chemistry analyzer relative to the reference wet chemistry analyzer. Data points were derived from canine (closed circle), equine (closed square), or feline (open circle) plasma samples. For Deming regression plots, lines of identity (y = x, dotted) and lines of best fit (solid) are plotted. For Bland–Altman plots, lines of no bias (y = 0, dotted), mean bias (solid), and 95% limits of agreement (dashed) are plotted.

Figure 2.

Deming regression and Bland–Altman plots of cholesterol (Chol), creatine kinase (CK), chloride (Cl), creatinine (Crea), and gamma-glutamyl transferase (GGT) for the in-clinic dry chemistry analyzer relative to the reference wet chemistry analyzer. Data points were derived from canine (closed circle), equine (closed square), or feline (open circle) plasma samples. For Deming regression plots, lines of identity (y = x, dotted) and lines of best fit (solid) are plotted. For Bland–Altman plots, lines of no bias (y = 0, dotted), mean bias (solid), and 95% limits of agreement (dashed) are plotted.

Figure 3.

Deming regression and Bland–Altman plots of glucose (Glu), potassium (K), magnesium (Mg), sodium (Na), ammonia (NH₃), and phosphorus (P) for the in-clinic dry chemistry analyzer relative to the reference wet chemistry analyzer. For Deming regression plots, lines of identity (y = x, dotted) and lines of best fit (solid) are plotted. For Bland–Altman plots, lines of no bias (y = 0, dotted), mean bias (solid), and 95% limits of agreement (dashed) are plotted.

Figure 4.

Deming regression and Bland–Altman plots of total bilirubin (TBili), total protein (TP), and triglyceride (Trig) for the in-clinic dry chemistry analyzer relative to the reference wet chemistry analyzer. For Deming regression plots, lines of identity (y = x, dotted) and lines of best fit (solid) are plotted. For Bland–Altman plots, lines of no bias (y = 0, dotted), mean bias (solid), and 95% limits of agreement (dashed) are plotted.