Lurye J.C., Behrend E.N. and Kemppainen R.J.
OBJECTIVE: To compare serum total thyroxine (T4) concentrations obtained with an in-house ELISA and a validated radioimmunoassay (RIA). DESIGN: Laboratory trial. SAMPLE POPULATION: 50 canine and 50 feline serum samples submitted for measurement of total T4 concentration with the RIA; samples were selected to represent a wide range of concentrations (< 6 to 167 nmol/L). PROCEDURE: Results of the ELISA and RIA were compared by calculating correlation coefficients, examining linearity, determining bias and precision, and evaluating clinical interpretations. RESULTS: Correlation coefficients for results of the 2 methods were 0.84 for the canine samples and 0.59 for the feline samples. Examination of bias plots revealed large variations in ELISA results, compared with RIA results. For the feline samples, the ELISA consistently overestimated total T4 concentration obtained with the RIA. When results of the 2 methods were categorized (low, borderline low, normal, borderline high, or high), results were discordant for 24 (48%) and 29 (58%) of the canine samples and for 18 (36%) and 28 (56%) of the feline samples (depending on whether borderline high ELISA results were considered normal or high). Reliance on results of the ELISA would have led to inappropriate clinical decisions for 31 (62%) canine samples and 25 (50%) feline samples. The ELISA coefficients of variation for the pooled canine and feline samples were 18 and 28%, respectively. CONCLUSIONS AND CLINICAL RELEVANCE: Substantial discrepancies between ELISA and RIA results for T4 concentrations were detected. Thus, we concluded that the in-house ELISA kit was not accurate for determining serum total T4 concentrations in dogs and cats.