Validation and method comparison for a point-of-care lateral flow assay measuring equine whole blood insulin concentrations

• Published in: Journal of veterinary diagnostic investigation 2023-03, Vol.35 (2), p.124-131
• Main Authors: Berryhill, Emily H., Urbina, Naomi S., Marton, Sam, Vernau, William, Alonso, Flavio H.
• Format: Article
• Language: English
• Subjects: Animals; Biological Assay - veterinary; Full Scientific Reports; Horses; Insulin; Point-of-Care Systems; Radioimmunoassay - methods; Radioimmunoassay - veterinary; Reproducibility of Results; Sensitivity and Specificity
• ISSN: 1040-6387; 1943-4936
• DOI: 10.1177/10406387221142288

The Wellness Ready Test (WRT) is a lateral flow, stall-side assay that measures equine insulin in whole blood and requires validation before recommending clinical use. We evaluated intra- and inter-assay precision and linearity and compared the WRT with a radioimmunoassay (RIA). Tested concentrations ranged from 695 pmol/L (100 μIU/mL). For 20 replicates at each insulin level, intra-assay CVs of the WRT for insulin were 13.3%, 12.9%, and 15.3% at low (139–278 pmol/L; 20–40 μIU/mL), intermediate (278–417 pmol/L; 40–60 μIU/mL), and high (>417  pmol / L ; >60 μIU/mL) concentrations, respectively. For 10 replicates at each level (3 assay lots), inter-assay CVs were 15.9%, 11.0%, and 11.7%, respectively. In the weighted linear regression of 5 measured insulin concentrations against expected concentrations, R2 = 0.98, slope = 1.02, and y-intercept = 14.4 pmol/L (2.08 μIU/mL). The Spearman correlation coefficient (rs) was 0.90 (95% CI: 0.85–0.94) between the WRT and RIA; the WRT = f(RIA) Passing–Bablok regression yielded the fit, y = 1.005x + 24.3 pmol/L (3.50 μIU/mL). The WRT result averaged 10.4% higher than the RIA result, with targeted bias of 25.9, 26.1, and 26.7 pmol/L (3.74, 3.76, and 3.84 μIU/mL) for cutoffs used to diagnose insulin dysregulation of 312, 347, and 451 pmol/L (45, 50, and 65 μIU/mL). Assay clinical sensitivities, specificities, and accuracies determined at the 3 selected clinical cutoffs and using the RIA as gold standard were 87–95%, 92–96%, and 91–95%, respectively (n = 99 samples). Observed total error was 28.4–30.4%. The WRT had acceptable precision, excellent linearity, and good association with the RIA.