PSIX-4 Interlaboratory comparison of vitamin D analysis in canned pet food.

Abstract Accurate vitamin D measurement in pet food is crucial to pet health due to the important role in animal health and subsequent risks of deficiency and toxicity. We therefore aimed to investigate the differences in performance between three commercial laboratories (Lab 1: Europhins, Des Moines, IA; Lab 2: Midwest Laboratories, Omaha, NE; Lab 3: Merieux NutriSciences, Crete, IL). Eight canned diets with varying vitamin D concentrations [from natural sources (67% chicken, and/or salmon) and synthetic vitamin D (+ 0, 100, or 200 IU)] were formulated and mixed in small batches. Ingredient pre-retort samples were tested (n = 1 sample laboratory). Paired t-test showed Lab 1 had significantly lower (P< 0.05) vitamin D concentrations than Lab 2 and Lab 3, which did not differ significantly (P >0.05) from each other. Utilizing laboratory-specific ingredient data we predicted vitamin D concentration for each diet (Table 1). Model fit (predicted vs measured) was high (R2 > 0.9) with Lab 1 (R2=0.98) and Lab 2 (R2=0.97) showing a slightly more accurate fit than Lab 3 (R2=0.91). Analysis of prediction errors using Paired t-tests on prediction errors (difference between measured and predicted) revealed no significant difference (P > 0.05) in accuracy between laboratories. Bland-Altman plots (i.e., difference vs. mean of predicted and measured) for each laboratory showed similar pattern with lower vitamin D diets (i.e., Chicken, and Chicken and Salmon diets) being closer to 0 with more values over-predicted, and higher vitamin D diets (i.e., Salmon diets) tending to be further from 0 and under-predicted. Predictive performance metrics (Table 2) were all lowest for Lab 1 indicating better predictive accuracy and lower variation. With the highest values seen for Lab 3. While all labs showed a high correlation (R2 > 0.9) between predicted and measured vitamin D, Lab 1 consistently showed lower values. Lab 1 also had the lowest variation and best predictive accuracy when examining performance metrics. Additional investigations are warranted.