ARS | Publication request: Quantifying Assay Variation in Nutrient Analysis of Feedstuffs

Submitted to: Journal of Dairy Science
Publication Type: Abstract
Publication Acceptance Date: April 11, 2006
Publication Date: July 9, 2006
Citation: Mertens, D.R. 2006. Quantifying assay variation in nutrient analysis of feedstuffs [abstract]. Journal of Dairy Science. 89 Supplement 1):383.

Technical Abstract: Analytical results from different laboratories have greater variation than those from a single laboratory, and this variation differs by nutrient. Objectives of this presentation are to describe methods for quantifying the analytical reproducibility among and repeatability within laboratories, estimate the expected variation for nutrient assays, and discuss sources of assay variation. Only carefully designed and replicated collaborative studies or proficiency testing programs can measure variation in nutrient assays. The National Forage Testing Association (NFTA) has developed a proficiency testing program that partitions variation in nutrient analysis into two components: accuracy and precision. This program and its statistical methodology will be described. Although they are often used interchangeably, accuracy and precision measure two independent sources of variation. Accuracy is related to the closeness of the result to the known or consensus reference value. The NFTA determines consensus values as the censored averages of only laboratories using the reference method. Precision is related to the consistency of results among repeated assays. Precision is related mostly to random variation and accuracy is related primarily to systematic error or bias. Much of the random variation associated with precision is related to differences in test samples from heterogeneous materials. Most systematic bias is related to differences in methodology among laboratories and technicians, although some systematic true error is associated with mistakes in calculation or corrections using inaccurate DM determination. Although consistency in analyses (precision) is desired by feed producers, accuracy is required by feed users because the animal provides an independent and ultimate evaluation of nutrient content and utilization. Variation in nutrient analysis is real and controllable, but not completely avoidable. For accurate determination of mean values for tables of feed composition, assays should be replicated across both test samples and analytical laboratories to verify consensus among results.