Skip to main content
ARS Home » Midwest Area » Madison, Wisconsin » U.S. Dairy Forage Research Center » Cell Wall Biology and Utilization Research » Research » Publications at this Location » Publication #269986

Title: A ring test of in vitro neutral detergent fiber digestibility: Analytical variability and sample ranking

Author
item Hall, Mary Beth
item MERTENS, DAVID - Former ARS Employee

Submitted to: Journal of Dairy Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/26/2011
Publication Date: 4/1/2012
Citation: Hall, M., Mertens, D.R. 2012. A ring test of in vitro neutral detergent fiber digestibility: Analytical variability and sample ranking. Journal of Dairy Science. 95(4):1992-2003.

Interpretive Summary: Laboratory measures of fiber digestibility by ruminal microbes are commonly used to evaluate forage quality for use in diet formulation. All feed quality assays have some variability, where a correct value for a feed will actually cover a range of values. The more precise the assay, the narrower the range of values one is likely to get. The precision of the fiber digestibility assays was not known, so a study was conducted to evaluate the precision. Three times over the course of a year, 9 laboratories were sent 14 forage samples which they analyzed for fiber digestibility after 30 hours of fermentation by rumen microbes. The precision varied somewhat among forage samples but the value for digestibility of a given forage was the average plus or minus 5 percentage units (for example, if the average was 50% fiber digestibility, the likely values would fall between 45% and 55% digestibility). However, the labs were able to rank the samples similarly from highest to lowest digestibility. So, although it may not be advisable to use the values as very precise numbers, the fiber digestibility assay is still quite useful for making relative comparisons of the fiber digestibility of different forages. This information will be useful for those who formulate dairy cow diets and for forage analysis laboratories.

Technical Abstract: In vitro neutral detergent fiber (NDF) digestibility (NDFD) is an empirical measurement used to describe fermentability of NDF by rumen microbes. Variability is inherent in assays and affects the precision that can be expected for replicated samples. The study objective was to evaluate variability within and among laboratories (labs) of 30 h NDFD values measured in repeated runs. Subsamples of alfalfa (n=4), corn silage (n=5), and grass (n=5) ground to pass a 6 mm screen were sent to 10 labs on 3 occasions over a 12 month period. Subsamples passed a test for homogeneity. Labs ground the samples and ran 2 or 3 replicates of each sample within run, and analyzed 2 or 3 sets of samples. A lab that did not provide in-run replicate data was not included in evaluation of standard deviations (SD). Mean and SD for sample within run within lab were calculated. Factors in the statistical model were lab, run within lab, sample, and lab by sample. All factors affected NDFD (P<0.01 for all) and within-run SD of NDFD (P<0.03 for all). The lab by sample effect suggests against a simple lab bias. Labs used 2 NDFD procedures: 8 labs used a procedure similar to Goering and Van Soest, 1970 (GVS) using fermentation vessels or filter bags, and 2 used a procedure with pre-incubated inoculum (PInc). Among GVS labs, NDFD results were affected by all factors (P<0.01 for all; mean 48.5%, range 42.7 to 53.2%, SED = 0.98). For PInc, mean NDFD was 30.7% (range: 31.8 to 29.6%); GVS and PInc NDFD differed (P<0.01, SED=0.95). Mean within-run SD were 1.9% (range: 0.5 to 3.4%) for GVS and 2.6% (range 1.8 to 3.4%) for PInc. The mean SD for all labs of 2.0% indicates that 95% of results for a sample within a run would be within a range of 11% NDFD. Labs ranked samples similarly within forage type. Spearman correlation coefficients between average rankings and those reported by labs were 0.83 for alfalfa, 0.70 for corn silage, and 0.90 for grass (P<0.01 for all). It is concluded that across all labs an average precision of 11% NDFD can be expected for a single analysis within run. Differences between GVS and PInc suggest using results in contexts appropriate to each procedure.