TDF averaged 48.2% DM, with one laboratory having higher values (49.3% vs 47.9%; P<0.001). The TDF values varied among samples (P<0.001), from about 40% DM for diets to 60% DM for grape marc and sugarbeet pulp. Repeatability (SR=0.95% DM) and reproducibility of TDF (SL=1.68% DM) were good with a low coefficient of variation among laboratory: CVL= 3.9%. The aNDFcorr (mean 37.1% DM) significantly differed from laboratories that used Ankom system (36.4% DM) to laboratories that used Fibertech equipment (38.3% DM). The differences among laboratories due to the equipment and to the corrections for ash and protein explained the poorer repeatability and reproducibility of aNDFcorr determination, with CVL=6.6%. SF values differed (P<0.001) among laboratories (from 9.6% to 12.0% DM) and samples (from 4.0% DM of wheat bran to 8-11% DM of diets and alfalfa meal to 24.3% DM of sugarbeet pulp). The among-laboratory variability of SF was higher (SL=2.97% DM; CVL=26.8%), due to the variability of both TDF and aNDFcorr analyses.
In conclusion, TDF analysis was characterized by good repeatability and reproducibility, but it was less reproducible in case of raw materials with high SF levels. The among laboratory variation increased with aNDFcorr, because of the differences in analytical equipment and the procedure for protein and ash corrections. Finally SF reproducibility appeared rather good for complete diets and raw materials with low or medium concentration (SF 4-10% DM), but it was affected by the analytical errors of both TDF and aNDFcorr and needs a better harmonization.