The increasing use in Italy of the Cornell Net Carbohydrate and Protein System (CNCPS) led researchers of five Italianuniversities to carry out a collaborative study to evaluate the precision of the CNCPS chemical analysis and derivate fractions.Each laboratory conducted in duplicate the chemical analyses according to the Weende (dry matter; crude protein;ether extract; crude fibre; ash), Van Soest (neutral and acid detergent fibre, NDF and ADF; acid detergent lignin; ADL)and CNCPS (soluble proteins, SP; non-protein nitrogen, NPN; neutral and acid detergent insoluble protein, NDIP andADIP; starch, ST) schemes on the same five feeds (barley meal, wheat straw, maize silage, dried lucerne and fieldbeans). Anomalous analytical data were identified and corrected by the "box-plot" graphic tool before the calculation ofthe CNCPS protein (B 1, B 2, and B 3,) and carbohydrate (A, B 2 and C) fractions. Finally, repeatability (chemical analysis) andreproducibility (chemical analysis and fractions) were calculated and expressed as relative values (repeatability andreproducibility standard deviation as percentage of the corresponding mean, RSDr and RSDR, respectively). Chemicalanalyses of the Weende scheme, together with NDF, ADF and ST analyses, have satisfactory repeatability (0.3-6.2%) andreproducibility (0.3-11.2%) values. On the contrary the ADL, NPN, NDIP and ADIP analyses showed high variability, bothwithin and between laboratories (RSD r and RSD R between 20 and 45%). The SP analysis had an intermediate value ofprecision (RSD r=10.6%; RSDR=16.4%).Finally, since different combinations of several chemical analyses with scarce (ADL, NPN, NDIP, ADIP, SP) or average precision(e.g. NDF and starch) are used to calculate CNCPS fractions (excluding B 2 protein fraction), also the reproducibilitiesof these fractions are poor and range from 10 to 20%.
