In the dairy sector, there has been a long run debate about an appropriate basis of calibration for milk protein content of routine indirect instrumental methods (IM) calibrated according to reference methods. Using traditional methods, milk protein content in per cent is given in terms of the content of so called crude protein, i.e. total nitrogen (according to Kjeldahl method) x 6.38. Indirect methods are usually calibrated for these values. In France, the content of milk true protein, that means the content of protein nitrogen (according to Kjeldahl method) x 6.38, has been used as a calibration basis for a longer time. Optimizing discussion about the aspects of suitability, accuracy and correctness as well as about the advantages and difficulties of these different approaches can be encountered in literature. The objective of the present paper is to discuss theoretical and practical problems: analyses of milk nitrogen matters; relationships of some methodical procedures used for analyses of milk nitrogen matters; interrelationships between milk nitrogen matters. Within a year, calibration of a milk infraanalyzer Milko-Scan 133 B (MSc, Foss Electric, Denmark) was done eleven times for the content of crude protein (CP) that was determined by Kjeldahl method (KM) on a Kjeltec instrument (Tecator AB, Sweden). The patterns and results of these calibrations (Tab. I) were very good with respect to generally respected requirements. At the same time, 288 individual samples of cow's milk from the middle part of lactation were analyzed in the middle of intercalibration intervals. The cows came from three herds of two breeds: Czech Pied and Black-Pied Lowland breed. The cows were at the second and higher lactations. Kjeldahl method, calculations and photometric method were used for the particular analyses in order to quantify the separate nitrogen fractions of milk according to a scheme shown in Tab. II and/or III. Absolute and relative proportions of nitrogen (N) fractions in milk are summarized in Tab. II. The absolute values of the levels of most N fractions were lower in comparison with those established in previous studies, but they were comparable, in absolute and relative terms, with the levels determined for the Holstein breed (Cerbulis and Farrell, 1975). Protein nitrogen equalled on average 96.3% of total N. The share of casein N in protein N was 82.6%. Urea nitrogen made 72% of nonprotein nitrogen (NN). The per cent of nonprotein N is important for calibrations of indirect methods (so called instrumental methods IM). In this case, it amounted to 3.7% of total N. The difference between the determination of CP by reference (RM), i.e. Kjeldahl method (KM) and MSc was insignificant (p > 0.05, the difference 0.006+/-0.081%). Figs. 1, 2 and 3 show the relationships between CP and TP contents determined by various procedures. Theoretically it is more appropriate to calibrate the infraanalyzers for the content of true protein (TP) in milk according to KM. But applying a practical procedurs, there was no closer relationship between infraanalyzer readings and TP content according to KM in comparison with CP content according to KM: r = 0.947 > 0.937 (Figs. 1 and 2). Manual analytical inaccuracies at KM method when determining TP content in curd (direct) may be a reason. Determination of TP content by Kjeldahl method is more labor- and time-consuming than the determination of CP content. These inaccuracies may overlap the theoretico-practical and expected effect of natural variability of NN on the agreement of infraanalysis and reference method (RM = KM). Although the better agreement of IM x RM was not achieved empirically, further evaluations show (Figs. 4, 5 and 7) that an improvement should be achieved theoretically. It is possible that determination of NN content (by KM) in filtrate that decrease the risk of inaccuracies, could help to improve empirically the IM x RM relationship. TP content for the purposes of IM calibration could be calculated as the difference: total N (by KM) minus NN (by KM), i.e. indirect. The relationships between the content of NN and/or urea (U) and the difference between the content of crude protein determined by infraanalysis and Kjeldahl method (CP, MSc - CP, KM) were as follows: r = -0.38 and r = -0.33 (p < 0.001), resp., Figs 4 and 5, resp. A relationship between U and difference CP, MSc - TP, KM was insignificant (Fig. 7). These three evaluations (Figs. 4, 5 and 7) confirm the theoretical assumptions for improvement of the IM x RM relationship at IM calibration for TP content by KM. On the other hand, the theoretical assumptions were not confirmed by evaluation in Fig. 6, which can also result from the methodical procedure of protein N determination by KM in filter curd. Eg. Grappin et al. (1980), Grappin (1992) and Grappin and Lefier (1993) did not confirm in several cases, but mostly they did the better agreement between IM and RM at IM calibration for TP content by KM. Therefore this procedure has been used for about 20 years in the only country - in France. Barbano and Lynch (1990, 1992) also recommend this procedure in theoretical discussions. Discussion of dairy specialists about this problem is going on, each system (calibration for CP or TP) has its pros and cons. But the reason for TP to be accepted as a calibration basis for IM seem to more serious. The relationships between the particular nitrogen matters are shown in Tab. III. The highest correlation was found between crude protein and true protein: r=0.96. All important correlation coefficients are positive, except one. coefficient. The relationship between U and NN content was relatively close: r=0.45. In order to decide what calibration basis for IM should be accepted in future, it is important to know the relationships between nitrogen matters of milk and methods of their determination. So this paper could contribute to the knowledge of this problem.
