Development of models to estimate milk urea nitrogen concentrations

This study aimed to estimate milk urea nitrogen (MUN) as a function of days in milk (DIM), lactation number (LN) and calving season. In the study, 7.006 observations were used. Milk production (kg), DIM and LN were collected on the milk sampling days. Calving seasons (CS) were divided into summer and winter. A model to describe the lactation curve was used to estimate MUN, adding LN and CS to it and verifying the coincidence among curves. Concentrations of MUN showed similar pattern to the lactation curve and differed among 1st, 2nd, 3rd and 4th-5th lactation, being infl uenced by CS.


INTRODUCTION
Urea is the main form of nitrogen excretion in mammals.For a long time, blood urea nitrogen has been known to refl ect the ineffi cient use of dietary crude protein by ruminants (Lewis, 1957) and has shown a good correlation with plasma and milk (MUN) urea nitrogen.Thus, MUN has been used as an indicator to monitor protein nutrition (Jonker et al., 1999), representing a simple, fast and cheap indicator to evaluate the nutritional status of cows during lactation (Roseler et al., 1993).Other factors have been shown to infl uence MUN: milk production, cow age, lactation stage, body weight, grazing system, milk composition and season of the year.In face of the confl icting results, the objective of this study was to develop models to estimate MUN concentrations.

MATERIAL AND METHODS
In this study was used 7,006 observations from 855 Holstein cows belonging to a commercial herd, Sao Paulo State (Brazil).Data were collected from September 2000 to January 2002 and included cows with 5 lactations or less, until 305 days in milk (DIM) and MUN concentrations less than 25 mg/dL.Milk production (kg/ day), DIM and lactation number (LN) were recorded on the milk sampling days.Calving seasons were divided into summer (from November to April) and winter (from May to October).
Animals were confi ned, fed 7 times/day and milked in milking parlour where each animal was identifi ed and had its production registered.They received TMR (48% roughage on dry matter basis) composed of maize silage, grass haylage, soyabean meal, maize germ, high moisture grain silage, maize gluten feed, citrus pulp and mineral mixture.
To estimate MUN concentration as a function of DIM, a model to describe the lactation curve, proposed by Wood (1967), was used.Later, LN and CS were added, verifying the existence of coincidence among curves.Likelihood ratio test was used to verify the signifi cance of model adjustments as well as other effects.A test for lack of fi t was also calculated as suggested by Neter et al. (1996).Models were adjusted by square means (signifi cance level of 5%), using the NLIN procedure of SAS (1999).

RESULTS AND DISCUSSION
On average, milk production of cows was 35.8 kg/d and MUN was 13.3 mg/dL.The MUN curve, estimated using Wood's model, showed a signifi cant adjustment for MUN concentrations (P<0.001) by the likelihood ratio test and in addition, the test for lack of fi t was not signifi cant (P=0.0576).Peak of MUN was 14.1 mg/dL, occurring on the 83 rd day in milk.In agreement with Carlsson et al. (1995), MUN concentration was lower in the beginning of lactation, considering all LN.
To verify the effect of lactation number, a model was built where the parameters of Wood's curve were estimated for each of the fi ve LN at the same time, proceeding then, the likelihood ratio test to compare the model containing just one curve for all LN with that one with a curve for each LN.There was an effect of LN (P<0.0001) on the considered curve; in other words, at least two lactation curves differed amongst themselves.However, it was observed that two curves (4 th and 5 th lactations) were coincident (P=0.6216).Coincidence among curves in LN from 3 to 5 was also tested, and this test showed signifi cant differences (P=0.0108)among at least two of the three LN studied in the model.Curves for the 2 nd and 3 rd lactation were different (P=0.0002), and so were curves for 1 st and 2 nd lactation (P<0.0001).Therefore, it is possible to affi rm that there VALIDATION OF MODELS FOR PREDICTING MILK UREA-N is no difference between curves in 4 th and 5 th lactation, i.e. it is only necessary to estimate four curves from the fi ve LN studied.These results agree with Canfi eld et al. (1990) who suggested that LN should be considered, because primiparous had lower MUN concentration than the multiparous.The effect of calving season was verifi ed on the MUN curves.The likelihood ratio test showed a signifi cant difference (P<0.0001)among CS for at least one of LN.However, coincidence was observed between curves for the 4 th and 5 th lactation (P=0.1390).Coincidence was also tested among CS for curves describing LN from 3 to 5 and this test showed signifi cant differences (P<0.0057)among at least two of the three LN in the studied model.So, it was necessary to estimate only eight curves (Table 1) out of the possible ten.For cows calved in wintertime, MUN concentrations followed the lactation curve pattern, increasing until the peak, and then declining until the end of lactation, independent on LN.Cows that calved in summertime began the lactation with higher or similar MUN concentration, increasing little and gradually during lactation, reaching peak much later compared with those that calved in the wintertime and remaining like that, with no decrease.

CONCLUSIONS
Milk urea nitrogen concentrations showed a similar pattern to the lactation curve as a function of days in milk, differing among 1 st , 2 nd , 3 rd and 4 th -5 th lactations, being infl uenced by calving season.

Table 1 .
Estimates of Wood's curve parameters with lactation number and calving season effects, standard error of mean and confi dence interval, 95%