Upper-normal prediction limits of lymphocyte counts for cattle not infected with bovine leukemia virus.

A study was conducted to develop valid estimates of lymphocyte count (LC; cells per microliter) of individual, clinically normal dairy cattle. Estimated weighted regression was used on repeated measures of individual LC to examine 6 models predicting LC as a function of age in cattle not infected with bovine leukemia virus. The generalized growth curve model of analysis of variance was used to estimate intercepts, slopes, and prediction limits for the models and to compare the LC-to-age relationship between Holstein and Guernsey breeds. The best-fitting model (P = 0.0001) with the narrowest prediction interval was LC = 4,414.4 - 84.6X, where X = (age -48) if age less than or equal to 48 months, and X = 0 if age greater than 48 months, and 163.6 and 8.1 are the SE of the estimates, respectively. Upper one-sided 95%-predicted normal LC tended to be higher than estimates derived from traditional hematologic keys that use confidence limits of mean LC. Difference was not found in the LC-to-age relationship between the Holstein and Guernsey cattle (P = 0.67). Results of this study provided estimates of normal LC that are more specific in diagnosing lymphocytosis in individual cattle.