H. Laevens et al., COMPARISON OF SOMATIC-CELL COUNTING WITH THE FOSSOMATIC-360 AND THE SOMASCOPE, Netherlands milk and dairy journal, 50(3), 1996, pp. 335-346
Duplicate samples of mixed milk (n = 13 927) were collected to compare
somatic cell counting with the Somascope and the Fossomatic 360. To e
stimate sampling and assay variability, somatic cell count (C) was mea
sured with the Fossomatic 360 on 295 duplicate milk samples. The total
variance was 1042 x 10(6). The contribution of the measuring instrume
nt (Fossomatic 360) and the sampling procedure were 12.7 and 87.3%, re
spectively. The overall variance of the paired differences of the loga
rithmically transformed numerical value of C (expressed in mu L(-1)) (
ln{C}) of the 295 duplicate milk samples all measured with the Fossoma
tic 360, was 0.08. Additionally, accuracy of the Somascope was evaluat
ed on 9558 duplicate milk samples. In one of the duplicate milk sample
s, C was measured with the Fossomatic 360; in the other, with the Soma
scope. The variance of the paired differences of ln{C} between the Som
ascope and the Fossomatic 360 was 0.14. Comparison of the In{C} per cl
ass (class width = 1 on the log, scale) showed 77.4, 28.1 and 11.3% ov
erestimation of In{C}, measured with the Somascope, in the classes les
s than or equal to 2,2-3 and 3-4, respectively, and 1.4, 2.6 and 4.6%
underestimation in the classes 6-7, 7-8 and >8, respectively, indicati
ng a lower accuracy of the Somascope in these classes. The Pearson cor
relation of ln{C} between the Fossomatic 360 and the Somascope was 0.9
6. The Pearson correlation of In{C} of the duplicate milk samples meas
ured twice with the Fossomatic 360 was 0.97. Simple linear regression
(Somascope against Fossomatic) showed a significant intercept, b(0) (/-s.e.) = 1.13 (+/-0.011) and the slope, b(1) (+/-s.e.) = 0.803 (+/-0.
0022), differed from the ideal value of 1 (P < 0.001). The overall res
idual variance of the linear regression analysis was 0.08. With a weig
hted non-linear regression analysis, a relationship X(som) = P-1 + X(f
os)/(1 + P(2)X(fos)) was calculated, where P-1, is a constant that des
cribes the overestimate of C (greatest at lowest values) with the Soma
scope, and P(2)X(fos) is a term that reflects the underestimate with t
he Somascope at higher values. The overall prevalence of cows with C h
igher than 250 x 10(3) mL(-1), based on the results of the Fossomatic
360 and on the Somascope, was 20.2 and 21.9%, respectively. Test agree
ment (K), sensitivity and specificity for detecting cows with C > 250
x 10(3) mL(-1) using the Somascope as compared to the Fossomatic 360,
was 0.89, 0.95 and 0.97, respectively. The Somascope, when correctly c
alibrated, is an acceptable alternative for measuring C for field tria
l purposes and for dairy herd improvement programme.