(METER) taken on the farm and to evaluate

Evaluation of the Hydrometer for Testing Immunoglobulin GI Concentrations in Holstein Colostrum1 ABSTRACT Hydrometer measurement in globulin and IgGl concentration measured by the radial immunodiffusion technique were compared for 915 samples of first milking colostrum from Holstein cows. Least squares analysis of the relationship between hydrometer measurement and IgGl concentration was improved by log transformation of IgGl concentration and resulted in a significant linear relationship between hydrometer measurement and loglo IgGl concentration; 9 =.469. At 50 mg of globulidml of colostrum, the recommended hydrometer cutoff point for colostrum selection, the sensitivity of the hydrometer as a test of IgGl concentration in Holstein colostrum was 26%, and the negative predictive value was 67%. The negative predictive value and sensitivity of the hydrometer as a test of IgGl in Holstein colostrum was improved, and the cost of misclassification of colostrum was minimized, when the cutoff point for colostrum selection was increased above the recommended 50 mg/ml. (Key words: colostrum, colostrometer, immunoglobulin GI, specific gravity) Abbreviation key: METER = hydrometer reading of globulin concentration, PV- = negative predictive value, SRID = single radial immunodiffusion IgGl concentration. Received October 7, 1993. Accepted January 28, 1994. *This research was funded by Grant 10A 3072 0656 from the Agricultural Research Center, Washington State University, and Grant 13C 2560 0075 from the Washington State Dairy Products Commission. ZField Disease Investigation Unit, Department of Veterinary Clinical Sciences. 3Department of Veterinary Microbiology and Pathol- OgY. 1994 J Dairy Sci 77:1761-1767 1761 LORI C. PRITCHElT,2 CUVE C. GAY,? DALE D. HANCOCK,z and THOMAS E. BESSER, College of Veterinary Medicine Washington State University Pullman 99164-6610 INTRODUCTION The importance of colostral Ig absorption by the neonatal calf in establishing passive immunity is well documented (1, 3, 8). Ingestion of a minimum of 100 g of Ig is recommended for adequate passive immunity (2, 6). In bovine colostrum, IgCl accounts for approximately 80% of the total globulin concentration and is the predominant Ig absorbed from the calf's intestine to the systemic circulation during the first 24 h postpartum (3). Holstein calves fed colostrum with an esophageal feeder or nipple bottle have a lower prevalence of failure of passive transfer of IgGl than naturally suckled calves (2). However, the artificially fed calf ingests adequate Ig only if the concentration of Ig in the colostrum is sufficient in the volume consumed. In 1980, Fleenor and Stott (4) described a hydrometer test to estimate Ig concentration in bovine colostrum from the specific gravity of fiesh whole colostrum. The hydrometer is calibrated in globulin concentration at intervals of 5 mgml from 0 to 180 mg/ml, displayed with three color-coded quality regions: poor (red) for <22 mg/ml, moderate (yellow) for 22 to 50 mg/ml, and excellent (green) for >50 mg/ ml. The objective of this study was to compare IgGl concentrations measured in the laboratory by the single radial immunodiffusion (SRID) technique with hydrometer readings in globulin (METER) taken on the farm and to evaluate the sensitivity, specificity, and negative predictive value of the hydrometer as a test for IgGl concentration in Holstein colostrum. MATERIALS AND METHODS Sample and Data Collection From September 1983 to January 1988,915 samples of first milking colostrum were obtained fiom Holstein cows on a single dairy farm in central Washington. Colostrum

1762 PRITCHEYIT ET AL. management, sample collection, and data were recorded as previously described (1 1). Briefly, the newborn calf was removed from the dam prior to any nursing and fed 2.84 L (3 qt) of stored colostrum with an esophageal feeder. The first milking colostrum of the dam was machine-milked, weighed, and stored for feeding the next calf born. In addition, colostrum globulin concentrations were measured with a commercially available hydrometer (Nasco, Fort Atkinson, WI) calibrated in globulin concentration in intervals of 5 mg of globulidml of colostrum. After the first milking colostrum was weighed, the bucket of colostrum was equilibrated to room temperature (18 to 22 C) for 6 to 8 h before a hydrometer was floated in the bucket and the globulin concentration recorded. Approximately 25 ml of colostrum sampled from the bucket were stored in a 30-ml polypropylene bottle at -20 C until analyzed for IgGl concentration. Ig Analysis Colostrum IgGl concentrations, expressed in milligrams of IgGl per milliliter of colostrum, were determined by results of the SRID of Heenor and Stott (9, which was modified as previously described (11). Statistical Analysis The data were analyzed initially by least squares linear regression (12). Estimates of the relationship between SRID and METER were calculated; SRID was the dependent variable, and METER was the regressor variable. Logarithmic transformation of SRID helped to linearize the sample data and improved the measured relationship between SRID and ME- TER (12). To calculate sensitivity, specificity, and predictive value of the hydrometer as a test of IgGl concentration, two-way frequency tables were constructed (13). For these calculations, METER was considered to be the diagnostic test value, and SRID was considered to be the true value. A globulin concentration of 50 mgl ml, which is the cutoff point between moderate (yellow region) and excellent (green region) on the hydrometer scale, was used to classify colostrum into one of two groups: low METER (40 mglml) or high METER (250 mglml). Because 80% of the total globulin TABLE 1. Two-way frequency table to calculate sensitivity, specificity> and negative predictive value3 of the hydrometet test to classify colostrum with IgGl concentrations c40 mglml or 240 mg/ml. SRID5 Low High Total METER6 Low 95a 12b 107 High 269 539d 808 Total 364 551 1Sensitivity = a/(a + c) x 100 = 95/364 x 100 = 26%. *Specificity = d/(b + d) x 100 = 539651 x 100 = 98%. JNegative pdctive value = d/(c + d) = 539/808 x 100 = 67%. 4- hydrometer estimates globulin concentration in bovine colostnun from the specific gravity of fresh whole COlOStnun. %ingle radial immunodiffusion IgGl concentration in milligrams of IgGl per milliliter of colostrum. Low SRID is <40 mg/ml, and high SRID is 240 mglml. SThe hydrometer reading of globulin concentration in milligrams of globulin per milliliter of colostrum. Low METER is d O mglml, and high METER is 250 mg/ml. concentration in colostrum is IgG1, the group classifications for colostrum based on the measured IgGl concentration were low SRID (c40 rnglml) or high SRID (240 mglml). Sensitivity was defined as the probability that a colostrum with an IgGl concentration of c40 mglml would be classified as low by the hydrometer. Sensitivity was calculated as the number of colostrums classified as low by METER and SRID divided by the total number of low colostrums by SRID: a/(a + c) x 100. Specificity was defined as the probability that colostrum with an IgGl concentration 240 mgl ml would be classified as high by the hydrometer and was calculated as the number of colostrums classified as high by METER and SRID, divided by the total number of high colostrums by SRID: d/(b + d) x 100. The negative predictive value (PV-) was defined as the probability that colostrum classified as high by the hydrometer was truly high; PV- was calculated as the number of colostrums classified as high by METER and SRID divided by the total number of lugh colostrum by METER: d/(c + d) x 100 (Table 1). For artificially fed calves, the volume of the colostrum fed is as important as the IgGl

OUR WDUSTRY TODAY 1763 concentration of the colostrum to ensure that 100 g of Ig are ingested by the newborn calf (2).A more practical approach to evaluate the hydrometer is to access the ability to discriminate between low and high colostrum, given the fixed feeding volume administered on the farm. During the study, the dairy farmer routinely fed each newborn calf 2.84 L of colostrum with an esophageal feeder. A colostrum must have a minimum IgGl concentration of 35.2 mg/ml to provide 100 g of IgGl in 2.84 L. Similarly, for feedings of 1.89 L (2 qt) or 3.78 L (4 qt), the minimum IgGl concentrations were 52.9 and 26.5 mg/ml, respectively. These IgGl concentrations were used as the SRID cutoff point to compute the sensitivity, specificity, and PV- of the hydrometer test for each of the three fixed feeding volumes. To determine optimal hydrometer cutoff points, misclassification costs were calculated using the formula: Figure 1. Frequency distribution and cumulative percentage of colostral hydrometer reading. model I.2 =.469. The least squares estimates of the linear relationship resulted in the regression equation for Holstein colostrum: loglo SRID = 1.2775 + (.8 x METER x.0055). where j is a hydrometer cutoff point, c is the number of false high colostrum at and above cutoff j, b is the number of false low colostrum below cutoff j, $h is the misclassification cost of false high colostrum, and $1 is the misclassification cost of false low colostrum. Rather than assign actual costs of misclassification, relative costs were assigned to the relationship $&$I as 1:1, 21, and 3:l. Hydrometer cutoff points determined by this procedure were then used as the METER value to compute the cost optimized sensitivity, specificity, and PV- of the hydrometer test. This relationship is shown in Figure 3. The equation fitted a curve with IgGl concentrations much lower than those predicted by the hydrometer. For the cutoff points of METER and SRID described previously, 12% of the colostrum had globulin concentrations 4 0 mg/ml measured by METER, and 40% of the colostrum had IgGl concentrations <40 mg/ml measured RESULTS The frequency distribution of METER appeared to be normally distributed; mean globulin reading was 82.1 mg/ml, and standard deviation was 30.6 mg/ml (Figure l). This result was in contrast to the frequency distribution of SRID, which was slightly skewed to the right, for the same sample (Figure 2). The mean IgGl concentration was 48.1 mg/ml, and the standard deviation was 21.9 mg/ml. The linear relationship between METER and loglo SRID was significant (Table 2); 100, Commtmtlon (mglml) Figure 2. Frequency distribution and cumulative percentage of colostral IgGl concentration. Journal of Dairy Science Vol. 77. No. 6, 1994

1764 PRITClETT ET AL. TABLE 2. Analysis of variance and least squares estimates for the regression model: loglo IgGl (mg/ml) =.8 x hydrometer reading (mdml). Analysis of variance ~ Source df ss MS F PsF Model 1 16.5957 16.5957 807.016.owl Error 913 18.7752.0206 Corrected total 914 35.3709 RZ.4692 Adjusted RZ.4692 Parameter estimates Parameter Variable df estimate SE Student's t P > t Intercept 1 1.2775,0136 94.066.o001 METER1 1.0055.m 28.408.OOO1 'Hydrometer reading of globulin concentration. by SRID. The sensitivity, specificity, and PVof the hydrometer as a test to classify colostrum with IgGl concentrations e40 mg/d or 240 mg/ml were 26,98, and 67%. respectively (Table 1). Using the METER cutoff point of 50 mghl to classify low or high colostrum for the 2.84-L feeding volume, the sensitivity, specificity, and PV- of the hydrometer test were 32, 97, and 78%. respectively. The misclassification cost of false highs to false lows was minimized at globulin concentrations of 60, 70, and 85 mg/ml, all above the recommended METER boundary of 50 mdml (between 160 r O o n I od io 40 80 80 loo 120 140 180 180 200 HYd--(mghnl) Figure 3. The relationship of colostral IgGl concentration to colostral hydrometer reading with the least squares regression line. moderate and excellent on the hydrometer) for the three relative costs: 1:1, 2:1, and 3:1, respectively (Figure 4, Table 3). The optimal cutoffs for the 1.89, 2.84, and 3.78-L feeding volumes are summarized in Table 4. DISCUSSION The distribution of Holstein colostral IgGl concentrations in this sample was similar to distributions reported for other Holstein dairy farms (2). Because the volume of colostrum fed to each calf on this farm was fixed at 2.84 L, the concentration of IgGl in the colostrum was the critical variable to control to ensure that adequate Ig (>IO0 g) was presented to the newborn calf. Approximately 29% of the colostrum on this dairy was low IgGl colostrum (<35.2 mg/ml); therefore, any method of selecting against the low colostrum would presumably improve chances for adequate passive transfer. Fleenor and Stott (4) calibrated the hydrometer based on a linear relationship between colostral globulin concentration and colostral specific gravity for 29 postpartum colostrums with an r2 of.699. The relationship measured in our study was neither linear nor as well correlated as that reported by Fleenor and Stott (4). In our sample, variation in SRID increased as METER increased. The log transformation of SRID improved the fit of the curve and increased the amount of variation in SRID explained by METER (9 =.469).

~ ~~ loo-. 0 ' f OUR INDUSTRY TODAY 1765 TABLE 4. Optimal hydrometer cutoff points for feedings of 1.89 L (2 qt), 2.84 L (3 qt), and 3.78 L (4 qt). Relative cost of misclassificationl Feedings 1:l 2:l 3:l Q 1.89 110 110 125 2.84 60 70 85 3.78 45 55 60 ]The relative cost of misclassification is defined as the cost of accepting and feeding false high colostrum to the cost of rejecting and discarding false low colostrum ($h: $1). I80 i-5 srn = c + b j 0 where j = hydrometer cutoff point, c = false high colostrum at cutoff point j and above, b = false low colostrum below cutoff point j, $h = cost of accepting and feeding false high colostrum, and $1 = cost of rejecting and discarding false low colostrum. The relative costs of false high colostrum to false low colostrum ($h:$1) are 1:l 0, 2:l (+), and 3:l (A). A recent study by Mechor et al. (10) reported a significant effect of temperature of the colostrum on hydrometer measurements. Although the colostrum temperatures were not measured in the present study, the hydrometer readings were consistently taken on fresh TABLE 3. Two-way frequency table results for a 2.84-L (3q) feeding volume. Hydrometer cutoff point 50 60 70 85 1:21 1:11 21' 3:11 Sensitivity, 8 32 47 63 86 Specificity, 9b 97 92 85 63 Negative predictive value, % 78 81 95 92 Colostrums saved. % 88 81 71 49 1The relative cost of misclassification is defined as the cost of accepting and feeding false high colostrum to the cost of rejecting and discarding false low colostrum ($h: $1). colostrum equilibrated to room temperature (18 to 22'C) for 6 to 8 h, which reflect farm conditions under which the hydrometer was intended for use. In our study, the recommended globulin cutoff point of 50 mg/ml for classifying colostrum with IgGl concentrations c40 mdml or 240 mg/ml was unsatisfactory because of very low sensitivity. Mechor et al. (9) also demonstrated that values measured by the hydrometer overestimated IgG content. However, given the 2.84-L feeding volume used by the dairy in the present study, determining the ability of the hydrometer to classify colostrum with IgGl concentrations <35.2 mg/ml or 235.2 mg/ml was considered to be a fairer assessment of the test. At the recommended METER = 50 mg/ml cutoff, the test has a specificity of 97% and a slightly improved sensitivity of 32% for correctly classifying colostrum with IgGl concentrations <35.2 mg/ml or 235.2 mg/ml. The low sensitivity in both of these cases indicated that more false high colostrum was accepted than false low colostrum rejected. Failure of passive transfer increases morbidity and mortality in neonatal calves (1,7, 8). Therefore, accepting and feeding false high colostrum can be viewed as worse than discarding false low colostrum. Given the tradeoff between sensitivity and specificity based on the cutoff point chosen, 97% specificity and 32% sensitivity were likely not optimized for cost. Our computations confmed this supposition, showing that the optimal METER cutoff point is between globulin concentrations of 60 and 85 mg/ml for a 2.84-L feeding, depending on the relative costs assigned to false highs Journal of Dairy Science Vol. 77, No. 6. 1994

1766 PRITCHE'IT ET AL and false lows (Figure 4). The higher relative cost assigned to false high colostrum was based on the assumption that feeding a calf colostrum with low IgGl concentration is a more costly mistake than falsely discarding colostrum with high IgGl concentration. The abundant production of first milking colostrum for Holstein cows [8.5-kg average for this farm (ll)] allows a higher METER cutoff point, which improves the sensitivity and PV- of the hydrometer test while maintaining a sufficient supply of colostrum for the first 2.84-L feeding of calves (Table 3). A 2-L feeding volume has been recommended for first colostral feedings of newborn calves (14), and commercially available nipple bottles (Nasco) designed for calf feeding hold 1.89 L. A feeding of 1.89 L of colostrum must contain a minimum IgGl concentration of 52.9 mg/ml to provide 100 g of IgG1. Using this cutoff for the classification of Holstein colostrum as SRID = low or high in the two-way frequency table and misclassification cost calculations (see Materials and Methods), the METER cutoff point must be increased to a globulin concentration of 110 mg/ml to minimize misclassification with a relative cost of 1: 1 (Table 4). The PV- of the hydrometer is increased from 37% for the 50 mg/ml cutoff to 75% for the 100 mg/ml cutoff point; however, less than 20% of the colostrum in this sample would be acceptable for this feeding volume (Figure 1). In contrast, a 3.78-L feeding regimen allows a much lower SRID cutoff point because colostrum with an IgGl concentration 126.5 mg/ml provides adequate Ig mass. For this SRID cutoff, a METER cutoff point of 45 mg/ ml minimizes misclassification with a relative cost of 1 : 1 (Table 4). The PV- of the hydrometer is 91% for the cutoff point of 45 mg/ml, and 89% of the colostrum would be acceptable. However, selection based on METER only slightly improves the chances of acceptable colostrum because 86% of colostrum samples on this dairy have IgGl concentrations 226 mg/ml (Figure 2). The study farm now routinely feeds 3.78 L of first milking colostrum with an esophageal feeder as soon as possible after birth. To date, no adverse effects of this feeding regimen have been noted. The exact ratio of costs of misclassifying colostrum is debatable. Other factors in addi- tion to adequate passive immunity influence neonatal calf morbidity and mortality (1). However, even if the relative misclassification cost is 1:1, with the distribution of colostral IgGl concentrations seen on this and other Holstein dairies (2), the hydrometer cutoff point must be increased above a globulin concentration of 50 mg/ml to minimize the cost of misclassification for 1.89 or 2.84-L feeding volumes. Alternatively, a 3.78-L feeding regimen alone could improve chances of providing adequate IgGl mass with little improvement from selection by METER. CONCLUSIONS This study was conducted on a commercial dairy farm under the conditions intended for hydrometer use. Forty-seven percent of the variation in Holstein colostral IgGl concentration can be explained by METER. By increasing the METER cutoff point above the recommended globulin concentration of 50 mg/ml, the calf feeder will improve the PV- of the hydrometer test and minimize the cost of misclassification. ACKNOWLEDGMENTS The authors gratefully acknowledge the assistance and cooperation of Gustafson Farms, Sunnyside, WA. REFERENCES 1 Besser, T. E., and C. C. Gay. 1985. Septicemic colibacillosis and failure of passive transfer of colostral immunoglobulin in calves. Vet. Clin. Noah Am. Food Anim. Pract. 1445. 2Btsser, T. E.. C. C. Gay, and L. C. Pritchett. 1991. Comparison of three methods of feeding colostnrm to dairy calves. J. Am. Vet. Med. Assoc. 198:419. 3 Butler, J. E.. 1983. Bovine immunoglobulins: an augmented review. Vet. Immunol. Immunopathol. 443. 4Fleenor, W. A., and G. H. Stott. 1980. Hydrometer test for estimation of immunoglobulin concentration in bovine. colostrum. J. Dairy Sci. 63:973. SFleenor, W. A.. and G. H. Stott. 1981. Single radial immunodiffusion analysis for quantitation of colostral immunoglobulin concentration. J. Dairy Sci. 64:740. 6 Kruse. V. 1970. Absorption of immunoglobulin hm colostrum in newborn calves. Anim. prod. 12627. 7 Logan, E. F. 1974. Colostral immunity to colibacillosis in the neonatal calf. Br. Vet. J. 130:405. 8Lomba, F., I. Fumiere, M. Tshibangu, G. Chauvaux, and V. Bienfet. 1978. Immunoglobulin transfer to calves and health problems in large bovine units. Ann. Rech. Vet. 9:353.

9 Mechor, G. D., Y. T. Grohn. L. R. McDowell, and R. J. Van Saun. 1992. Specific gravity of bovine ~010smm immunoglobulins affected by tempem and colostrum components. J. Dairy Sci. 793131. 10 Mechor, G. D,, y. T, ~ ~ and h R, J,, van saun. 1991. Effect of temperature on colostrometer readings OUR INDUSTRY TODAY 1767 influencing immunoglobulin GI concentration in Holstein colostrum. J. D;ury Sci. 74:2336. 12SASe user s Guide: StatlStiCS, Version 5 Edition. 1985. SAS Inst., Inc., Cary, NC. 13 Smith, R. D. 1991. Veterinary Clinical Epidemiology, A Problem-Oriented Approach. Butterworth- Heineman, Stoneham, MA. for Of in 14StoK G. H., D. B. Mm, B. E. Men&=, and G. T, bovine colostrum. J. Dairy Sci. 743940. Nightengale. 1979. Colostral immunoglobulin transfer 11 Pritchett, L. C., C. C. Gay, T. E. Besscr, and D. D. in calves. II. The rate of absorption. J. Dairy Sci. 62: Hancock. 1991. Management and production factors 1766.