Soo-Han Lee and Jun Jaekal contributed equally to this work.
Enzyme-Linked Immunosorbent Assay, Single Radial Immunodiffusion, and Indirect Methods for the Detection of Failure of Transfer of Passive Immunity in Dairy Calves
Version of Record online: 14 FEB 2008
Copyright © 2008 by the American College of Veterinary Internal Medicine
Journal of Veterinary Internal Medicine
Volume 22, Issue 1, pages 212–218, January–February 2008
How to Cite
Lee, S.-H., Jaekal, J., Bae, C.-S., Chung, B.-H., Yun, S.-C., Gwak, M.-J., Noh, G.-J. and Lee, D.-H. (2008), Enzyme-Linked Immunosorbent Assay, Single Radial Immunodiffusion, and Indirect Methods for the Detection of Failure of Transfer of Passive Immunity in Dairy Calves. Journal of Veterinary Internal Medicine, 22: 212–218. doi: 10.1111/j.1939-1676.2007.0013.x
- Issue online: 14 FEB 2008
- Version of Record online: 14 FEB 2008
- Submitted March 27, 2007; Revised July 4, 2007; Accepted September 20, 2007
- Diagnostic performance;
- Immunoglobulin G;
- Receiver operating characteristic
Background: Confirmatory tests for failure of transfer of passive immunity (FTPI) in dairy calves require direct measurements of the serum immunoglobulin G concentration. Enzyme-linked immunosorbent assay (ELISA) has advantages over single radial immunodiffusion (SRID) in terms of cost and time.
Objectives: To evaluate the agreement between ELISA and SRID, and to compare the diagnostic performance of ELISA with indirect methods, in the detection of FTPI in calves.
Animals: One hundred and fifteen dairy calves (aged 0–10 days) from 23 calf-rearing facilities.
Methods: Prospective, observational study. The agreement between SRID and ELISA was determined by the Bland-Altman method. Fixed bias (SRID − ELISA) was calculated. For comparison of the diagnostic performance of ELISA with indirect methods, sensitivity, specificity, and area under the curve (AUC) of receiver operating characteristic (ROC) curves were calculated at cut-off values of 500 and 1,000 mg/dL.
Results: The agreement between SRID and ELISA was 94%. Fixed bias (SRID − ELISA) was 140 ± 364 mg/dL. The AUC and sensitivity of ELISA at the cut-off value of 1,000 mg/dL were higher than those of indirect methods (P<.004). The specificity of ELISA at the cut-off value of 1,000 mg/dL was not higher than that of indirect methods, except for serum total protein concentration assay.
Conclusion and Clinical Importance: ELISA exhibited good diagnostic performance and good agreement with SRID. ELISA is an adequate method for both screening and confirmatory tests for FTPI in dairy calves at the cut-off value of 500 mg/dL.
The transfer of immunoglobulins from the dam to the neonate through colostrum is important for protection of the neonate from infectious diseases. Serum or plasma concentration of immunoglobulin G (IgG) is a good indicator of the concentration of other immunoglobulins, which explains why the concentration of IgG can be used as the reference value in most discussions of colostral immunoglobulin content and serum immunoglobulin concentration in calves.1
The diagnosis of failure of transfer of passive immunity (FTPI) requires the serum immunoglobulin concentration of the calf to be measured directly or indirectly. Several indirect methods, such as serum total protein (TP) concentration assays,2 sodium sulfite precipitation (SSP) test,3 10% glutaraldehyde reagent test,4 and zinc sulfate turbidity (ZST) test,5 have been used widely in the field. However, serum IgG concentration cannot be measured quantitatively with these indirect methods. A specific laboratory evaluation of immunoglobulin concentration is preferable to these field techniques under conditions when confirmation of FTPI is essential, such as plasma administration therapy.1 Single radial immunodiffusion (SRID) has been used for the confirmatory diagnosis of FTPI and is currently considered the gold standard for the direct measurement of serum IgG concentrations in neonate calves.2,5,6 However, it requires a long diffusion time (18–24 hours) and is expensive and laborious for analysis of a large number of samples. Furthermore, there are several commercially available bovine SRID kits. However, agreement between bovine SRID kits has not been reported.
Enzyme-linked immunosorbent assay (ELISA) has recently become available for use in calves.8 Being capable of directly measuring IgG concentration,9,10 ELISA has advantages over SRID in terms of cost, time, and the capacity for measuring a large number of samples at once, which might render it useful for the confirmatory diagnosis of FTPI in herds.
The aims of this study were to evaluate the agreement and bias between the IgG concentrations measured with 2 commercially available bovine SRID kits, and to evaluate the agreement and bias between the IgG concentrations measured by ELISA and SRID. Additionally, the diagnostic performance of ELISA for FTPI was compared with indirect methods using receiver operating characteristic (ROC) curve analysis in condition of reference value from SRID.
Materials and Methods
Sample Collection and Processing
Blood samples were obtained by jugular venipuncture from 115 dairy calves, aged 0–10 days and housed at 23 calf-rearing facilities. Samples were collected in serum separation tubesa and stored at room temperature until clotting (about 10–20 minutes). Serum was harvested after centrifugation and stored at − 80 °C until assay.
Indirect Measure Methods
Using 14, 16, and 18% sodium sulfiteb solutions, the SSP test was performed according to the method described previously.3 Test results were recorded as 0, 1, 2, or 3, where 0=no turbidity in all 3 solutions, 1=turbidity with the 18% solution only, 2=turbidity with the 16 and 18% solutions, and 3=turbidity in all 3 solutions.
The modified glutaraldehyde coagulation (MGC) test was performed according to the method described previously,4 using 5, 7.5, and 10% glutaraldehydec solutions. Test results were recorded as 0, 1, 2, or 3, where 0=no coagulation in any solution, 1=coagulation with the 10% solution only, 2=coagulation in the 7.5 and 10% solutions, and 3=coagulation in all 3 solutions.
The ZST test was performed according to the method described previously,5 using zinc sulfate (ZnSO4)d solutions of 200, 250, 300, 350, and 400 mg/L concentration. The solution was used within 24 hours after preparation and maintained in a sealed container until use. Test results were recorded as 0, 1, 2, 3, 4, or 5, where 0=no turbidity in any of the 5 concentrations, 1=turbidity with the 400 mg/L concentration only, 2=turbidity in the 350 and 400 mg/L concentrations, 3=turbidity in the 300, 350, and 400 mg/L concentrations, 4=turbidity in the 250, 300, 350, and 400 mg/L concentrations, and 5=turbidity in all 5 concentrations.
Serum TP concentrations were measured with a nontemperature-compensating refractometer.e
Direct Measure Methods (SRID and ELISA)
Two commercially available kits (SRID1f, SRID2g) were used to determine serum IgG concentration. Test references and serum samples (5 or 3 μL of each) were applied to serial SRID plates containing agarose gel with anti bovine IgG. The plates were left undisturbed for 18 hours at room temperature after adding the samples. The resulting ring diameters were measured with an imaging analyzerh and the IgG content of the samples was calculated by regression analysis. A standard curve was generated with reference sera supplied by the manufacturer (r2≥0.979 in SRID1; r2≥0.997 in SRID2). If sample IgG concentration was below lowest standard sera, these sample concentrations were calculated by means of a new standard curve that was generated by diluted standard sera.
Serum IgG determination by ELISA was carried out following the manufacturer's instructions. Briefly, the assays were performed in microtiter platesi at room temperature. For isotype ELISA, wells were coated for 1 hour with affinity-purified sheep anti bovine IgGj diluted 1 : 100 in coating buffer (0.05 M sodium carbonate, pH 9.6). Plates were washed 3 times with 50 mM Tris, 0.14 M NaCl, 0.05% Tween 20, pH 8.0, incubated for 30 minutes in the same buffer (blocking), and then washed 3 more times. After the 3rd wash, 100 μL of bovine reference serumk or diluted calf sera were added to each well, and the plates were incubated for 1 hour. Appropriate dilutions of reference serum (500, 250, 125, 62.5, 32.25, 15.623, and 7.8 ng/mL) were made to generate a standard curve and were assayed in triplicate; unknown serum samples were assayed in triplicate at the dilution 1 : 100,000. Wells were then washed 5 times, 100 μL of sheep anti bovine IgG (1 : 150,000) conjugated to horseradish peroxidase (HRP)l was added, and plates were incubated for 1 hour. After incubation, plates were washed 5 times and enzyme substratem was added. Reactions were stopped after 30 minutes with 2 M H2SO4 and the optical density at 450 nm was determined with an automated plate reader.n Median values were calculated for the triplicate readings for each standard, control, and sample, and the 0 reading from each mean value was subtracted. The standard curve was generated by means of a 4-parameter logistic (4-PL) curve fit (r2≥0.999).
For each sample, the IgG concentration measured with SRID1 was used as the reference to which the other assays were compared, because the reference serum of ELISA in this study was the bovine reference serum of SRID1. This was confirmed by the lot numbers of the bovine reference sera of SRID1 and ELISA kits, which were provided by Bethyl Laboratories.
The agreement between SRID1 and SRID2 and between SRID1 and ELISA was determined by the Bland-Altman method11 and intraclass correlation coefficients (ICCs).12 In clinical comparisons of a new measurement technique with an established one, the 2 methods are required to agree sufficiently for one to replace the other. An approach using the degree of agreement between 2 methods based on graphical techniques and simple calculations may be more useful than correlation analysis.11 When one is interested in the relationship among variables of a common class, which means variables that share both their metric and variance, ICCs are alternative statistics for measuring homogeneity, not only for pairs of measurements but also for larger sets of measurements. For the Bland-Altman method, the IgG concentrations measured with SRID1, SRID2, and ELISA were log-transformed to standardize the limits of agreement in overall concentrations.11,13 The agreement was calculated by Bland-Altman as follows:
where A is the number of samples located within the limits of agreement and B is the number of total samples.
We calculated the mean difference between 2 methods (SRID1 − ELISA or SRID1 − SRID2), the standard deviation (SD) of the differences, and the limits of agreement (mean ± 1.96 × SD of the difference). The mean difference is the estimated bias, and the SD of the differences represents the random fluctuations around this mean value. A mean value of the differences that differs significantly from 0 on the basis of a 1-sample t-test is indicative of fixed bias. When there exists a proportional bias, the difference tends to increase with increasing mean value. The existence of proportional bias indicates that the methods do not agree equally across all measurements; that is, the limits of agreement depend on the actual measurement. To formally evaluate this relationship, the differences in measurements (SRID1 − SRID2 or SRID1 − ELISA) were regressed using the mean value of 2 measurements (SRID1 and SRID2, or SRID1 and ELISA). The Kruskal-Wallis test was used to compare the biases in IgG concentrations of <500 mg/dL, 500–1,000 mg/dL, and >1,000 mg/dL, and then pairwise multiple comparisons were performed using the Mann-Whitney U-test.
Calves with <500 mg IgG/dL are particularly prone to septicemic Escherichia coli infection, and those with 500–1,000 mg IgG/dL are defined as having partial FTPI.1 In this study, calves with <500 mg IgG/dL and with 500–1,000 mg IgG/dL are therefore defined as having serious and partial FTPI, respectively. Therefore, the cut-off concentrations of IgG were 500 mg/dL and 1,000 mg/dL for the comparison of diagnostic performance between ELISA and indirect assays. For each sample, IgG concentrations obtained from SRID1 were used as a reference. Sensitivity, specificity, and area under the ROC curves (AUC) were calculated at each cut-off IgG concentration. Several summary indices are associated with the ROC curve. One of the most popular measures is AUC.14 AUC is a measure of the overall performance of a diagnostic test and is interpreted as the average value of sensitivity for all possible values of specificity.15 It can take on any value between 0 and 1, because both the x- and y-axis have values ranging from 0 to 1. The closer the AUC to 1, the better the overall diagnostic performance of the test. A test with an AUC value of 1 is one that is perfectly accurate. The AUCs for ELISA and indirect methods were compared using the ROC curves statistic of Medcalc.o The sensitivity and specificity determined by picking the point closest to the upper left corner of the ROC curve (minimal false-negative and false-positive results) of ELISA and indirect methods were compared using the McNemar test for paired proportions. Positive test results for FTPI were defined as test results less than or equal to the criterion value corresponding to the point closest to the upper left corner of the ROC curve for ELISA and indirect methods. Positive and negative predictive value (PPV and NPV, respectively) were calculated based on the following equations:
where Se is the sensitivity, Sp is the specificity, and Pr is the prevalence.
A value of P < .05 was considered statistically significant. The statistical analyses were performed with Medcalc and SPSS.p
Serum IgG concentrations determined by SRID1 were lower than 500 mg/dL in 14 samples (serious FTPI, 12%), 500–1,000 mg/dL in 27 samples (partial FTPI, 24%), and higher than 1,000 mg/dL in 74 samples.
The agreement between SRID1 and SRID2 was 97% (Fig 1A). The ICC between SRID1 and SRDI2 was 0.93 (95% confidence interval [CI], 0.91–0.95). The fixed bias (SRID1 − SRID2) between SRID1 and SRID2 was − 134 ± 251 mg/dL, which was significantly different from 0 (P < .0001). There was proportional bias, as indicated by the linear regression curve (dependent variable: differences of the IgG concentrations measured with SRID1 and SRID2; independent variable: mean IgG concentrations of SRID1 and SRID2, Fig 1B). On increasing the mean values of IgG concentrations, the differences tend to be enlarged, particularly toward negative values, reflecting higher IgG concentrations in SRID2 than in SRID1 with increasing IgG concentrations.
The bias between SRID1 and SRID2 was significantly smaller at serum IgG concentrations of <500 mg/dL than at concentrations of 500–1,000 mg/dL and >1,000 mg/dL (P < .037, Table 1). The SDs of the mean biases and hence the limits of agreement were increased with increasing IgG concentrations.
|IgG (mg/dL)||Mean Bias (± SD)||Limits of Agreements|
|SRID1 − SRID2||SRID1 − ELISA||SRID1 − SRID2||SRID1 − ELISA|
|<500||− 7 ± 101a||92 ± 71||− 204 to 191||− 48 to 233|
|500–1,000||− 108 ± 176b||143 ± 141||− 452 to 237||− 133 to 419|
|>1,000||− 167 ± 284b||148 ± 446||− 725 to 391||− 726 to 1,021|
The agreement between SRID1 and ELISA was 94% (Fig 2A). The ICC between SRID1 and ELISA was 0.85 (95% CI, 0.80–0.90). The fixed bias between SRID1 and ELISA (SRID1 − ELISA) was 140 ± 364 mg/dL, which was significantly different from 0 (P=.02). There was proportional bias between SRID1 and ELISA, as indicated by the linear regression curve in Figure 2B. Even though the mean biases between SRID1 and ELISA at serum IgG concentrations of <500, 500–1,000, and >1,000 mg/dL did not show significant differences (Table 1), the random fluctuations around this mean value seemed to increase at serum IgG concentrations >1,000 mg/dL, particularly >2,000 mg/dL (Fig 2B). However, the overall trend of mean biases, SDs of mean biases, and the limits of agreement was to increase with increasing IgG concentrations (Table 1).
The AUC of ELISA was highest at the cut-off value of 500 mg/dL, even though there was no statistically significant difference between ELISA and the indirect methods except MGC (Table 2). The sensitivity of ELISA was 100% at the cut-off value of 500 mg/dL; however, the only indirect method that showed 100% sensitivity at the cut-off value of 500 mg/dL was SSP. MGC showed the worst diagnostic performances in terms of the AUC, sensitivity, and specificity.
|Assay||Criteriona||AUC (95% CI)||Sensitivityb (95% CI)||Specificity (95% CI)|
|SSP||≤ 0||0.95 (0.89–0.98)a||100 (77–100)||90 (82–95)a|
|ZST||≤ 2||0.95 (0.89–0.98)a||93 (66–99)||89 (81–94)ab|
|MGC||≤ 2||0.86 (0.79–0.92)b||86 (57–98)||79 (70–86)b|
|TP||≤ 5.4||0.93 (0.87–0.97)ab||86 (57–98)||89 (81–94)ab|
|ELISA||≤ 459.42||0.98 (0.93–1.00)a||100 (77–100)||92 (85–96)a|
The AUC and sensitivity of ELISA at the cut-off value of 1,000 mg/dL were significantly higher than those of indirect methods (P < .004 Table 3). However, the specificity of ELISA at the cut-off value of 1,000 mg/dL was significantly lower than that of SSP (P=.031), similar to those of ZST and MGC, and significantly higher than that of TP (P < .0001).
|Assay||Criteriona||AUC (95% CI)||Sensitivity (95% CI)||Specificity (95% CI)|
|SSP||≤ 1||0.88 (0.80–0.93)bc||76 (60–88)b||99 (93–100)a|
|ZST||≤ 3||0.92 (0.85–0.96)b||68 (52–82)b||97 (91–100)ab|
|MGC||≤ 2||0.85 (0.78–0.91)bc||73 (57–86)b||96 (89–99)ab|
|TP||≤ 5.8||0.83 (0.75–0.90)c||73 (57–86)b||78 (67–87)c|
|ELISA||≤ 999.86||0.98 (0.93–1.00)a||98 (87–100)a||91 (82–96)b|
The relationships between the positive and negative predictive values and the prevalence of FTPI in ELISA, SSP, ZST, MGC, and TP at the cut-off values of 500 and 1,000 mg/dL are found in Figures 3 and 4, respectively. The prevalence of FTPI was known to be 19–42% in calves.5,10,16,17 For the cut-off value of 500 mg/dL, ELISA showed the highest PPV (75–90%) and NPV (100%) at 19–42% prevalence of FTPI. For the cut-off value of 1,000 mg/dL, ELISA showed the highest NPV (99%) at 19–42% prevalence of FTPI, but the PPV (71–88%) of ELISA at 19–42% prevalence of FTPI was lower than those of indirect methods except for TP. The NPVs and PPVs of ELISA at all prevalences of FTPI showed the same features as those at 19–42% prevalence.
Sample volumes, analysis times, and costs of indirect and direct assays for the detection of FTPI in this study are shown in Table 4. ELISA showed better cost-effectiveness than SRID1 and SRID2 in all aspects but the cost per kit.
|Assay||Sample Volume (mL)||Time (hour)||Tests per Kit or Reagent (packing unit)||Cost per Kit or Reagent (USD)||Cost per Test (USD)|
|SSP||0.3||1||115 (500 g)||65.15||0.57|
|ZST||0.5||1||11,111 (100 g)||119.95||0.01|
|MGC||1.5||1||444 (10 mL)||85.83||0.19|
|SRID1||0.005||18–24||48 (1 kit)||315.40||6.57a|
|SRID2||0.003||18||60 (1 kit)||423.15||7.05a|
|ELISA||<0.001||4||1,000 (1 kit)||674.54||0.67a|
For the direct measurement of serum IgG concentration to diagnose FTPI in dairy calves, ELISA showed good agreement with SRID1 and good diagnostic performances. However, fixed and proportional biases between ELISA and SRID1 were observed. The agreements and biases between measurements of SRID1 and SRID2 showed similar patterns, which are in accordance with a previous study in foals,18 and are considered to be attributed to the nonequal reference sera of 2 commercially available bovine SRID kits in this study, as pointed out by Hutchison et al.7 In addition, the differences in reference-standard concentration increments and the high-concentration reference standards provided in the SRID kits may contribute to the gradual increase in bias, as pointed out by Davis and Giguère.18 For SRID1, the reference standard has 3 concentrations (620, 2,500, and 5,000 mg/dL), whereas the reference standard step for SRID2 has 4 concentrations (400, 800, 1,600, and 3,200 mg/dL).
The fixed and proportional biases between SRID1 and ELISA may be caused by sample dilution and serial dilutions of standards during analysis of ELISA. Regardless of the sources of these biases, the magnitude of fixed biases between SRID1 and ELISA was similar to that between 2 SRID assays. The magnitude of proportional biases, assessed by visual inspection of Figures 1B and 2B, was found to be similar to each other. In terms of bias, the potential assay error of ELISA may be nearly equivalent to that attributed to using different kinds of SRID kits, the reference sera of which differ from kit to kit. Therefore, this potential assay error of ELISA may be clinically acceptable. The presence of proportional biases observed not only for SRID1 versus SRID2 but also for SRID1 versus ELISA may justify the use of log-transformed IgG concentrations for the Bland-Altman method in this study.11,13
As can be seen in Figure 2B, the dispersion of the differences between SRID1 and ELISA seemed to be wider at higher concentration of IgG (>1,000 mg/dL, and particularly 2,000 mg/dL) than that between SRID1 and SRID2. This may account for the failure to show a significant difference between SRID1 and ELISA over the IgG concentration (Table 1), even though proportional bias was identified by linear regression. In turn, the wider dispersion of the difference between SRID1 and ELISA at a higher concentration of IgG (>1,000 mg/dL) may reduce the specificity of ELISA for the cut-off value of 1,000 mg/dL. For this reason, the PPV of ELISA was lower than most indirect methods in this study, when the cut-off IgG concentration is set to 1,000 mg/dL. A test with high sensitivity and hence with high NPV is desirable for screening FTPI. However, a test with a high specificity and hence a high PPV is needed to confirm FTPI.18 From these points of view, ELISA may be appropriate for screening FTPI, regardless of the cut-off values chosen, and may be used as a confirmatory test only when the cut-off value is set to 500 mg/dL, in other words, for serious FTPI. However, ELISA may not be appropriate as a confirmatory test for partial FTPI because specificity and PPV at all prevalences were lower than those of indirect methods at the cut-off value of 1,000 mg/dL.
The higher AUC, sensitivity, and specificity of ELISA, particularly at the cut-off value of 500 mg/dL, may be because of the ability of ELISA to detect very low concentrations of target such as protein and peptides.8,9
Although the bovine IgG quantitative ELISA protocol consists of many steps, including coating with capture antibody, blocking, conjugation of HRP detection antibody, and enzyme substrate reaction, the analysis time is 4.5 times less than that of SRID. As mentioned previously, the large number of tests per plate and the low cost per test make ELISA suitable for analyzing serum IgG concentration in herds rather than an individual calf. In addition to the serial dilution of standards and samples, many analytical steps of ELISA may be cumbersome for analyzing a small number of samples.
In conclusion, ELISA exhibited high diagnostic performance at all cut-off values, and good agreement with SRID. Clinically, ELISA can be used as a screening and confirmatory test for serious FTPI (<500 mg IgG/dL) and as a screening test for partial FTPI (500–1,000 mg IgG/dL) in herds of dairy calves. SRID should be used for the confirmatory diagnosis for the latter situation. However, for the detection of FTPI in an individual dairy calf, screening with indirect methods and confirmatory diagnosis with SRID may be more appropriate.
The authors thank Hyun-Je Ha, DVM (Head Veterinarian, Korea Large Animal Clinic, Ansung City, Korea), Dong-Hee Lee, DVM, MS, PhD (Head Veterinarian, Bakdoosan Large Animal Clinic, Pocheon City, Korea), Jae-Kyung Kim, DVM, MS (Head Veterinarian, Paulus Large Animal Clinic, Kanghwa City, Korea), and Kyung-Moo Lee, DVM (Head Veterinarian, Sung-Sin Large Animal Clinic, Namyangju City, Korea), for sampling blood and recording clinical history. The authors also thank Do-Yang Park, BS (Researcher, Asan Institute for Life Science, Seoul, Korea), for measurements of serum IgG concentrations.
aSST tube, BD, Franklin Lakes, NJ
bSodium sulfite, Sigma-Aldrich, St Louis, MO
dZinc sulfate, Sigma-Aldrich
eClinical Refractometer, NOW, Tokyo, Japan
fBovine IgG VET-RID Kit, Bethyl Laboratories, Montgomery, TX
gImmunocheck, VMRD, Pullman, WA
hVersaDoc Multi Imaging Analyzer System, Bio-Rad, Hercules, CA
iMaxisorp, Nalgene Nunc International, Rochester, NY
jSheep anti bovine IgG-affinity purified, Bethyl Laboratories
kBovine reference serum, Bethyl Laboratories
lSheep anti bovine IgG-HRP conjugate, Bethyl Laboratories
mStable Stop ELISA TMB peroxidase substrate, Kirkegaard & Perry Laboratories, Gaithersburg, MD
nAutomated plate reader, Spectramax 340PC, MDC, Sunnyvale, CA
oMedcalc, version 18.104.22.168, Medcalc Software, Mariakerke, Belgium
pSPSS, version 9.0, SPSS Inc, Chicago, IL
- 1Diseases of Dairy Cattle. Baltimore, MD: Williams & Wilkins; 1995:155–223., , .
- 12Statistical Principles in Experimental Design, 2nd ed. New York: McGraw-Hill; 1971:283pp..
- 15Statistical Methods in Diagnostic Medicine, 1st ed. New York: John Wiley & Sons; 2002:15–164., , .