The Comparison of 5 Diagnostic Tests for Diagnosis of Johne’s Disease in Western Canada

Sylvia L. Checkley, DVM*

Cheryl L. Waldner, DVM, PhD*

Greg D. Appleyard, MSc, PhD†

John R. Campbell, DVM, DVSc*

LeeAnn Forsythe, DVM§

Eugene D. Janzen, DVM, MVS‡

*Department of Large Animal Clinical Sciences, Western College of Veterinary Medicine, 52 Campus Drive, University of Saskatchewan, Saskatoon, SK S7N 5B4, Canada

†Department of Veterinary Microbiology, Western College of Veterinary Medicine, 52 Campus Drive, University of Saskatchewan, Saskatoon, SK S7N 5B4, Canada

‡Feedlot Health Management Services, P.O. Bag Service 5, Bay 7, 87 Elizabeth Street, Okotoks, AB, T1S 1A6

§Saskatchewan Agriculture, Food and Rural Revitalization, Regina, Saskatchewan, Canada

KEY WORDS: Johne’s disease, diagnostic test evaluation, test agreement

Abstract

Five diagnostic tests for Johne’s disease were compared, including a commercial enzyme linked immunosorbent assay (ELISA), routine fecal culture, agar gel immunodiffusion (AGID), and 2 different polymerase chain reaction (PCR) assays. Five populations of cattle were used in the evaluation of these tests. The CS-PCR was carried out on washings from fecal culture slants containing Herrold’s Egg Yolk Medium (HEYM) that were cultured for 6 weeks, and the fecal-PCR was applied directly to DNA extracts of feces. The 2 tests with substantial agreement were the CS-PCRand fecal culture (k = 0.79 [95% confidence interval (CI), 0.68–0.90]). Fecal culture is the existing “gold standard” test for Johne’s disease, although its sensitivity is low. The sensitivity (72.1% [95% CI, 58.7–85.5]) and specificity (98.9% [95% CI, 97.6–100.0]) of the CS-PCRwere estimated using fecal culture as the gold standard. AGID consistently rated only poor or fair in agreement with all the other tests. Agreement among the other tests as measured by kappa varied among different populations examined during the study, partly as a result of changes in prevalence across these groups. Therefore, before using and interpreting diagnostic tests, the population of interest must be characterized.

INTRODUCTION

Johne’s disease is caused by Mycobacterium avium subspecies paratuberculosis (MAP). This bacterium is thought to infect neonatal calves in most cases during the first 30 days of life. Infected animals move from a silent phase with no clinical signs and no detectable shedding of the organism to a subclinical phase without clinical signs but intermittent shedding of the organism in the feces. Most infected animals eventually exhibit symptoms of severe weight loss and diarrhea, but remain bright and alert with a good appetite. Symptoms of more advanced clinical disease include ventral edema and “water hose” diarrhea.1 Economic losses can be substantial when a beef or dairy herd becomes infected with MAP. The most apparent losses are from symptomatic animals that are culled. However, premature culling of positive asymptomatic animals results in loss of potential income from reproduction, show, or sales. Herds labeled as MAP-infected can be restricted from both domestic and export sales. Milk production losses in asymptomatic dairy cattle have also been documented.2

The screening tests available for MAP have only moderate sensitivity; however, most of these tests do have high, although still imperfect, specificity. The specificity of fecal culture approaches 100%.3 Antibodies to the MAP antigen can develop during the subclinical and clinical stages of disease, but some animals never have a strong antibody response. The specificities of the enzyme linked immunosorbent assay (ELISA) and AGID are estimated between 97% and 100%.3,4 Some false-positives will occur, even with high-test specificity, and this should be considered when testing low prevalence herds or when using the test on asymptomatic animals. The sensitivities of the ELISA and AGID tests are low, in the range of 15% to 60%.3,4 The resulting high probability of false-negative results for individual animals must be considered when designing control programs. The sensitivity of these diagnostic tests also varies greatly with the disease prevalence and severity within a herd. This variation is primarily the result of biologic factors within the host such as the stage of disease and immune status of the host.5 The variation in sensitivity is also related to the fact that these are not truly binary tests, but rather test protocols that dichotomize results into positive and negative subsets based on variable cut points.5,6

A gold standard test definitively establishes the health status of an animal with respect to the disease in question. In theory it should have 100% sensitivity and specificity. Ideally new tests should be compared with a gold standard to determine their sensitivity, specificity and, therefore, validity. A true gold standard or definitive diagnosis often is not available as a result of practicality, expense, or the nature of the disease. For example, subclinical animals with MAP infection might test negative in all serologic tests, fecal culture, and even biopsy. Definitive diagnosis might only be attainable on postmortem examination when the disease is advanced far enough along to diagnose with histology. However, when no alternatives are available, new methods must be compared with an imperfect gold standard. Kappa is a measure of agreement beyond chance that can be used to compare diagnostic tests without designating one test as a gold standard.7

Laboratories and veterinarians have sought alternate diagnostic tests for MAP because of limitations in the sensitivity of fecal culture as well as the long time required for culture results. Polymerase chain reaction (PCR) assays work on the principle of enzymatic amplification of DNA. The assay uses extracts of DNA from the test sample, in this case the washings from a fecal culture slant. The test amplifies a region of the insertion sequence (transposon) in MAP called IS900, which is unique to MAP and a few strains of M. avium subspecies avium.8,9 The reported advantages of PCR technology as a diagnostic test include a low minimum detection limit for the organism of interest, rapid results when compared with traditional culture, and high versatility in sample type and pathogen type.

The objectives of this study was to compare the results of 5 different diagnostic tests for Johne’s disease: a commercial ELISA, standard fecal culture, AGID, and 2 different PCR tests. The second objective was to estimate test performance of an optimized PCR performed on HEYM fecal culture slants relative to traditional HEYM fecal culture.

Materials and Methods

Sample Handling

Blood and fecal samples were collected in western Canada over a 3-year period between 1998 and 2001. Samples came from 5 different populations of cattle. All cattle sampled were over 2 years of age. All samples used were chosen based on their sample to positive (S/P) ratio of optical densities from the Johne’s ELISA (HerdChek* M.pt., IDEXX Laboratories, Westbrook, ME. Population 1 included samples collected randomly from 29 herds involved in a large independent multidisease prevalence survey in dairy cattle that included testing for Johne’s disease. Owners were chosen randomly to participate in the disease study. All of these samples used were chosen based on initial ELISA results. Further testing was carried out if the ELISA S/P ratio was greater than 0.10. Population 2, the population of “negative” animals in this study, included samples from animals with S/P ratios less than 0.01 in the prevalence survey described previously. These animals originated from 6 herds in which the ELISA S/P ratios for the 30 cows tested were all less than 0.2 with no positive cases. This gives a fairly high likelihood that the herds were negative for the disease.10 Populations 3 and 4 included samples from known MAP-infected beef herds. Population 5 included samples from a known MAP-infected dairy herd. In the individual herds (populations 3, 4, and 5), if the ELISA S/P ratio was suspicious or positive, then further tests were performed. Prevalence of disease in these 3 herds had been estimated by previous herd serologic studies and fecal cultures. The results from population 4 were used in overall calculations, but the individual population results were not reported because of the small sample size. In all 5 populations, animals with complete results for at least 4 of the 5 diagnostic test results were chosen for inclusion in this study. The 5 populations used are described further, including each population’s fecal culture prevalence of Johne’s disease, in Table 1.

Blood samples were centrifuged and the serum was separated within 24 hours of collection. Fecal samples were stored at –4˚C for 24 to 48 hours and then cultured. Further diagnostics were either done immediately or after storage at –20 or –70˚C.

Commercial ELISA

A solid-phase, enzyme-linked immunosorbent assay was carried out on 581 samples per the manufacturer’s instructions (HerdChek* M.pt., IDEXX Laboratories, Westbrook, MA). This is an ELISA with a rapid absorption step that removes cross-reacting antibodies with M. phlei. Serum samples with a S/P ratio greater than or equal to 0.25 were considered positive and serum samples with S/P ratios greater than or equal to 0.1 but less than 0.25 were considered suspicious. Samples with S/P ratios less than 0.1 were considered negative. The manufacturer supplied positive and negative controls.

AGID

The agar gel immunodiffusion test was carried out on 570 samples. One central and 6 surrounding wells were made in an agar plate with a punch and template. Paratuberculosis Protoplasmic Antigen (30 mL; Allied Monitor, The Paratuberculosis Laboratory, Fayette, MO) was placed in the 3-mm center well. Control sera (50 mL) were placed in 2 of the 4-mm outer wells directly opposite each other. Test antisera (50 mL) were placed in 2 of the outer wells, directly opposite each other, and with each sera being directly beside the control antisera. The plate is placed in a sealed, shallow plastic container that has dampened paper towel in the bottom to control humidity. Test samples were incubated at reverse transcriptase for 48 hours, with readings taken at both 24 and 48 hours. The manufacturer supplied positive and negative controls.

Fecal Culture

All 579 fecal samples were cultured for MAP. The procedure began with mixing, sedimentation, and decontamination of 2 g of fecal material with hexadecylpyridinium chloride in brain heart infusion broth, followed by centrifugation and resuspension in an antibiotic broth for final decontamination. Inoculation of 3 slants (2 with mycobactin and 1 without) of HEYM and incubation in a slanted position at 37˚C followed. Caps on the slants were tightened after 1 to 2 weeks and slants were read starting at 6 weeks for culture growth. Suspicious colonies were evaluated for mycobactin dependence along with morphology and acid-fast staining. This procedure was U.S. Department of Agriculture-approved and recommended by Fyock and Whitlock from the University of Pennsylvania School of Veterinary Medicine, New Bolton Center in 1999.

CS-PCR

The PCR, carried out on 315 samples, was adopted from previously published protocols.9 Briefly, DNA was extracted from HEYM slants on which feces had been cultured for 6 weeks. The slants were rinsed with 1 mL lysis buffer (Tris, EDTA, 10% SDS) and harvested cells killed by heating to 80˚C for 20 minutes. Cells were lysed through sequential treatment with 0.2 mg/mL Proteinase K for 2 hours at 65˚C. Two phenol–chloroform (50:50 v/v) extractions were performed and nucleic acid was recovered through cold 95% ethanol precipitation. The DNA pellet was then dried for 5 to 10 minutes under a vacuum and resuspended in 30-m L sterile water. Two microliters of resuspended DNA were added to PCR tubes containing 48 mL of Master Mix, which consisted of 36 mL sterile ultrapure water, 5 mL 10x PCR buffer, 3 mL MgCl₂, 0.3 mL dNTP’s (25 mM), 50 rMol of each primer (P36, P11), and 5U (Taq) polymerase. For amplification, the samples were treated at 92˚C for 4 minutes followed by 35 cycles at 94˚C for 45 s denaturation, annealing at 58˚C for 60 s, and extension at 72˚C for 90 s in a thermal cycler (PTC 200, MJ Research, Inc., Reno, NV). Positive and negative controls were used.

Fecal PCR

This PCR was performed on 153 fecal samples. DNA from the fecal samples was extracted using the QIAamp DNA Stool Mini Kit and the corresponding protocol for the isolation of DNA from stool for pathogen detection (Qiagen Inc., Mississauga, Ontario). The PCR was performed as previously described. The manufacturer supplied positive and negative controls.

Statistical Analysis

Descriptive statistics were generated with a commercial software program (SPSS for Windows 11.0, SPSS Inc., Chicago, IL). Agreement among tests was measured by calculating the kappa coefficient (WinEpiscope 2.0, Epidecon, Wageningen University, The Netherlands). An analysis with latent class modelling was also attempted (TAGS V2.0, Agence Francaise de Securite Sanitaire des Aliments, France). Sensitivity, specificity, and other diagnostic parameters were calculated.11

RESULTS

Descriptive statistics of the test results were examined. Summaries of the test results observed were stratified and tabulated. Agreement between test pairs using the kappa statistic for all populations of cattle was highly variable. The prevalence and total number of animals in the comparison was included to facilitate the interpretation of kappa coefficients. Agreement was assessed across all groups, including the entire population (Table 2). The ELISA has a large number of suspicious results; therefore, agreement was calculated for both positive and negative interpretations of these suspicious results. This approach is consistent with interpretation of these results in veterinary practice in which animals with ELISA-suspicious results are noted and often undergo further testing. The CS-PCR had one suspicious result that was interpreted as negative.

Sensitivity, specificity, and predictive values of the ELISA, AGID, CS-PCR, and fecal-PCR relative to fecal culture are presented in Table 3. Latent class models were attempted to examine the validity of the CS-PCR_.The assumptions required for this maximum likelihood method were violated.

Assessing new diagnostic tests in the absence of a gold standard is common but not ideal. New tests that are compared with an imperfect gold standard will have bias in the error rates of the new test as a result of the imperfection of the gold standard.12,13 This is especially true for tests with a greater detection limit than the gold standard. Therefore, techniques for evaluation of diagnostic tests using maximum likelihood methods with a latent class model have been developed and refined.14,15 A latent class is one in which the true disease status is not known. These techniques can estimate sensitivity and specificity for diagnostic tests as well as disease prevalence estimates in unknown populations. This method provides better estimates for test performance characteristics by avoiding the bias resulting from the imperfect gold standard. This technique has been used previously in veterinary medicine for Johne’s disease diagnostic tests2 and is applicable to the type of data in this analysis, but 2 major assumptions must be met. First, conditional independence must exist between the tests used, and second, the test accuracy must remain constant across all subpopulations tested. This type of analysis could not be used in this study because both the assumptions were violated: ELISA, AGID, and the 2 PCRs were not independent tests and test accuracy varies with prevalence.14,16

Cohen’s kappa coefficient is another method of evaluating agreement between 2 tests and can be used when a gold standard test is not available. Kappa is an expression of the agreement found between 2 tests beyond what would be expected as a result of chance. The kappa statistic varies between 0 and 1, in which 0 is chance agreement and 1 is perfect agreement. Kappa does not reveal which test is better, but rather it indicates how often 2 tests have the same results within the same subject. Two tests with relatively high sensitivities and specificities might have a low kappa because they detect different populations of diseased animals. Kappa has a number of limitations.7,17,18 The value of kappa depends on the proportion of individuals in each category and on the number of categories used. The criteria for judging kappa vary and are not completely objective. The discussion of agreement in this study is based on interpretation of values of the kappa statistic as described previously in the literature in which k<0.00 is “poor,” 0<k<0.02 is “slight,” 0.21<k<0.40 is “fair,” 0.41<k<0.60 is “moderate,” 0.61<k<0.80 is “substantial,” and 0.81<k<1.00 is “almost perfect” agreement.19

Agreement among all combinations of tests was calculated in each population and for all populations combined. The only test combination with “substantial” agreement was the CS-PCR compared with fecal culture in all populations combined and in the moderate prevalence populations. In the high prevalence population, agreement among ELISA, CS-PCR, and fecal culture was “perfect” or “almost perfect.” The agreement between the AGID and fecal-PCR in the high prevalence population was “substantial,” probably at least partly related to the small sample size for this comparison. AGID and fecal-PCR did not agree well with fecal culture however. Comparisons of agreement involving AGID consistently rated low. In the moderate prevalence populations, agreement in all its permutations was “poor” or “fair” except CS-PCR, which had an “almost perfect” agreement with fecal culture.

DISCUSSION

Poorer agreement was expected for the low prevalence population because kappa varies with prevalence within subpopulations of animals conditional on the distribution of covariates within the population.5 With all other variables held constant, kappa will be lower in a low prevalence population and higher in a higher prevalence population. In the disease-free population, agreement was difficult to determine as a result of little variation in the distribution of results. Clearly kappa is not an appropriate measure of agreement between diagnostic tests in a disease-free population alone. Different kappa values for different populations were expected; therefore, before using and interpreting diagnostic tests, the population of interest must be characterized. Prevalence stratum-specific calculations were used to increase the generalizability of the results.5

Fecal culture is reported to have 100% specificity but a much lower sensitivity.3 An approximation of sensitivity and specificity for CS-PCR and the other tests can be made using fecal culture as a gold standard (Table 3). CS-PCR sensitivity for predicting fecal culture results in this trial was approximately 72.1% (95% confidence interval [CI], 58.7–85.5), and the specificity was 98.9% (95% CI, 97.6–100.0). The positive predictive value was 91.4% and the negative predictive value was 96.0% in a population with a disease prevalence of approximately 8%. These values would indicate a highly useful test in herds with prevalence above that expected in the general population. The specificity estimate might be misleading because the amount of MAP in these samples might be below the detection limit for fecal culture but not below the detection limit for this PCR. When the sensitivity of the purported gold standard is less than one, like in this case, the specificity and positive predictive value for the new test are underestimated and the negative predictive value is overestimated.12 Therefore, CS-PCR might perform even better than we can estimate at this point.

The direction of bias in this calculation is related to the sensitivity and specificity of the gold standard.13 The CS- PCR had a somewhat lower sensitivity than another real-time PCR assay described in the literature, performed directly on the feces with 1 of 2 different extraction techniques.20 The sensitivity for the real-time PCR was reported between 93.0% and 98.0% compared with fecal culture, and the specificity was between 95% and 100%. That PCR was tested on 63 samples from populations of unknown prevalence, which might explain why different results differed from those in this study. The fecal-PCR in this study also had a low sensitivity and lower specificity than reported in the literature.20 The sample size for its evaluation in this study was lower than for the other tests; further analysis is indicated. The AGID in this study had a low sensitivity and high specificity similar to other reported values when compared with fecal culture.3 The ELISA appeared to have a much higher sensitivity and lower specificity in this trial than those reported in the literature.3,4 This is the result of the sampling method used in this study, in which samples were chosen by ELISA S/P values.

The CS-PCR assay performed on the fecal culture slant after 6 weeks incubation was a rapid and accurate method of Johne’s disease diagnosis. The sensitivity and specificity of the CS-PCR, compared with traditional fecal culture, were higher than for the serologic tests. The faster processing time of the CS-PCR compared with traditional HEYM fecal culture, along with the higher sensitivity and direct identification of MAP DNA in the feces, make this a highly desirable test for Johne’s disease in cattle in western Canada. This advantage might overcome the higher cost of this test. In bovine practice, faster test results lead to faster culling of animals that are shedding MAP bacteria in the feces, therefore reducing transmission of the disease within the herd.

ACKNOWLEDGMENTS

Many people were involved in obtaining samples and laboratory analysis of these data. The authors specifically thank Dr. Maria Spinato at Prairie Diagnostic Service in Regina, Saskatchewan. The Saskatchewan Agriculture Development Fund supplied funding in part for this project.

REFERENCES

1. Whitlock RH, Buergelt C: Preclinical and clinical manifestations of paratuberculosis (including pathology). Vet Clin North Am Food Anim Pract 12:345–356, 1996.

2. Spangler E, Bech-Nielsen S, Heider LE: Diagnostic performance of two serologic tests and fecal culture for subclinical paratuberculosis, and associations with production. Prev Vet Med 13:185–195, 1992.

3. Sockett DC, Conrad TA, Thomas CB, Collins MT: Evaluation of four serological tests for bovine paratuberculosis. J Clin Microbiol 30:1134–1139, 1992.

4. Sweeney RW, Whitlock FH, Buckley CL, Spencer PA: Evaluation of a commercial enzyme-linked immunosorbent assay for the diagnosis of paratuberculosis in dairy cattle. J Vet Diagn Invest 7:488–493, 1995.

5. Greiner M, Gardner IA: Epidemiologic issues in the validation of veterinary diagnostic tests. Prev Vet Med 45:3–22, 2000.

6. Brenner H, Gefeller O: Variation of sensitivity, specificity, likelihood ratios and predictive values with disease prevalence. Stat Med 16:981–991, 1997.

7. Maclure M, Willett WC: Misinterpretation and misuse of the kappa statistic. Am J Epidemiol 126:161–169, 1987.

8. Cousins DV, Whittington R, Marsh I, et al: Mycobacteria distinct from Mycobacterium avium subsp. paratuberculosis isolated from the faeces of ruminants possess IS900-like sequences detectable by IS900 polymerase chain reaction: implications for diagnosis. Mol Cell Probes 14:431–442, 1999.

9. Moss MT, Green EP, Tizard ML, Malik ZP, Hermon-Taylor J: Specific detection of Mycobacterium paratuberculosis by DNA hybridization with a fragment of the insertion element IS900. Gut 32:395–398, 1991.

10. Collins MT: Interpretation of a commercial bovine paratuberculosis enzyme-linked immunosorbent assay by using likelihood ratios. Clin Diagn Lab Immunol 9:1367–1371, 2002.

11. Sackett DL, Haynes RB, Guyatt GH, Tugwell P: Clinical Epidemiology. A Basic Science for Clinical Medicine, 2nd ed. Boston: Little, Brown and Company; 1991:69–152.

12. Staquet M, Rozencweig M, Lee YJ, Muggia FM: Methodology for the assessment of new dichotomous diagnostic tests. J Chron Dis 34:599–610, 1981.

13. Valenstein PN: Evaluating diagnostic tests with imperfect standards. Am J Clin Pathol 93:252–258, 1991.

14. Enoe C, Georgiadis MP, Johnson WO: Estimation of sensitivity and specificity of diagnostic tests and disease prevalence when the true disease state is unknown. Prev Vet Med 45:61–81, 2000.

15. Pouillot R, Gerbier G, Gardner IA: ‘TAGS’ a program for the evaluation of test accuracy in the absence of a gold standard. Prev Vet Med 53:67–81, 2002.

16. Vacek P: The effect of conditional dependence on the evaluation of diagnostic tests. Biometrics 41:959–968, 1985.

17. Chmura Kraemer H, Bloch DA: Kappa coefficients in epidemiology: and appraisal of a reappraisal. J Clin Epidemiol 41:959–968, 1988.

18. Thompson WD, Walter SD: A reappraisal of the kappa coefficient. J Clin Epidemiol 41:949–958, 1988.

19. Landis JR, Koch GG: The measurement of observer agreement for categorical data. Biometrics 33:159–174, 1977.

20. Christopher-Hennings J, Dammen MA, Weeks SR, et al: Comparison of two DNA extractions and nested PCR, real-time PCR, a new commercial PCR assay, and bacterial culture for detection of Mycobacterium avium subsp. paratuberculosis in bovine feces. J Vet Diagn Invest 15:87–93, 2003.

Table 1. Description of Populations Used

Fecal Culture

Prevalence of Population

Johne’s Disease (%) No. Description

Population 1 2.5 119 Moderate

Population 2 0 64 Negative

Population 3 12.5 40 High

Population 4 6 10 Moderate

Population 5 6 349 Moderate

Overall ? 582 Combined

Table 2. Agreement of Diagnostic Tests in Populations of Animals With Varying Prevalence of Disease

Population No. positive on tests Kappa Interpretation

Comparison (Prevalence) Notes n (No. negative on tests) (95% confidence interval) of Kappa

Culture & ELISA Overall @ ELISA susp as +ve 578 37 (373) 0.205 (0.148–0.262) Fair

@ ELISA susp as -ve 34 (442) 0.322 (0.253–0.391)

Culture & ELISA 1 @ ELISA susp as +ve 118 10 (0) 0 (na) Slight

@ ELISA susp as -ve 9 (42) 0.073 (-0.006–0.152)

Culture & ELISA 2 – 62 62 (0) 1 (na) Perfect

Culture & ELISA 3 @ ELISA susp as +ve 39 20 (18) 1 (0.686–1.314) Almost perfect

@ELISA susp as –ve 20 (19) 0.949 (0.635–1.262)

Culture & ELISA 5 @ ELISA susp as +ve 349 3 (293) 0.053 (–0.026–0.132) Slight

@ ELISA susp as -ve 3 (313) 0.114 (0.020–0.207)

Culture & AGID Overall – 567 9 (516) 0.266 (0.190–0.342) Fair

Culture & AGID 1 – 118 1 (106) 0.119 (–0.030–0.269) Slight

Culture & AGID 2 – 61 61 (0) 1 (na) Perfect

Culture & AGID 3 – 39 7 (19) 0.344 (0.107–0.581) Fair

Culture & AGID 5 – 349 1 (330) 0.074 (-0.031–0.178) Slight

Culture & CS-PCR Overall – 311 31 (266) 0.790 (0.680–0.901) Substantial

Culture & CS-PCR 1 – 110 0 (99) -0.031 (-0.172–0.110) Poor

Culture & CS-PCR 2 – 62 0 (61) 0 (na) Poor

Culture & CS-PCR 3 – 19 19 (0) 1 (na) Perfect

Culture & CS-PCR 5 (@ PCR susp as –vet) 111 10 (100) 0.947 (0.762–1.133) Almost perfect

Culture & Fecal-PCR Overall – 151 7 (121) 0.307 (0.165–0.450) Fair

Culture & Fecal-PCR 1 (@ PCRf susp as –ve) 118 0 (104) -0.051 (-0.212–0.110) Poor

Culture & Fecal-PCR 3 – 33 7 (17) 0.445 (0.161–0.729) Moderate

ELISA & AGID Overall @ ELISA susp as +ve 570 19 (383) 0.132 (0.091–0.173) Slight

@ ELISA susp as -ve 18 (446) 0.208 (0.156–0.259)

ELISA & AGID 1 @ ELISA susp as +ve 119 3 (1) 0.000 (-0.005–0.006) Slight

@ ELISA susp as -ve 3 (44) 0.030 (-0.014–0.370)

Table 2 continued on next page

Table 2. continued

ELISA & AGID 2 – 63 63 (0) 1 (na) Perfect

ELISA & AGID 3 @ ELISA susp as +ve 39 7 (18) 0.316 (0.087–0.545) Fair

@ELISA susp as –ve 7 (19) 0.344 (0.107–0.581)

ELISA & AGID 5 @ ELISA susp as +ve 349 9 (301) 0.285 (0.211–0.358) Fair

@ ELISA susp as -ve 8 (320) 0.409 (0.321–0.498)

ELISA & CS-PCR Overall @ ELISA susp as +ve 315 26 (146) 0.108 (0.040–0.175) Slight

(@ CS-PCR susp as –ve) @ ELISA susp as -ve 23 (200) 0.204 (0.116–0.292)

ELISA & CS-PCR 1 @ ELISA susp as +ve 111 2 (1) 0.000 (-0.004–0.005) Slight

@ ELISA susp as -ve 0 (37) -0.036 (-0.073–0.000)

ELISA & CS-PCR 2 - 64 0 (64) 0 (na) Poor

ELISA & CS-PCR 3 – 19 19 (0) 1 (na) Perfect

ELISA & CS-PCR 5 @ ELISA susp as +ve 111 3 (82) 0.073 (-0.096–0.241) Slight

(@ CS-PCR susp as –ve) @ ELISA susp as -ve 3 (93) 0.030 (0.004–0.056)

ELISA & Fecal-PCR Overall @ ELISA susp as +ve 152 11 (17) 0.019 (-0.012–0.051) Slight

@ ELISA susp as -ve 8 (58) 0.032 (-0.037–0.101)

ELISA & Fecal-PCR 1 @ ELISA susp as +ve 119 1 (40) -0/042 (-0.091–0.008) Poor

@ ELISA susp as -ve 1 (41) -0.041 (-0.091–0.009)

ELISA & Fecal-PCR 3 @ ELISA susp as +ve 32 7 (15) 0.396 (0.120–0.672) Moderate

@ELISA susp as –ve 7 (16) 0.438 (0.151–0.724)

AGID & CS-PCR Overall (@ PCRs susp as –ve) 304 8 (267) 0.314 (0.213-0.414) Fair

AGID & CS-PCR 1 — 111 0 (106) -0.022 (-0.204–0.160) Poor

AGID & CS-PCR 2 — 63 0 (62) 0 (na) Poor

AGID & CS-PCR 3 — 19 7 (0) 0 (na) Poor

AGID & CS-PCR 5 @ PCR susp as –vet) 111 1 (99) 0.117 (-0.037–0.271) Slight

AGID &Fecal-PCR Overall - 152 5 (136) 0.437 (0.279-0.596) Moderate

AGID &Fecal-PCR 1 - 119 0 (112) -0.030 (-0.207–0.148) Poor

AGID &Fecal-PCR 3 - 32 5 (23) 0.634 (0.288–0.981) Substantial

Fecal-PCR & CS-PCR Overall - 128 7 (106) 0.421 (0.254–0.588) Moderate

Fecal-PCR & CS-PCR 1 - 111 0 (105) -0.025 (-0.199–0.150) Poor

Fecal-PCR & CS-PCR 3 - 16 7 (0) 0 (na) Poor

ELISA, enzyme linked immunosorbent assay; AGID, agar gel immunodiffusion; PCR, polymerase chain reaction.

Table 3. Diagnostic Parameters of the 4 Diagnostic Tests Compared With Fecal Culture

No. Sensitivity Specificity True P PPV NPV

ELISA (with suspicious as positive) 578 82.2 (71.1–93.4)* 70.0 (66.1–73.9)^*7.8 (5.6–10.0)^* 18.8 (13.3–24.2)^*97.9 (96.5–99.3)^*

ELISA (with suspicious as negative) 578 75.6 (63.0–88.1) 82.9 (79.7–86.1) 7.8 (5.6–10.0) 27.2 (19.4–35.0) 97.6 (96.2–99.0)

AGID 567 22.0 (9.3–34.6) 98.1 (97.0–99.3) 7.2 (5.1–9.4) 47.4 (24.9–69.8) 94.2 (92.2–96.1)

PCR_slant (with suspicious as positive) 312 74.4 (61.4–87.5) 98.9 (97.6–100.0) 13.8 (10.0–17.6) 91.4 (82.2 –100.0) 96.0 (93.7–98.3)

PCR_slant (with suspicious as negative) 312 72.1 (58.7–85.5) 98.9 (97.6–100.0) 13.8 (10.0–17.6) 91.2 (81.6–100.0) 95.7 (93.3–98.1)

PCR_feces 151 26.9 (9.9–44.0) 96.8 (93.7–99.9) 17.2 (11.2–23.2) 63.6 (35.2–92.1) 86.4 (80.8–92.1)

*95% confidence interval.
P, prevalence; PPV, positive predictive value; NPV, negative predictive value; ELISA, enzyme linked immunosorbent assay; AGID, agar gel immunodiffusion; PCR, polymerase chain reaction.