• Open Access

Indirect Fluorescent Antibody Test and Surface Antigen ELISAs for Antemortem Diagnosis of Equine Protozoal Myeloencephalitis

Authors


  • Cases were evaluated and samples were collected at New Bolton Center. IFAT was performed commercially at UC Davis School of Veterinary Medicine. SAG2, 4/3 ELISA was performed at Equine Diagnostic Solutions, LLC.
  • This study was presented during the EPM SIG meeting at the 2012 ACVIM Forum, New Orleans, LA

Corresponding author: A.L. Johnson, Department of Clinical Studies – New Bolton Center, University of Pennsylvania School of Veterinary Medicine, 382 W. Street Road, Kennett Square, PA 19348; e-mail: amyjohn@vet.upenn.edu.

Abstract

Background

Recent research suggests that serum : CSF titer ratios could provide the most accurate antemortem diagnosis of equine protozoal myeloencephalitis (EPM) caused by Sarcocystis neurona.

Objectives

The purpose of this study was to assess the utility of two commercially available tests, the indirect fluorescent antibody test (IFAT) and the surface antigen 2, 4/3 ELISA (SAG2, 4/3 ELISA), using archived paired serum and CSF samples.

Animals

Samples were obtained from 4 types of clinical patients. Confirmed positive cases (n = 9 horses; 11 sample sets) had neurologic deficits and postmortem lesions consistent with EPM. Confirmed negative cases (n = 28) had variable clinical signs and postmortem lesions consistent with another disease. Suspected positive cases (n = 6) had neurologic deficits consistent with EPM, marked improvement after treatment, and exclusion of other diseases. Suspected negative cases (n = 14) had variable signs with a strong presumptive diagnosis of another disease.

Methods

For each test, descriptive statistics were calculated using serum results alone, CSF results alone, and a serum : CSF titer ratio.

Results

Overall accuracy was highest for SAG2, 4/3 ELISA titer ratio at 0.97 (95% CI 0.88–0.99) with sensitivity = 0.88 (95% CI 0.66–0.97) and specificity = 1 (95% CI 0.92–1). IFAT CSF and titer ratio results also showed high accuracy at 0.88 (95% CI 0.77–0.94), but lower sensitivity = 0.65 (95% CI 0.41–0.83).

Conclusions and Clinical Importance

Using serum results alone was least accurate for both test types. The more accurate methods, such as the SAG2, 4/3 ELISA serum : CSF titer ratio, should be utilized.

Abbreviations
BBB

blood-brain barrier

CI

confidence interval

CNS

central nervous system

CSF

cerebrospinal fluid

EPM

equine protozoal myeloencephalitis

IFAT

indirect fluorescent antibody test

ROC

receiver operating characteristic

SAG2

surface antigen 2

4/3 ELISA

4/3 enzyme-linked immunosorbent assay

The antemortem diagnosis of equine protozoal myeloencephalitis (EPM) remains a clinical challenge. Most commonly used diagnostic tests are based on detection of antibodies against Sarcocystis neurona, the primary causative agent. Tests can be differentiated based on both methodology and the specific antibodies detected. Although the Western blot was the first available immunologic test, many practitioners are now using quantitative tests, including the indirect fluorescent antibody test (IFAT) and various ELISAs.[1-3]

Laboratory and clinician recommendations vary as to the relative accuracy of the tests and the ideal sample type for submission. Widespread equine exposure to S. neurona with subsequent antibody production in horses without central nervous system (CNS) infection confounds interpretation of serum titers.[4] Normal partitioning of serum antibodies across the blood-brain barrier (BBB) can lead to false positive results when only CSF is tested.[5] Previous recommendations included testing CSF rather than serum,[4] linking serum titers to posttest probabilities of infection,[6] and most recently, evaluating serum and CSF titers concurrently.[5] The recommendation to evaluate paired serum and CSF samples is based on the premise that horses with CNS infection will have intrathecal antibody production resulting in proportionately higher CSF antibody titers than would be expected by normal blood/CSF partitioning.

The objective of this study was to evaluate two of the most commonly utilized diagnostic tests, the indirect fluorescent antibody test (IFAT) and the surface antigen 2, 4/3 ELISA (SAG2, 4/3 ELISA), in the antemortem diagnosis of EPM using naturally infected cases. Test performance was evaluated for each potential type of sample submission: serum alone, CSF alone, and paired serum and CSF.

Materials and Methods

Archived paired serum and CSF samples were utilized for this study. Fifty-nine sets of samples were collected from 57 horses that were presented to the George D. Widener Hospital for Large Animals at New Bolton Center for evaluation over a 2-year period (December 2009–2011). Samples were acquired during the initial evaluation of horses that were presented for neurologic disease; after utilizing portions of these samples for routine testing as indicated by the horses' clinical signs, remaining samples were stored at −80°C until analysis. Serum samples were acquired from neurologically normal horses before euthanasia, and CSF samples were obtained immediately after euthanasia; these samples were also stored at −80°C until analysis.

Samples from 4 different types of cases were used. Confirmed positive cases (n = 9 horses and 11 sample sets) had neurologic deficits as determined by attending clinicians and postmortem lesions consistent with EPM as determined by university pathologists. Criteria for a postmortem diagnosis of EPM included the presence of multifocal or focally extensive lymphocytic, lymphohistiocytic, or lymphoplasmacytic myelitis, encephalitis, or both. Immunohistochemistry, PCR testing, or both for S. neurona were used at the discretion of the pathologist to confirm a diagnosis, but were not performed in every case. The 2 confirmed positive cases with 2 sample sets each had samples collected at the time of initial onset of neurologic signs, before treatment, and again several months later after treatment failure and recrudescence of signs. Suspected positive cases (n = 6) had neurologic deficits consistent with EPM, marked improvement after standard-of-care treatment for EPM, and exclusion of other likely diseases by appropriate diagnostic testing. This testing was determined by the attending clinician based on neuroanatomical diagnosis and likely differential diagnoses, and it included skull or spinal imaging or both (radiography, myelography, nuclear scintigraphy, ultrasonography, endoscopy, magnetic resonance imaging), immunologic testing (for West Nile virus, neosporosis, and neuroborreliosis), and PCR (for equine herpesvirus-1). Not all diagnostic tests were performed for each case. Confirmed negative cases (n = 28) had variable clinical signs and postmortem lesions consistent with another disease. Twenty of the confirmed negative cases had neurologic diagnoses, including cervical vertebral stenotic myelopathy (8), degenerative myelopathy (3), neuroborreliosis (2), idiopathic epilepsy (2), West Nile encephalomyelitis (1), hepatic encephalopathy (1), skull fracture (1), pelvic fracture (1), and spinal infarct (1). Eight of the confirmed negative cases were considered neurologically normal and had diagnoses of abdominal disease (3), thoracic disease (3), bilateral ocular disease (1), and cardiac disease (1). Suspected negative cases (n = 14) had variable clinical signs with a strong presumptive diagnosis of another disease. Diagnoses for suspected negative cases included cervical vertebral stenotic myelopathy (5), neoplasia (2), trauma (2), intoxication (2), idiopathic epilepsy (2), and idiopathic headshaking (1).

Paired samples were submitted for routine diagnostic testing at two laboratories.12 SAG2, 4/3 ELISA results for each sample set reported the serum titer, CSF titer, and serum : CSF titer ratio. IFAT results for each sample set reported the serum titer and CSF titer; the serum : CSF titer ratio was calculated using the reported values. All cases were included. Sample results from confirmed and suspected positive cases were grouped for analysis, as were sample results from confirmed and suspected negative cases. For each test and each sample type (serum, CSF, serum : CSF titer ratio), receiver operating characteristic (ROC) curves were constructed to assess test performance as previously described.[7] The area under each ROC curve (AUC) was calculated using the trapezoidal rule, and the standard error of the AUC and 95% confidence limits were calculated.[8] The differences in the AUCs were compared by calculating z statistics adjusted for correlation because of the use of the same patients for both tests.[9]

Descriptive statistics, including sensitivity, specificity, overall accuracy, and 95% exact binomial confidence intervals, were calculated for each sample type: serum alone, CSF alone, and a serum : CSF titer ratio. Although titers were returned from each laboratory, results were dichotomized into positive and negative results for the purposes of this study and determination of sensitivities and specificities. ROC curve analysis was used to select cut-off titers that optimized sensitivity and specificity for each test as previously described.[7]

Finally, test agreement for each sample type was assessed by a kappa statistic.

Results

Table 1 displays the AUCs and 95% exact binomial confidence intervals for the ROC curves for each test and sample type. For the IFAT, although the highest degree of discrimination was seen using the CSF titer (AUC 0.82; 95% confidence limits, 0.67–0.96), there was no statistically significant difference between results for the 3 sample types. For the SAG2, 4/3 ELISA, there were statistically significant differences between the 3 sample types, with the serum : CSF titer ratio (AUC 0.98; 95% confidence limits 0.95–1.0) significantly more discriminatory than the CSF titer, and the CSF titer significantly more discriminatory than the serum titer. When similar sample types were compared between tests, there was no significant difference between the IFAT and SAG2, 4/3 ELISA for serum or CSF. However, the SAG2, 4/3 ELISA serum : CSF titer ratio was significantly more discriminatory than the IFAT serum : CSF titer ratio or IFAT CSF titer.

Table 1. Comparison of ROC curves
TestAUC95% CI
IFAT on serum0.660.50–0.82
IFAT on CSF0.820.67–0.96
IFAT serum : CSF titer ratio0.810.66–0.96
SAG2, 4/3 ELISA on serum0.540.38–0.69
SAG2, 4/3 ELISA on CSF0.910.83–0.99
SAG2, 4/3 ELISA serum : CSF titer ratio0.980.95–1.0

Cut-off titers were selected using ROC analysis for each test and sample type to maximize sensitivity and specificity. Optimal results for IFAT were obtained at a serum cut-off titer of 1 : 10, CSF cut-off titer of 1 : 5, and serum : CSF titer ratio of <100. Optimal results for SAG2, 4/3 ELISA were obtained at a serum cut-off titer of 1 : 500, CSF cut-off titer of 1 : 10, and serum : CSF titer ratio of <100. Analysis of test performance is shown in Table 2. Highest sensitivity (88%) was obtained using either CSF or serum : CSF titer ratio and SAG2, 4/3 ELISA. Highest specificity (100%) was obtained using serum : CSF titer ratio and SAG2, 4/3ELISA, although using CSF or serum : CSF titer ratio and IFAT achieved nearly equivalent specificity (98%). Highest overall accuracy (97%) was achieved using serum : CSF titer ratio and SAG2, 4/3 ELISA.

Table 2. Comparison of test performance
TestSensitivity (%)95% CISpecificity (%)95% CIOverall Accuracy (%)95% CI
IFAT on serum5936–787156–836855–78
IFAT on CSF6541–839888–1008877–94
IFAT serum : CSF titer ratio6541–839888–1008877–94
SAG2, 4/3 ELISA on serum7147–875036–645643–68
SAG2, 4/3 ELISA on CSF8866–978672–938675–93
SAG2, 4/3 ELISA serum :  CSF titer ratio8866–9710092–1009788–99

Comparison of IFAT and SAG2, 4/3 ELISA results on serum samples revealed fair agreement, with κ = 0.37 (95% confidence limits, 0.14–0.61). Comparison of IFAT and SAG2, 4/3 ELISA results on CSF samples revealed moderate agreement, with κ = 0.47 (95% confidence limits, 0.21–0.72). Comparison of IFAT and SAG2, 4/3 ELISA results using serum : CSF titer ratios revealed substantial agreement, with κ = 0.67 (95% confidence limits, 0.43–0.90).

Discussion

This study was performed to compare the utility of 2 quantitative immunologic tests for the diagnosis of EPM in clinical horses and to identify the ideal sample type for each test. As expected for a disease with high incidence of seroprevalence, but low incidence of CNS infection, serum testing alone was the least accurate means of diagnosis regardless of test. However, 1 surprising result was the lower sensitivity of serum tests compared with previous estimates.[1, 7] In addition, there was a larger disparity between serum and CSF sensitivity estimates for the SAG 2, 4/3 ELISA than is typical for EPM studies. These findings are likely the result of small sample size combined with the nature of a referral caseload. Several positive cases were serologically negative on 1 or both tests; these cases tended to be horses with acute severe disease that might not have had sufficient time for seroconversion above cut-off levels despite having developed measureable antibodies in CSF. There was limited agreement between tests on serum samples, with test disagreement on approximately one-third of the samples. Approximately 80% of the time, this disagreement occurred when the SAG2, 4/3 ELISA was positive and the IFAT was negative, suggesting that the SAG2, 4/3 ELISA might detect lower levels of antibody. Testing CSF improved the agreement between tests, particularly when serum : CSF titer ratios were used.

When results of this study are compared with those of previous studies evaluating the IFAT,[1, 7, 10] IFAT test performance was not as good, with lower sensitivity and specificity estimates regardless of whether serum or CSF was used. This finding was consistent despite the fact that one of the previous studies was performed at the same institution, albeit with different cases.[7] As both studies were conducted in a very similar fashion, the most likely explanation of the difference is the small sample size, with a total of 19 samples from positive cases in the previous study and 17 samples from positive cases in this study. An alternative explanation would be a change in test methodology at the testing laboratory.

Based on the results of this study, clinicians electing to use the IFAT will obtain most accurate results by submitting CSF for analysis. There was no added benefit to testing both serum and CSF and calculating a serum : CSF titer ratio. However, clinicians using the SAG2, 4/3 ELISA are best served by submitting both serum and CSF samples, as the serum : CSF titer ratio showed superior performance characteristics to testing CSF alone. Overall, the SAG2, 4/3 ELISA titer ratio performed best in this sample population. Additional research is warranted, so that test performance may be better defined using a larger sample set.

Acknowledgment

This study was supported by the Neurologic Research Endowment Fund at New Bolton Center.

Conflict of Interest Declaration: Dr Jennifer Morrow is a co-owner of Equine Diagnostic Solutions, LLC, the laboratory that commercially offers the SAG2, 4/3 ELISA.

Footnotes

  1. 1

    SAG2, 4/3 ELISAs were performed at Equine Diagnostic Solutions, LLC, Lexington, KY

  2. 2

    IFATs were performed by the Immunology/Virology Laboratory at the University of California-Davis, Davis, CA

Ancillary