Antibody Coefficients for the Diagnosis of Equine Protozoal Myeloencephalitis
The work was performed at the Equine Medical Center, Leesburg VA, and the Gluck Center, Lexington KY. Results were reported in part at the Annual Meeting of the American College of Veterinary Internal Medicine, Anaheim, CA, June 2010.
Corresponding author: M. Furr, Marion DuPont Scott Equine Medical Center, Virginia-Maryland Regional, P.O. Box 1938, Leesburg, VA 20176; e-mail: firstname.lastname@example.org.
Background: Diagnosis of equine protozoal myeloencephalitis (EPM) remains a challenge for equine practitioners. Current utilized methods have inadequate sensitivity and specificity, because of a high number of false positive results.
Hypothesis/Objective: Evaluation of antibody indices to Sarcocystis neurona should provide high sensitivity and specificity for diagnosis of EPM.
Animals: Archived samples from 29 clinical patients.
Methods: Archived serum and cerebrospinal fluid (CSF) samples from clinical patients with either EPM (14) or cervical vertebral compressive myelopathy (CVM) (15) were examined and tested for anti-S. neurona antibodies by the SnSAG2 ELISA. The results were used to calculate the antibody index (AI) and C-value. Sensitivity and specificity were calculated, and the AI, C-value, immunoglobulin G (IgG) concentrations, and anti-S. neurona titers compared. In addition, negative CSF was spiked in varying concentrations with blood from a horse with a high anti-S. neurona titer, and the tests repeated.
Results: Results demonstrated that the IgG concentration, anti-S. neurona titer, AI, and C-value were significantly higher (P < .05) in horses with EPM than in those with CVM. Sensitivity and specificity of the AI was 71 and 100%, respectively, and that of the C-value was 86 and 100%, respectively. In addition, the AI and C-value from the samples spiked with S. neurona positive blood remained below 1 (eg, negative) in CSF with a red blood cell (RBC) count up to 105 RBC/μL.
Conclusions/Clinical Importance: Results of the study demonstrate the value of calculating the AI and C-value in the diagnosis of EPM in horses. In addition, the test is robust in the presence of blood contamination.
central nervous system
cervical vertebral compressive myelopathy
equine protozoal myeloencephalitis
indirect fluorescent antibody test
Equine protozoal myeloencephalitis (EPM) remains a diagnostic challenge for veterinary practitioners. Current standards for diagnosis include the presence of clinical signs consistent with the disease, elimination of other illnesses that might result in the clinical signs observed, and confirmation of Sarcocystis neurona-specific antibodies in the serum, or preferably, the cerebrospinal fluid (CSF) of the horse.1 Although there are a number of commercially available tests to detect immunoconversion to S. neurona, all have similar shortcomings. A major impediment shared by current tests is evidenced by the results of various studies showing that 32–89% of tested animals are seropositive, even though the disease is considered to affect <1% of horses.2,3 The ability to differentiate horses that have been exposed to the organism and developed a serologic response (ie, exposure) from those that have the organism within the central nervous system (CNS) (ie, infection) is problematic. The presence of antigen specific antibody in the CSF is thought to confirm CNS infection. This supposition has been found to be incorrect, however, because antigen-specific antibodies to peripherally administered, nonreplicating proteins can be found in the CSF of healthy animals in direct proportion to serum titers.1 This observation suggests that basing a diagnosis on the presence of CSF antibodies alone may lead to many false positive results, which has been found to be the case using immunodiagnosis by Western blot (WB) analysis of CSF.4,5
Proteins are partitioned between the CSF and plasma at a constant ratio, which depends upon the restriction coefficient of the blood-brain barrier (BBB). As plasma concentrations of a particular protein increase, there is a proportionate and predictable increase in the corresponding CSF concentration of that same protein.6–8 This phenomenon also is applicable to antibodies, and occurs because immunoglobulins of different antigenic specificities have the same physicochemical characteristics, and consequently cross the BBB into the CSF in the same proportion of total protein as that found in peripheral blood. This phenomenon can be exploited in the diagnosis of CNS infection, because the CSF titers of antigen-specific antibodies produced within the CNS in response to infection should constitute a greater proportion of total immunoglobulin than that found in the serum. When CSF antibody is of non-CNS origin, the antigen-specific antibody in the CSF should be proportionately equal to that found in the serum.8 Two major mathematical approximations (termed CSF indices) have been used to evaluate these proportions: the Goldman-Witmer coefficient (C-value) and the antibody index (AI). Both methods are standard in the neuroimmunology literature, and have been utilized in human medicine in the diagnosis of a variety of neuroinfections caused by Borrelia burgdorferi,9Angiostrongylus cantonensis,10Toxoplasma gondii,11 and Trypanosoma brucei.12
The application of the CSF indices has been established in the horse with a challenge model of meningitis, in which the ability of the AI and C-value to distinguish intrathecal versus passively acquired antibodies in the CSF was demonstrated.13 In addition, the AI was further tested in an S. neurona challenge model in which all horses challenged with S. neurona sporocysts had positive serum titers, and 2 of 5 had positive CSF titers but no clinical signs or histologic changes in the spinal cord were found.14 AI values in all of the horses were <1, confirming that the antibody was of extraneural origin.14 The purpose of the present study was to evaluate the performance of the C-value and AI using results from an S. neurona antigen-specific ELISA15 for the diagnosis of EPM in clinical patients.
Materials and Methods
To determine the value of the CSF indices (AI and C-value) for the diagnosis of EPM, archived CSF and serum from horses suffering from either EPM or cervical vertebral compressive myelopathy (CVM) were utilized. Serum was collected at the time of diagnosis or euthanasia, separated after clot retraction, and aliquoted for storage at −80°C. CSF was collected at the time of clinical evaluation from either the atlanto-occipital or lumbosacral sites, aliquoted, and stored at −80°C until analysis.
ELISA Assay and CSF Index Determination
Serum and CSF were assayed for anti-SnSAG2 antibodies by a previously described ELISA.15 Reciprocal end-point titers were determined by the predefined ELISA cut-off of 20% positivity. Serum and CSF total protein, albumin, and immunoglobulin G (IgG) concentrations were determined by a commercial laboratory.a These data were used to calculate the AI and C-value by standard formulas, as described previously.13 Briefly, C-values were calculated as the reciprocal CSF titer times the total serum IgG divided by the total CSF IgG times the reciprocal serum titer. The AI was calculated as the ratio of the specific antibody quotient (QAb) over the equine albumin quotient (QAlbe). The QAlbe is calculated as: (CSF albumin concentration/serum albumin concentration) × 1,000, and the QAb is calculated as: the reciprocal CSF anti-S. neurona titer × 1,000/the reciprocal serum anti S. neurona titer. After the CSF indices were calculated, results were categorized as negative or positive at a variety of cut-off values ranging from 0.5 to 2.0.
To determine the impact of blood contamination on the resultant calculated values, a CSF sample from a seronegative horse was spiked with whole blood from a horse that had a SnSAG2 serum titer of 1 : 1,000. Blood was added in amounts to achieve red blood cell (RBC) counts of 105, 104, 103, and 102 RBCs/μL of CSF, and the assays described above were repeated. The range of RBCs added for testing was chosen to represent the range of values expected in a clinical setting with blood contaminated samples.
One test group was comprised of horses with CVM. Horses were classified as having CVM based on neurologic examination and a myelogram demonstrating compression as well as postmortem and histologic examination of the cervical spinal cord demonstrating pathologic changes characteristic of CVM. These findings include, but were not limited to, neuron fiber degeneration in ascending and descending tracts, demyelination of white matter tracts, and astrocyte gliosis. Horses were included if they met this diagnostic standard, and appropriate samples were available for assay.
A 2nd group consisting of horses with EPM also was identified. To be considered a case, the medical records of horses were closely examined and the following criteria met: horses had to demonstrate clinical signs of a neurologic abnormality consistent with EPM, have serum and CSF collected, and have radiographs of the cervical spine performed and within normal limits. The range of clinical neurologic abnormalities associated with EPM is wide, including almost any abnormality. Most commonly clinical abnormalities include ataxia and muscle wasting.
Postmortem examination was performed in some cases, and horses were included if changes characteristic of EPM (lymphoplasmacytic perivascular cuffing, presence of organism) or positive culture for S. neurona organism were noted. For horses with a clinical diagnosis, diagnosis was based on observation of characteristic clinical signs (eg, ataxia, muscle atrophy), seropositivity (CSF WB) normal cervical radiography, negative CSF serology for herpesvirus 1, and prompt response to treatment with antiprotozoal medications (ponazuril). Horses were excluded from the EPM group if cervical radiographs found narrowing of the vertebral ratios, cervical arthritis, or other neurologic conditions. In addition, horses were removed if no histologic changes consistent with EPM were found on postmortem examination.
The determination of inclusion was made before performing the SnSAG2 ELISA and calculating CSF index values, and was solely determined by review of clinical and pathological data available on each case, as described above.
Results were entered into a desktop computer with a commercial statistical analysis program.b After data entry validation, summary statistics were determined for continuous variables and reported as mean ± 1 SD. Sensitivity and specificity, with 2-tailed 95% confidence intervals, were calculated for the C-value and AI of all horses, at various cut-off values, by standard methods. Continuous data were evaluated by Student's t-test, with a predetermined P-value for significance of <.05.
A total of 37 horses met the criteria for inclusion and had appropriate samples available. Eight horses that had a recorded, historical diagnosis of EPM were removed because they did not meet the inclusion criteria described above, leaving a total of 29 horses for evaluation. The cervical compression group consisted of 15 horses, with 14 horses in the EPM group.
Of the horses with EPM, 9 were confirmed by postmortem examination, and 5 were included based on a rigorous clinical evaluation. Of horses (9) with postmortem confirmation, S. neurona was cultured from 2, whereas characteristic histologic changes were noted in the other 7; organisms were observed in only 1 horse.
Results of serum and CSF testing are summarized in Table 1. Thirteen of 14 horses in the EPM group had positive serum and CSF titers, whereas 12/15 of the horses in the CVM group had positive serum titers ranging from 0 to 1,000 (400 ± 264). Using a cut-off of >1 being positive, the C-value was positive in 12 of 14 horses with EPM, and was negative in 15 of 15 horses with CVM, resulting in a sensitivity of 86% and a specificity of 100%. The AI was positive in 10/14 horses with EPM, and negative in 15 of 15 horses with CVM, for a sensitivity of 71% and a specificity of 100%. Sensitivity and specificity data for various cut-off points for the CSF indices are presented in Table 2.
Table 1. Summary data for cerebrospinal fluid (CSF) and serum values in horses with either cervical vertebral compressive myelopathy (CVM) or equine protozoal myeloencephalopathy (EPM).
|Serum titer−1||1,714.8 (2,233.2)||400 (263.9)||.04|
|CSF titer−1||114.8 (211.8)||0.2 (0.8)||.04|
|Serum IgG (mg/dL)||2,319.1 (682.1)||2,568.5 (798.9)||.37|
|CSF IgG (mg/dL)||15.3 (6.0)||10.1 (5.6)||.02|
|Serum albumen (mg/dL)||3,335.8 (396.2)||3517.1 (541.4)||.31|
|CSF albumen (mg/dL)||37.9 (12.9)||30.0 (10.1)||.08|
|C-value||13.3 (20.0)||0.03 (0.13)||.02|
|Antibody index||7.8 (11.5)||0.17 (0.07)||.02|
Table 2. Sensitivity and specificity at various cut-off points of the cerebrospinal fluid (CSF) indices with 95% confidence intervals (lower, upper).
|(0.64, 0.99)||(0.64, 0.99)||(0.39, 0.87)||(0.75, 1)|
|(0.56, 0.97)||(0.75, 1)||(0.39, 0.87)||(0.75, 1)|
|(0.56, 0.97)||(0.75, 1)||(0.42, 0.90)||(0.75, 1)|
|(0.49, 0.94)||(0.75, 1)||(0.42, 0.90)||(0.75, 1)|
|(0.42, 0.90)||(0.75, 1)||(0.42, 0.90)||(0.75, 1)|
|(0.42, 0.90)||(0.75, 1)||(0.42, 0.90)||(0.75, 1)|
|(0.42, 0.90)||(0.75, 1)||(0.39, 0.87)||(0.75, 1)|
Analysis of CSF spiked with blood from a seropositive horse (1 : 1,000 serum titer) indicated that ELISA results were impacted at only the highest amount of blood contamination (100,000 RBCs/μL CSF). With 105 RBC/μL, the CSF sample exhibited a 1 : 12 endpoint titer. In addition, this degree of blood contamination resulted in an increase in the CSF albumin concentration to 217.5 mg/dL and the CSF IgG to 57.4 mg/dL. Calculation of the CSF indices resulted in a C-value of 0.64, and an AI of 0.23. In CSF spiked with 104 or fewer RBC/μL, none of the sample dilutions yielded a percent positivity value that exceeded the cut-off of 20% (ie, were considered negative).
This study was designed to evaluate the CSF indices C-value and AI as aids in the diagnosis of EPM. Results demonstrated the value of the AI and C-value methods in aiding the diagnostic classification of horses both with and without EPM. Of notable value was the ability of the C-value in particular to discriminate between seropositive horses with CVM and horses with EPM. Test performance also was strong in the presence of blood-contaminated CSF, a common complicating factor in the clinical diagnosis of EPM. In the diagnosis of EPM, the challenge is to differentiate animals that are exposed to the parasite, and hence have an immunologic response, from those that truly have the disease (ie, infection of the CNS). Infection of the CNS is necessary for the development of clinical signs, and demonstration of this antigenic stimulation within the CNS (ie, intrathecal antigen-specific antibody production) is crucial to the diagnosis of EPM. This has been attempted in a variety of ways, yet interpretation of results is confounded by the observation that there is passive movement of antibodies into the CSF in normal animals. Thus, it is often not possible to determine if the antibodies detected in the CSF are a consequence of passive diffusion or were locally produced. The CSF indices represent a mathematical method to evaluate the CSF results and differentiate between these 2 possibilities.
Diagnosis of clinical disease based on a titer is sometimes challenging because infection is not always directly correlated with titer, and the influence of CNS inflammation, BBB integrity, iatrogenic blood contamination, and CNS hemorrhage on the resultant CSF titers is difficult to determine. In addition, factors such as individual immune response, variations in titer because of antigen dose or virulence, and the timing of the sampling relative to infection, for example, cannot be accounted for with a single titer determination. The CSF indices attempt to control for such variability by evaluating the CSF titer as a proportion of the serum titer, as influenced by passive diffusion of antigen-specific immunoglobulins into the CSF. Hence, the CSF indices are based on patient-specific characterization of the properties of the BBB, and this is a key strength of the indices.
In the current study, horses with well-characterized EPM and CVM were used to evaluate the performance of the indices for the diagnosis of EPM. To determine antibody titers, an ELISA based on the S. neurona surface antigen SnSAG2 was utilized. This ELISA previously was shown to provide a highly accurate method (95.5% sensitivity, 92.9% specificity) for detecting the presence of antibodies against S. neurona.15 Results of the index calculations utilizing end-point titers from the SnSAG2 ELISA indicated a high specificity and sensitivity for both indices described. The C-value had the best overall performance, with a specificity of 100% and a sensitivity of 86% in this study sample, and using a cut-off value of >1 as positive. This is to be expected, because the theoretical upper limit of normal for both CSF indices is <1 in normal horses. The same overall performance was achieved using a cut-off of 0.5 for the C-value; however, this resulted in 1 CVM horse being misclassified as an EPM case as well as correct classification of an EPM case that was misclassified using a cut-off of >1. Additional studies by large data sets are necessary to optimize the critical value for test interpretation. Until such studies are completed, it seems prudent to use a cut-off value of >1.
In the current study, the value of the AI was always lower than the C-value, and in some marginal cases this resulted in the AI being <1, and hence were considered negative in a true positive case. The overall performance of the indices was strong, given that almost all horses with CVM had serum titers that might have been considered clinically relevant. A few horses with CVM also had positive CSF titers, and easily could have been clinically diagnosed as EPM cases.
The CSF indices utilized in this report have been described previously in horses.8,13 In a model of equine neuroinflammation, CSF indices increased dramatically in horses with intrathecal immune challenge, but remained normal, even in the presence of substantial CNS inflammation, for an infectious agent (herpesvirus) with which the horse was not challenged.13 Although this demonstrated proof of concept, the value of the indices in the diagnosis of EPM was not specifically evaluated.
The use of a similar CSF AI was described by Heskett and MacKay14, after intragastric challenge with S. neurona sporocysts. The AI used in the Heskett study was more similar to the mathematical form of the C-value in the present study, but the theoretical maximal normal value still is <1. In this model, 11 of 14 horses challenged with intragastric S. neurona were subsequently found to have a positive CSF WB, yet the AI did not change, although the CSF titer increased by 31-fold after challenge.14 No clinical signs were observed in any challenged horse, and no histologic abnormalities of the CNS were identified after challenge. The authors concluded that AI effectively demonstrated that the increase in CSF S. neurona-specific IgG was not attributable to synthesis within the CNS, but rather was a consequence of passive diffusion from the general circulation.
An alternative approach to the paired evaluation of serum and CSF anti-S. neurona titers has been suggested by other authors, by the indirect fluorescent antibody test (IFAT), in which the ratio of serum to CSF anti-S. neurona titers was evaluated in a group of S. neurona-vaccinated horses and horses with EPM.16 In this report of preliminary findings, the authors suggested that a ratio of serum titer to CSF titer <16 discriminated vaccinated and infected animals with a high sensitivity and specificity (100 and 95%, respectively). These results further strengthen the argument that serum and CSF titers should be evaluated in concert to provide the best possible diagnostic classification of horses with EPM.
A complicating issue for the interpretation of CSF titers and WB tests has been the concern that iatrogenic blood contamination during the CSF collection can lead to spurious results. Such false positives have substantial importance because they lead to an incorrect diagnosis precluding or at least delaying the correct diagnosis, they lead to unnecessary treatment with attendant costs, and the diagnosis may negatively impact sales. A previous study by Miller et al17 found that contamination of a truly negative CSF sample with a quantity of blood from a seropositive horse resulting in a CSF RBC count of 8 cells/μL was adequate to convert the sample from WB negative to WB positive. A study by Finno et al18, however, found that using the IFAT, blood contamination had no effect on the resulting titer at RBC counts up to 10,000 RBCs/μL CSF. The results of the current study also indicate that application of the described CSF indices by the SnSAG2 ELISA is robust in the presence of blood contamination >10,000 RBCs/μL CSF, allowing proper assignment of cases in which there is substantial blood contamination of the CSF sample. This performance is expected for the CSF indices because any blood introduced into the sample or entering the CSF will have S. neurona specific antibodies in the same proportion as in serum that passively crosses the BBB, hence the resultant calculated value will be unaffected by the contamination.
The present study demonstrates the value of the C-value and AI as aids in the diagnosis of EPM. The indices are readily calculated from data generated by straightforward assays, thus providing an easily interpreted value. In addition, their application allows analysis of CSF with blood contamination, precluding repeat sampling or disallowing the use of a diagnostic sample that has been collected.
a Equine Diagnostic Solutions, Lexington, KY
b SAS 9.6.1, SAS Institute, Cary, NC
Funded by an internal grant from Virginia Tech.