Example 1

Equine Protozoal Myeloencephalitis (EPM) Review

EPM is one of the most common and serious neurologic diseases in horses. The disease is caused by an infection with Sarcocystis neurona, a protozoan parasite that exhibits a predilection for the central nervous system. This parasite has also been identified to cause similar symptoms in cattle, sheep, dogs, cats, raccoons, skunks, mink, chickens, a golden eagle, a rhesus monkey and harbor seals. The clinical signs associated with this disease are classically an asymmetric ataxia with associated muscle atrophy.

The ante-mortem diagnostic tests that are currently used are a Western blot analysis for antibody to Sarcocystis neurona in cerebrospinal fluid or a PCR analysis for detection of DNA of Sarcocystis neurona. Originally, the diagnosis was based on post-mortem findings of necrosis of any portion of the central nervous sytem and presence of the organism using immunohistochemistry. However, clinical cases have had to rely on the ante-mortem tests for diagnosis and appropriate treatment. Based on 295 post-mortems, it has been determined that the sensitivity and specificity of the Western blot for detection of antibody to Sarcocystis neurona is 89%. This finding, however, is based on the most severe cases due to the fact they had a post-mortem.

The problem with the diagnosis of this disease has been the populations to which the test has been applied. Due to the seriousness of EPM, paranoia has swept through the horse industry and numerous attempts have been made to use this antibody test as a screening test. However, due to the difference in prevalence in different populations, interpretation of the test is difficult.

For example, normal horses are being screened for EPM either as a stable policy or as part of a prepurchase examination. The prevalence of EPM in the general horse population has been estimated at 1%. If we use 1000 animals as an example, using a sensitivity and specificity of 89%, the breakdown will be as follows:

 Disease +Disease -TotalPredicted value
Test +9109118PV+ 7.6%
Test -1881882PV- 99.9%
Total109901000 

The result of a positive test in the above example would be very difficult to interpret. When the sample was obtained from the cerebrospinal fluid collection, was there blood contamination? Does this mean the horse is infected but not developed clinical signs? Has the horse cleared the infection and the resultant positive antibody test is due to the remaining antigens not yet cleared from the central nervous system? The "negative predictive value" is very high, and if forced to perform this test in normal horses, this is the only interpretation of value.

The next example we will use is a horse with equivocal mild neurologic signs and we extimate that the prevalence of EPM in this population is 10%. If we use 1000 animals as an example, using a sensitivity and specificity of 89%, the breakdown will be as follows:

 Disease +Disease -TotalPredicted value
Test +8999188PV+ 47.3%
Test -11801812PV- 98.6%
Total1009001000 

In this group, the "positive predictive value" has increased to almost 50%, therefore, if one receives a positive test, there is a 50-50 chance that the horse is truly diseased. Once again, the "negative predictive value" is extremely high, and therefore, is of significant value.

If a neurologic horse is admitted to the OSU Veterinary Medical Center, approximately 50% of the neurologic horses are diagnosed with EPM. This represents a prevalence of 50%. If we use 1000 animals as an example, using a sensitivity and specificity of 89%, the breakdown will be as follows:

 Disease +Disease -totalpredicted value
Test +44555500PV+ 89%
Test -55445500PV- 89%
Total5005001000 

In the above instance, we have much more confidence in a positive test representing a "truly diseased" animal. The "negative predictive value" is reduced and you would likely have a difficult time believing the horse was "non-diseased". However, the usefulness of the diagnostic test (western blot) would be best where the prevalence is close to 50%.

The last example we will use is the horse that has neurologic deficits worse on one side of its body than the other. Along with the asymmetric ataxia, there is evidence of severe muscle atrophy of the shoulder and quadriceps muscles. For the purpose of this discussion, very few health problems with these presenting neurologic signs would be anything other than EPM. Therefore, the prevalence in this population of animals would be approximately 90% or greater. If we use 1000 animals as an example, using a sensitivity and specificity of 89%, the breakdown will be as follows:

 Disease +Disease -TotalPredicted value
Test +80111812PV+ 98.6%
Test -9989188PV- 47.3%
Total9001001000 

In this particular example, if the horse presented to you, testing any further after your clinical exam would be questionable. First, you would not believe a negative test result if that was the outcome. Since the "positive predictive value" is so high when the animal presents as this horse does, instituting immediate treatment for EPM and monitoring for response to therapy is the most likely approach.

As has been illustrated, as the prevalence increases, so does the "positive predictive value". Conversely, the "negative predictive value" decreases as the prevalence increases. Therefore, it is important when applying the Western blot analysis to consider the population in which you are applying the test. This test should not be used in normal horses and is likely of little benefit in horses that are exhibiting the classical signs of EPM.