This blog has MOVED!

Please visit for the most updated content. All these posts and more can be found over at the new URL.

Tuesday, January 29, 2013

Testing the Test

Since we last talked I have finished the Onco block, started the Infectious Dz block, taken up snowboarding (shared activity with the boyfriend) and have watched Farley become almost 99.99% sound again. It was almost certainly not an abscess, "just" a softening of that old wire scar which caused some pain/discomfort. I say "just" because while I'm glad that she didn't have an abcess, I'm not exactly happy that the old injury raised its ugly head - I purchased her with it and have never had any problems with it.....but now I know that it is something I'll have to keep in mind and manage. At least it wasn't in the middle of my ride season!!!!!!

I'm in the middle of doing some final edits on a post that gives some pointers on how to evaluate literature, and while I was doing so I got on a bit of a tangent. There are 2 terms that I think you should be familiar with in literature, that is most likely to come up when looking at different tests that detect diseases or conditions. This isn't strictly evaluating literature.....but I do think it helps to understand some of the statistical background to some of the tests that you will run across in your everyday horsey world (like the Coggins test).

Sensitivity versus specificity

Sensitivity is the number of "True positives" (animal has the disease AND the test says they have the disease) divided by the total number of animals that actually have the disease, whether or not they tested positive or not (true positive + false negatives).

--> a test that is 100% sensitive means that there is zero false negatives. Every single animal that has the disease tested positive.

Specificity is the number of "True negatives" (animals without the disease that tested negative) divided by the total number of animals that are actually negative, whether or not they tested negative or not (true negative + false positives).

--> a test that is 100% specificity means that there are zero false positives. Every single animal that doesn't have the disease tests negative.

As you can imagine there is usually an inverse relationship between the two --> there is a trade off in specificity when there is high sensitivity and vice versa!

How does this apply to your horsey life?

Let's take the tests for Equine Infectious Anemia (EIA), which is most commonly tested with something called a "Coggins".

There is actually a second type of test that which is a type of ELISA that also tests for EIA.

The Elisa is faster - if the elisa is negative, than the horse is documented as EIA negative. HOWEVER, if the Elisa is positive, the standard coggins has to be positive too to call the horse positive. If the Coggins is negative, then the horse is certified as negative no matter what the Elisa says. If the Coggins is positive, than the horse is EIA positive.

Why might this be?

We are euthanizing horses based on these test results. A positive EIA horse is usually euthanized. We better be darn sure that the horse has it.

The Elisa test is highly sensitive. Remember as a test approaches 100% there are fewer and fewer false negatives - the Elisa test is going capture more diseased animals and call them diseased. But what about the false positives? Those horses without disease that the test mistakenly calls positive?

The Coggins test is very specific. As specificity approaches 100%, there are fewer and fewer false positives. ie - it is unlikely to call a healthy animal diseased.

Which test would you rather have deciding whether your horse lives and dies? A compromise has been reached. The Elisa is fast and if it's negative the animal IS negative and the paperwork is sent off and you get your Coggins cert...but if it's positive, the Coggins test is there as a back up, and we err on the side of not euthanizing a healthy horse, and perhaps missing a few infected animals.

The issue of specificity versus sensitivity is also why it doesn't work to get screened for a large amount of diseases in a shot gun approach - better to match testing with clinical signs and suspected disease because each test has it's own unique specificity and sensitivity.


1 comment:

  1. Who knew? I'm so glad we have you to explain this stuff...


Note: Only a member of this blog may post a comment.