If you are recommending that your client spend several hundred dollars on an allergy test for their dog, it would be nice to have confidence that the test is reliable.  It should demonstrate both test-retest reliability (repeatability) and inter-lab reliability (reproducibility).  In human medicine, allergy testing companies are required to participate in an interlaboratory proficiency program using standardized serum pools. Unfortunately, such a program has not yet been adopted in veterinary medicine, although the International Committee for Atopic Diseases of Animals is working on it.  This leaves us with relatively little information about which serum allergy test (SAT) to recommend.

Over the years, most studies examining SAT reliability, including one I presented to the ACVD in 1994, have found most to be less reliable than desired. Since then, many of the companies have changed their methodology, although some do not readily disclose their anti-IgE antibody source or enzyme conjugate, which can profoundly influence results.

Two more recent studies have been published. Thom et al. evaluated the reliability of three independent European allergen-specific IgE testing laboratories all employing Heska techniques and reagents. In terms of differences in reporting positive versus negative reactions, they reported a 3% intralaboratory discrepancy rate and a 9% interlaboratory discrepancy rate for the 3 labs.  What I am unable to glean from this study, is how this would have differed from chance alone.  If positive reactions were either very common or very uncommon, that would have a significant effect on the expected agreement. The kappa statistic is often used to account for this, but was not reported in this paper.

In the second report, SAT results were compared between two labs that employ Greer macELISA methodology and also between Greer’s reference lab and Heska’s fceELISA test. Pools of sera were created from samples of known macELISA reactivity and duplicate samples submitted for each of the comparisons.  Agreement was reported in terms of percent concordance of positive and negative reactions, and it was found to exceed 90% for both interlaboratory comparisons. Kappa was not reported, nor was the proportion of the 18 serum pools that had high, intermediate, or low (negative control) reactivity.  Again, if the samples were skewed to very high or low reactivity, positive/negative agreement would look better than for samples with reactivity closer to the cutoff value.

Following the second report, I submitted duplicate samples from two atopic dogs, with one sample going to Heska and the other to Greer. From one dog, both samples were negative.  From the second dog, one of the samples showed no positive reactions, while the other lab reported several highly positive reactions to mites and one weed.  I have since followed up on this finding by evaluating the agreement of four commercial allergen-specific IgE testing laboratories, both in terms of reported positive allergens, and allergens recommended for immunotherapy by laboratory personnel. A colleague submitted four identical samples from each of four dogs to different laboratories. Since each lab tests for a unique set of allergens, I examined the results a) as tested (ungrouped) and b) treating similar or cross-reacting allergens as the same (grouped).  Figure 1 shows the total number of pairwise agreements and disagreements.

The relatively high number of disagreements compared to positive agreements results in kappa values that are for the most part in the poor range (< 0.2), with just 1/6 lab comparisons displaying agreement in the fair range for treatment recommendations.  I’m not able to present the entire study here, since I hope to publish it.

Another way of putting it, 81% of the allergens recommended for treatment for these four dogs in their treatment sets were recommended by only one of four laboratories, and just 4% were recommended by 3 or 4.

My bottom line:  if you choose to send in a blood sample for serum allergy testing, your choice of labs has a major influence on their immunotherapy prescription recommendation, yet we don’t know which test is more accurate. Most clients are not keen on spending hundreds of dollars on a test if given that bit of information.