What the index is, exactly.
Dysbiosis is a word that does a lot of work in marketing and much less in the literature. In dogs it has one concrete, published, reproducible definition, and it is worth knowing precisely what that definition contains.
AlShawaqfeh and colleagues built the canine dysbiosis index in 2017 from faecal DNA from 95 healthy dogs and 106 dogs with histologically confirmed chronic inflammatory enteropathy. The final panel is eight qPCR assays: total bacteria, Faecalibacterium, Turicibacter, Escherichia coli, Streptococcus, Blautia, Fusobacterium and Clostridium hiranonis. Seven of those were used to build the model and the total bacteria assay normalises the rest, which is what makes results comparable between samples.
The index itself is a distance. Each sample is placed relative to the average healthy dog and the average diseased dog in the training set, and the index is the difference between those two distances. A negative value indicates normobiosis, a positive value indicates dysbiosis, and zero means the sample sits equally far from both. A sample at 8 is further from the healthy reference than a sample at 2.
Six of the eight measures fall in disease: total bacteria, Faecalibacterium, Turicibacter, Blautia, Fusobacterium and C. hiranonis. Two rise: E. coli and Streptococcus. All eight differed significantly between the groups at p less than 0.001.
How well it performs, and where it fails.
On the combined dataset the index achieved an area under the curve of 0.93, with a 95 per cent confidence interval of 0.89 to 0.96. That is a strong result, and considerably better than any single assay in the panel: the individual AUCs ran from 0.69 for total bacteria to 0.86 for C. hiranonis, with Fusobacterium at 0.80 and Faecalibacterium at 0.78. Combining the assays is what buys the performance.
At the usual threshold of zero, the index gave 74 per cent sensitivity and 95 per cent specificity. Lowering the threshold to minus one raised sensitivity to 82 per cent and dropped specificity to 91 per cent. Raising it to plus two pushed specificity to 100 per cent and sensitivity down to 63 per cent.
Read that asymmetry carefully, because it determines what the number is good for. At threshold zero, a positive result is unlikely to be a false alarm: only about one healthy dog in twenty crosses the line. But roughly one dog in four with confirmed chronic enteropathy sits below it. The index is much better at confirming a suspicion than at dismissing one.
The distributions make the same point. Median dysbiosis index in healthy dogs was minus 4.8, with a range from minus 9.1 to plus 1.7. In dogs with chronic enteropathy the median was 3.3, with a range from minus 6.2 to plus 9.3. The medians are far apart and the ranges overlap substantially. The authors are direct about why: chronic enteropathy is multifactorial and the microbiota is just one of several factors in its pathogenesis, which is also the likely cause of the overlap, and the same overlap is reported in humans with inflammatory bowel disease.
Two further limits are worth stating. The index tracked disease severity, correlating with the clinical activity index at r equal to 0.682, p less than 0.001. But it could not localise the problem: there was no significant difference between dogs with predominantly small intestinal signs and dogs with predominantly large bowel signs, p equal to 0.416. It tells you the community has shifted. It does not tell you where, or why.
On the credit side, the assay is stable. Coefficients of variation across five aliquots of five faecal samples ran from 2.9 to 10.5 per cent, and sequencing of the amplicons confirmed primer specificity. Whatever its interpretive limits, it measures the same thing twice.
One disclosure belongs here. The paper states that three of its authors are employees of the Gastrointestinal Laboratory at Texas A&M, which offers microbiome testing as a diagnostic service to veterinarians, and it also carries the line that no conflict of interest is declared. Both statements appear in the paper. Readers can weigh that themselves; we mention it because we are a commercial laboratory writing about a commercial laboratory's index.
Why this matters more than a number usually does.
Chronic enteropathy is not a minor inconvenience, and the evidence on what it costs the household is worth putting next to the assay performance.
Díaz-Regañón and colleagues in 2023 surveyed 110 dog owners, 30 whose dogs had been diagnosed with inflammatory bowel disease under WSAVA criteria and 80 with healthy dogs, using a validated 1 to 10 scale. Dogs with the disease scored 6.30 for general quality of life against 9.10 in healthy dogs, p less than 0.0001, and 6.07 for health against 8.95, p less than 0.0001.
The owner findings are the ones that stay with you. Owners of affected dogs rated their own general quality of life at 7.20 against 8.39 in the control group, p equal to 0.001. They reported feeling burdened by caring for their dog at 5.40 against 3.63, p equal to 0.006, and reported greater limitation of their social life, leisure time and daily activities, 6.77 against 5.28, p equal to 0.015. Asked directly what impact the disease had had on their own quality of life, on a scale where 1 was negative and 10 positive, the median answer was 4 out of 10.
The study reports that the effect held regardless of the severity of the disease. A mild case and a severe case were both hard to live with. The study received no external funding and declares no conflict of interest.
That is the context in which a dysbiosis number gets read. It lands in a household that is already tired, and it will be over-read in both directions unless what it can support is stated plainly.
What a dysbiosis index should and should not be asked to do.
The index was trained on dogs with histologically confirmed disease against dogs verified as healthy. Its published performance is performance in that comparison. Applying it to a well animal with no clinical signs is a different task with a different prior probability, and the paper reports no performance figures for that use.
The authors position it accordingly: the index may be helpful to assess whether the microbiota is normal or dysbiotic in chronic enteropathy, as part of an assessment that would also encompass intestinal and systemic inflammation. It is one input into a clinical picture, not the picture.
So the honest reading of a dysbiosis value is narrow. A clearly positive value in a dog with gastrointestinal signs is a real finding worth taking to a vet. A normal value in a dog with ongoing signs rules out very little, because a quarter of confirmed cases sit in that range. And a value of either kind in a dog with no signs at all is a description of a community, not evidence of a disease.
Where BAARK fits.
A number with its limits attached is more useful than a number without them.
BAARK reports the composition of the community found in the sample. Where a published index exists for a measure we report, we name the study it came from, the population it was trained on and the performance it achieved, so the figure can be read against what it was built to do rather than what a marketing page would prefer it to mean.
BAARK is wellness profiling, not a diagnostic test. It does not detect, diagnose, stage or predict the progression of chronic enteropathy or any other disease. A dog with ongoing vomiting, diarrhoea, weight loss or blood in the stool needs a veterinarian, and needs one whatever any microbiome report says.
The bottom line.
The canine dysbiosis index is a genuinely useful instrument, built on 201 dogs, reproducible to within about 10 per cent, and reaching an AUC of 0.93 in the comparison it was designed for. At its standard threshold it clears 95 per cent of healthy dogs and misses about 26 per cent of sick ones. It cannot tell small bowel from large bowel disease, its healthy and diseased ranges overlap, and it was never validated as a screening test for animals without signs. Used for what it was built for, it is good. Used as a verdict on a well animal, it is not evidence of anything.