Among the 97 dogs included in the study, 50 were classified as healthy controls and 47 were identified as MMVD cases. Following assessment of disease severity by the attending cardiologist, MMVD cases were further divided from stage B1 to stage D including 21 dogs in stage B1 (44.7%), eight dogs were in stage B2 (17%), eight dogs were in stage C (17%), and 10 dogs were in stage D (21.3%).
The expression profiles of the selected 15 miRNAs allowed convincing distinction of healthy from MMVD cases in the study, with the optimized predictive model providing an AUC value of 0.93, with an overall cross-validated accuracy of 0.83, a sensitivity of 0.85 and a specificity of 0.82. Variability between MMVD and healthy samples was mostly related to differential expression of a small number of miRNAs including cfa-mir-206-5p, most highly expressed in the healthy group, and cfa-mir-29a-3p, most highly expressed in the MMVD group. cfa-mir-206-5p has previously been associated with atrial fibrillation and autonomic nerve remodeling in dogs37, something commonly encountered in MMVD although not present in the included cases of this study. Increased expression of cfa-mir-29a-3p was seen in MMVD cases and fits with observations in rat models in which increased levels of this marker were associated with cardiac hypertrophy and remodelling37. Other miRNAs such as cfa-mir-133b-3p were poorly associated with any group. This contradicts previously obtained observations in Dachshunds where miR-133b was down-regulated in Stage C compared with Stage A dogs and thus thought to be a possible marker of CHF17. This discrepancy could be linked to variation in size and the wider range of breeds in the current study population.
One possible limitation of the study is some structuring in the data due to age. Although efforts were made to obtain a similar population in the control and MMVD groups, the median age of the healthy control group was lower (4 years) than the median age of the MMVD group (10 years), although the age range of both groups does overlap (2- to 15-year-old and 6.1- to 13.8-year-old, respectively). This is largely related to age being an important factor in development of MMVD1,2,4 making it difficult to design experiments to counter this. However, the possibility of an age-related effect in our data should be considered. Sex has also been reported as an important risk factor in MMVD38. With 35 females (52% intact and 18% neutered) and 15 males (20% intact and 10% neutered) in healthy controls, and 22 females (14.9% intact and 31.9% neutered) and 25 males (19.1% intact and 34% neutered) in MMVD dogs, the current study also concludes statistically significant differences in the distribution of sexes between healthy controls and MMVD dogs. However, sex was not identified as a significant factor of disease when included in the predictive models along with the miRNA profiles.
For the comparison of miRNA profiles between pre-clinical (stage B1/B2; n = 29) from clinical (stage C/D; n = 18) MMVD cases, expression patterns from the 47 MMVD dogs allowed some distinction between the two groups. ROC analysis suggested a moderately high AUC value of 0.82, with sensitivity of 0.61 and specificity of 0.79. The 95% confidence intervals for these metrics showed a marked variability across cross-validation runs, and the overall cross-validated accuracy was 0.73, with misclassification being here more frequent for B1/B2 samples than in stage C/D samples, and the F1 score was moderate with 0.63. Taken together, these data suggest a higher level of uncertainty in separating pre-clinical from clinical MMVD cases, compared to the comparison of control and MMVD cases. Such reduction in sensitivity and accuracy is likely related to the reduced number of samples within this comparative group and would likely be enhanced with an increased sample size.
In the MMVD group, there was a breed over-representation of Cavalier King Charles Spaniels (CKCS) (40.4%). Breed is an important risk factor for MMVD1,4 and the potential influence of specific breed miRNA expression profiles cannot be excluded. There is breed variation for NT-proBNP39–41 and the same is possible for miRNA expression. However, formal statistical assessment and comparison between breeds and miRNA profiles were not possible in this study due a high diversity of breed types providing a low number of individuals per breed. In the future, greater numbers of specific breeds with and without MMVD in different stages will allow to explore any breed influence further.
It was not possible to statistically compare miRNA results and conventional CBs, since there were only a low number of CB results available within the study. Subjectively in these cases, miRNA profiling performed comparably or better, especially in early-stage disease, but more detailed inferences cannot be made at present. With CBs, although there is a trend of increasing values with stage of MMVD42,43, there is no distinct cut-off for a specific stage of the disease (e.g. B2). Since progression of MMVD through the stages is a continuum, this is likely true for miRNA expression. However, the use of a panel of markers gives rise to the possibility of producing a fingerprint profile at each stage, with the potential to identify individual or patterns of miRNA expression which may be predictive of impending stage progression. If a future biomarker panel could reliably identify stage B2, this would a great asset for primary care practice, for making decisions in a dog with a heart murmur, where echocardiography is unavailable. It is anticipated that forthcoming serial, longitudinal monitoring of cases will provide further insight into this possibility.
Although limited in scope, comparison of the diagnostic capability of miRNA profiling to existing NT-proBNP and cTnI biomarkers showed strong promise. The miRNA profiling method performed well in classifying controls correctly, showed an apparent advantage in accurate detection of early-stage B1 and B2 cases over the existing biomarkers, and was comparable in stage C and D cases (Table 3). This provides potential for an improved method of early disease diagnosis, following further validation with a larger sample cohort.
In summary, this study provides strong evidence that a multiplexing detection assay of blood miRNAs can be a useful diagnostic tool for identification of MMVD cases in canine populations. Additionally, miRNA profiling provides the possibility to differentiate between pre-clinical (stage B1/B2) and clinical MMVD (stage C/D). This predictive modelling analysis would now benefit from training on a larger canine cohort constituted of a variety of MMVD cases to continually improve the probabilistic classification algorithms. MicroRNA diagnostic technology has great upcoming potential not only in the field of veterinary cardiology, but current and future applications within the broader veterinary sphere. This technology is patent pending.