- The therapy is biologically plausible
- The evidence is weak
- The claims for these devices routinely go well beyond the available evidence
A reader recently drew my attention to a new study that is a follow-up of a previous pilot trial.
Pankratz K, Korman J, Emke C, Johnson B, Griffith EH, Gruen ME. Randomized, Placebo-Controlled Prospective Clinical Trial Evaluating the Efficacy of the Assisi Anti-anxiety Device (Calmer Canine) for the Treatment of Canine Separation Anxiety. Front Vet Sci. 2021 Dec 20;8:775092.
The good news is that the study is very well designed, with appropriate controls for many important sources of bias and error. The subjects were selected and randomized appropriately, and everyone was properly blinded. Both a subjective owner measure of effect and objective blinded analysis of video were used to assess effects. There were also active and matched sham devices employed.
As always, there were some limitations to the study. Dogs with behavioral problems other than separation anxiety (SA), such as noise phobia, were excluded. While this makes sense to simplify the study, such problems are common in dogs with SA, so the subjects may not represent the population of SA dogs likely to be treated in the real world.
Dogs were also on varied medications, and none were given a behavior modification program, as would normally be done for this condition, so again how treatment tested might interact with these factors is unclear. The authors also pointed out that some of the study took place during the COVID-19 pandemic, and the activities of owners was likely affected by this event. Finally, there was some missing video due to technical issues, and it is unclear if this might have affected treatment or control dogs differently or otherwise influenced the results.
The authors also mention in the paper that, “after the first 40 dogs completed the study, the devices were sent to the manufacturer for testing; devices that were no longer active were removed from the study and replaced.” They never discuss, however, how many devices malfunctioned and needed replacement, which raises the question of how many dogs may have been “treated” with devices that were not actually operating properly. Presumably, this wasn’t mentioned since it was not a common occurrence, but it would have been useful to include this information.
The less good news is that the results were mixed and inconsistent, yet as is usually the case, they were reported as positive. It is common for research studies that measure multiple outcomes for some to show an effect and others not to. It is often recommended to designate one outcome measure as most important in advance and to use only that to decide if the test treatment works. This avoids the problem of having both successful and unsuccessful outcomes and simply having authors emphasize the positive and de-emphasize the negative.
In this study, the subjective owner assessment showed improvement for all dogs and no difference between treatment and placebo. This is a great illustration of the importance of control groups and placebo treatments in veterinary studies. In this case, the owners were as likely to see improvement with no treatment as with the PEMF device. This was probably due to a combination of caregiver placebo effects and the nonspecific impact of being in a clinical study, where patients get more monitoring and treatment than when they are not in a research trial.
The failure to find any difference in this outcome does not necessarily mean the treatment was not effective, but it does indicate that any benefit it may have had was indistinguishable from placebo to the owners. This is critical since it is mostly the owner perception that determines whether a treatment is thought to be working, and whether treatment is changed or a pet is rehomed or euthanized, in the real world of clinical management of SA.
The more objective analysis of video recordings showed mixed results. In terms of negative behaviors, there was no significant effect of treatment overall, and the comparison between the treatment and sham group was not significantly different at 4 weeks but was different at 6 weeks. The difference from baseline was significant for both groups at week 4 and for only the treatment group at week 6.
In terms of successful treatment, defined as at least a 100% increase in desirable behavior in the video recordings, the treatment group was significantly higher at 4 weeks but not at 6 weeks, and the difference was borderline overall (a p-value of 0.05 is considered statistically significant, and this difference showed a p-value of 0.047).
These results illustrate the problem with multiple outcomes measures and no clear statement before the study of which is primary or what the expectations are. Based on the questionnaire, one could conclude the treatment didn’t work. This is probably not justified since this is a weak measure of effect, but as I mentioned it does signify that the treatment may be no different from placebo in the perception of owners, which is important for how dogs with SA are ultimately managed.
The more objective video assessment showed no overall difference in negative behaviors, which again is consistent with no meaningful treatment effect. There was a difference at 6 weeks, however, and this could be cited to argue that the treatment took time to have a measurable benefit. Unfortunately, this is inconsistent with the results for positive behaviors, which were better for the treatment group at 4 weeks than at 6 weeks, and which again showed a borderline level of statistical significance that could be interpreted as either a real effect or no real difference from chance.
Finally, the importance of consistency across studies and replication cannot be overemphasized. It has become apparent in the last ten years or so that there is a huge problem in biomedical science generally, and in human behavioral science in particular, with replication. Decisions about the causes and treatment of disease are made on a few studies, or even a single study, and then it turns out those results cannot be reproduced, which strongly suggests they were not accurate in the first place.
Veterinary medicine has always suffered from a lack of efforts to replicate research studies, as well as from a lack of good studies in the first place, and this makes it very likely that many of our treatments don’t actually work. All studies should be reproducible before widespread adoption of treatments based on their results occurs, and the mixed results of this particular study make it especially important to show that the findings are repeatable.
This is a well-designed and well-conducted study that shows a mixture of positive and negative results. The most appropriate conclusion would be that the results suggest PEMF might be useful to separation anxiety, but whether or not it is, and how effective it might be, remain to be determined. Given the minimal risks, use of PEMF for this condition is not unreasonable so long as it is not substituted for therapies with better supporting evidence and owners are made aware that the evidence for PEMF is weak and inconclusive.