Every day, we are bombarded with new information about dogs that arises from a variety of sources – via the internet, through our smart (or not so smart) phones, from our colleagues, friends and family, and of course from our neighbor Joe next door (who happens to know a lot about dogs). In this day and age of information overload, it should come as no surprise that deciding which information is trustworthy and which to view with a healthy dose of skepticism is increasingly difficult. Lucky for dog folks everywhere, science comes to the rescue once again.
Even within science however, all evidence is not created equal. Though not quite as dramatic as walking on water, the construction of a handy “evidence pyramid” helps us to sort the various categories of scientific study and to rate the types of information that they provide.
Interpreting the Pyramid: First, keep in mind that this graphic presents only scientific evidence and does not include the host of other types of information that we come across each day, such as anecdotes, testimonials, stories/experiences, and (non-expert) opinion. (For a review of all of these types of evidence, see my new book, Dog Food Logic). As you move from the bottom to the top of the pyramid, the amount of information (published literature) decreases, but its relevance and reliability generally increases. Starting at the bottom:
- Editorial and expert opinions: Experts in a field will often produce textbooks and review papers that can provide a good foundation about the chosen topic. This information presents a helpful summary, but because it is not reporting results of a scientific study, cannot provide evidence that supports or refutes a new scientific hypothesis.
- Case reports: These are individual reports, usually published by a practicing veterinarian, of one or more dogs with a condition who are found to respond to a particular treatment. People are often surprised that case reports are not regarded as stronger scientific evidence. However, while case reports may generate new hypotheses, they cannot be used as strong support for an existing hypothesis because of their anecdotal nature and the lack of control groups (remember our Steve Series).
- Case-controlled and cohort studies: Case controlled studies occur when the researcher finds cases (dogs, in our example) that have the condition under question and then matches and compares those cases with other dogs who are similar, but lack the condition. Similarly, cohort studies compare large groups of dogs with or without a condition, over time. While providing a type of control, results of these studies are limited because showing a statistical relationship (usually correlation) between two groups does not mean than one factor necessarily caused the other.
- Randomized controlled trials (RCT): This type of study is the most important and reliable source of scientific research. It includes methodologies that use the scientific method, reduce the potential for bias (via randomization, blinding, and the use of placebos) and allows for comparison between intervention groups and carefully selected control groups.
- Systematic Reviews: At the pinnacle of scientific evidence, this is is a specific type of review in which experts in a field assess all of the relevant studies and the data (called a meta-analysis) that address a particular topic. Systematic reviews require enormous commitments of effort, time, and money and are only possible once a hypothesis has been studied in depth. Therefore, these studies are few in number, especially for many topics of that are important to dogs. (For a good source of a few systematic reviews in canine health see Best Bets for Vets).
AN EXAMPLE – THYROID ON TRIAL
An example: I recently came across a great example of a hypothesis about canine health and behavior that progressed from a few initial case reports, through case-controlled studies and culminated recently with the gold standard – the completion of a randomized, controlled trial. The issue had to do with a common endocrine disorder in dogs, hypothyroidism, and its potential relationship with aggressive behavior.
Background information: Thyroid hormone is produced by the thyroid gland. The active form of this hormone regulates cellular metabolism and so has effects in virtually all body systems. The condition of hypothyroidism refers to a reduction in thyroid hormone production and resulting clinical signs. Hypothyroidism generally develops in middle-aged or older dogs and certain breeds show a genetic predisposition. Documented clinical signs of hypothyroidism include lethargy, decreased interest in exercise, weight gain, changes to coat quality and hair loss, and skin problems such as seborrhea, hyperpigmentation and secondary bacterial infections.
Hypothesis: In recent years, it has been speculated that certain types of aggressive behaviors in dogs may be related to suboptimal or low thyroid hormone levels. Starting with case reports, this hypothesis has gradually worked its way up the evidence pyramid:
- Case reports: Two case reports were published in 2002 and 2003. Together, they involved a total of 5 dogs with owner-directed aggression, who were subsequently also diagnosed with hypothyroidism. The dogs responded to thyroid hormone replacement therapy with a reduction in aggressive episodes, leading to the hypothesis that some cases of aggression in dogs may be associated with hypothyroidism (1,2).
- Case-controlled studies: A case-controlled report was conducted a few years later (3). Records of over 1500 dogs were reviewed. Of these dogs, 61 per cent were classified as either hypothyroid or with suboptimal thyroid function. A statistically significant correlation was found between thyroid dysfunction and dog-to-human aggression in this group of dogs (p < 0.001). However, two subsequent case-controlled studies failed to find a connection between thyroid hormone levels and behavior problems (3,4). Conflicting results – jury still out.
- The RCT: Most recently, in 2013, the connection between suboptimal thyroid hormone levels and aggression in dogs was examined using the Gold Standard of designs – a double-blind, randomized, placebo-controlled study (5). Dr. Nick Dodman and his colleagues at the Tufts Cummings School of Veterinary Medicine enrolled a group of 40 dogs, all of whom were exhibiting owner-directed aggression and were also diagnosed with suboptimal or low thyroid hormone levels. Following screening and a 2-week pre-treatment (baseline) period, the dogs were randomly assigned to either the treatment group (thyroxine replacement therapy) or to the control group (placebo). Neither the researchers nor the owners knew which group each dog was assigned to. Dogs were medicated twice daily for a period of 6 weeks. During both the baseline period and throughout the study period, owners recorded the number and type of aggressive episodes that their dog exhibited.
- Enrollment and attrition: The highly specific inclusion criteria for the dogs in this study (aggressive dogs that had borderline or low thyroid function) coupled with difficulties associated with working with dogs with aggression problems led to a relatively small initial sample size (n = 40). In addition, attrition was high, due to owner non-compliance or other problems.
- Frequency of aggressive behavior within groups: The frequency of aggressive episodes significantly decreased in both groups from baseline levels over the six-week experimental period. The change in the control group demonstrates a significant placebo effect in this study.
- Treated group vs. placebo group: Owner-measured aggression did not differ between the treated group and the placebo group during the first five weeks of the study period. During the final week of the study (week six), dogs who were treated with thyroxine and had normalized serum thyroxine levels showed slightly lower frequencies of aggression when compared with dogs who were receiving the placebo, but this difference was not statistically significant (P = 0.08).
Take away for dog folks: This type of design is truly the “gold standard” of experimental studies for several reasons:
- Methodology: An RCT design reduces bias and allows relevant comparisons because a matched control group is used, treatments are randomly assigned, and neither the experimenters nor the subjects (in this case the owners of the dogs) know which treatment each dog is receiving. In addition, the inclusion of a placebo treatment (as opposed to simply not treating the control group at all) allowed the researchers to measure and account for a placebo effect (which clearly was important in this study).
- Clinical trials using dogs in homes: This study was a clinical trial, meaning that it was conducted with owned dogs living at home with their owners. This differs from studies conducted with dogs living in kennels (typically at a university setting or at a pet food company’s kennel). If you remember back to The Steve Series, a cornerstone of the scientific method is selecting a study sample that is representative of the population that you will make conclusions about. Therefore, while we can control many of the “variables of life” with kenneled dog studies (making those studies much easier to conduct and to detect differences) , such a sample is by definition, less representative of the population of dogs than is a sample that includes dogs living at home.
Challenges: In-home clinical trials with dogs are wrought with enormous challenges, all of which make it difficult to demonstrate real effects when they exist and which can require larger sample sizes to detect any true differences. These include:
- Variations in daily life: Every owner lives with his/her dog in ways that are idiosyncratic to that person’s demographics, lifestyle, and values. These differences all impact an owner’s perceptions of his dog’s behavior (in this example, displays of aggression) as well as a tendency to show a placebo effect (see below).
- Dog differences: Generally speaking, the differences among pet dogs enrolled in a clinical trial are going to be greater (spread more widely around the mean) than those among a group of kenneled dogs. The most obvious difference is the variability in living situations and daily routines among households. These are not present when studying a group of dogs who are housed under the same conditions and experience the same daily routines. For a researcher, this means that being able to identify a treatment effect (in this case, a measurable reduction in aggressive episodes in dogs treated with thyroid hormone), is much more difficult when studying dogs in homes compared with studying dogs in kennels.
- Owner perceptions and compliance: When the owner is the data collector in a study (which is sometimes the only feasible approach with in-home studies), there will be error (variability) introduced by the different perceptions among owners as well as by varying levels of compliance. Extreme non-compliance usually leads to removal from the study, but this too is a problem since removing subjects from an already limited sample will further reduce the power of the experiment (i.e. the ability to detect a true difference when it exists).
- Placebo effects: Just as in studies with human subjects, the placebo effect is a real effect that must be accounted for in dog studies. When owners are aware that their dog is enrolled in an experimental trial, even though they are blinded to the treatment that their dog is receiving, the mere participation in the study will affect their perspective of their dog’s behavior and their judgement of possible effects or side effects of the treatment (that their dog may or may not be receiving). Including a placebo control group in a study that includes subjective measures of behavior (such as measuring the number and intensity of an aggressive response) is even more important since subjective scales are generally less reliable than objective measures.
Bottom line? The RCT that examined the effects of thyroid hormone replacement therapy on borderline or frankly hypothyroid dogs with owner-directed aggression showed a slight numerical reduction in aggression that was not statistically significant. As a result, the researchers concluded that thyroid replacement therapy could not be wholeheartedly recommended as a treatment for aggression in hypothyroid dogs and that additional studies of this type may be helpful to further examine this potential connection.
Personally, I think that this is also an excellent example of the progression of science from a set of initial case reports, followed by case-controlled studies, culminating in a randomized, controlled, clinical trial. An examination of the final study illustrates the enormous commitment of labor, time, and money that is required when conducting clinical trials as well as the importance of including placebos and double-blinding in scientific studies. Kudos to the investigators – not only for conducting what was clearly a very challenging clinical trial, but also for reporting informative negative results in a peer-reviewed journal.
Fatjó J, Stub C, Manteca X. Four cases of aggression and hypothyroidism in dogs. Veterinary Record 2002;151: 547-548.
- Dodds WJ, Aronson LP. Behavioral Changes Associated with Thyroid Dysfunction in Dogs. Proceedings 1999 American Holistic Veterinary Medical Association Annual Conference, pp. 80-82.
- Carter GC, Scott-Moncrieff JC, Luescher AU, Moore G. Serum total thyroxine and thyroid stimulating hormone concentrations in dogs with behavior problems. Journal of Veterinary Behavior 2009; 4:230-236.
- Radosta LA, Shofer FS, Reisner IF. Comparison of thyroid analytes in dogs aggressive to familiar people and in non-aggressive dogs. Veterinary Journal 2011;192:472-475.
- Dodman NH, Aronson L, Cottam N, Dodds JW. The effect of thyroid replacement in dogs with suboptimal thyroid function on owner-directed aggression: A randomized, double-blind, placebo-controlled clinical trial. Journal of Veterinary Behavior 2013;8:225-230.