Abstract
Dear Editors,
We write with respect to the study published by Conti et al, ‘Evaluation of environment and a feline facial pheromone analogue on physiologic and behavioral measures in cats’. 1 The stated aim of this study was to assess ‘behavioral and physiologic stress parameters in cats placed in two environments: home and the veterinary hospital…’ and ‘whether a feline facial pheromone analogue (FFPA) had an effect on any of these measures in either environment’. The results state, ‘We found that exposure to FFPA did not reduce the effects of stress’, and the conclusion was made that ‘this study shows a lack of effect of FFPA on any of these parameters’. We have several concerns about this study, which we believe question the validity of the work and make the results and conclusions unsound.
A prerequisite for any study investigating the effect of any intervention on stress is that both the stressor and the ‘stress response’ are clearly defined. In this study, the stressor is the environment, and it seems reasonable to assume that the hospital should be more stressful than the home, although the results fail to show this: surprisingly, the systemic blood pressure (SBP) measured at the veterinary hospital was not higher than at home, and the behavioural measures of stress were found to be higher at home. This study used an interesting range of measures but failed to show a consistent effect of the environment. Additionally, an arbitrary range of parameters, which emerged as significant post-hoc, were used to define a ‘stress response’. These are non-specific signs and do not make a coherent body of measures to infer that the cats are indeed stressed; respiratory rate (RR) and heart rate (HR), for example, can be affected by body temperature, which was not measured in this study. Accordingly, it is not surprising that the use of a pheromone analogue, which has been shown to reduce stress in cats during a veterinary examination, 2 failed to have a reliable effect on these arbitrary measures. The fundamental flaw here is that the authors have assumed because these measures can indicate stress, that they necessarily do, which is not the case, and so their claims in relation to stress are invalid.
Several key pieces of information also seem to be missing from this paper, such as the study design itself, which apparently was a crossover. If so, the test for any sequence effect or period effect should be specified, to validate the study, which is not the case here. Nor is any information provided regarding the baseline values for these cats (to make sure the two groups were balanced, and that no cat had elevated blood pressure or HR values). It is not described how the cats were randomised (the procedure is unclear), and if the tests were always performed in the same order. The latter is known to lead possibly to a habituation effect over exposure to the tests, especially for such tests performed within a short period of time (limited washout).
There is a range of other errors in the study’s methodology. For example, the authors indicated that they used a power analysis to ensure the study was properly powered, but then failed to use the correct number of independent samples required. Indeed, although there were 30 subjects, the authors actually considered 120 cases, each cat being analysed separately in each context – home or hospital – and exposed to treatment or placebo: this failed to recognise that variables were dependent, as each cat actually acted as its own control subject. The choice of statistical test is of concern, too, as the Dunn test is not appropriate for such paired data (same cat in different environments). Also, given that the environment is the variable of interest, the use of multiple cats from the same home represents a form of pseudo-replication (ie, the test units should be the number of homes, not the number of cats). This greatly reduces the power of the study, and mandates that subjects are nested within home within a parametric ANOVA.
All this suggests, given the sample size, that sampling was not appropriate, there was an error in the measurement process or some other factor has skewed the results. Indeed, the results are not in accordance with either the authors’ own initially hypothesised effect (a 15 mmHg difference between the two groups for SBP and 35 beats/min for HR) or that of Quimby et al, 3 from whom they borrowed some of the methods. These factors further question the validity of the data reported here as measures of stress, but this issue is not considered by the authors.
In sum, the authors insufficiently described their study design to make it understandable, they failed to demonstrate that their measures were related to stress and that these cats were actually stressed, and they inappropriately analysed their results and therefore made invalid conclusions about the efficacy of FFPA on stress.