This Letter to the Editor reveals broad disagreement between our empirical approach to animal pain science and research methodology and Dr. Mitek's beliefs. We feel that this difference has heavily influenced Dr. Mitek's opinion of our study. Through this response, we wish to answer the criticisms raised in as clear a fashion as possible, in particular as regards our statistical analyses. We also want to clarify the ethical environment of our research, and highlight the requirements of the publication process as well as the importance of supporting substantial concerns about research methods by reference to scientific literature.
First, we would like to emphasize that our study protocol was carefully reviewed and approved both by the funding agency's independent scientific committee, and by our Institutional Care and Use Committee (IACUC). In this and all our studies, we adhere strictly to the guidelines published by the Canadian Council on Animal Care, which are recognized internationally. As veterinarians, we also have a deep concern for animal health and welfare.
As was stated in the article's Introduction, critical anthropomorphism supports that pain is associated with bone marrow aspiration (BMA); however, scientific documentation of its presence in dogs was needed. Currently, recommended methods for performing canine BMA do not necessarily include sedation, and clinical practices do not routinely use either sedation or local anesthesia for sternal BMA. The objectives of our study were to compare the intensity of pain caused by 3 different BMA procedures in dogs, and to compare the results of different pain scales when used to assess this type of pain. The primary hypothesis we wished to verify was whether we could adequately assess the pain associated with these procedures by using currently accepted pain scales. As sedation is well known to introduce bias in evaluations using pain scales, a group of unsedated dogs was needed to assess the magnitude of the sedation effect. We limited the nonsedated procedure to the sternal site puncture, which is acknowledged to be less painful than the iliac site. This was accepted by the IACUC because of the high prevalence of this practice in veterinary medicine. We would never endorse the use of sedation or general anesthesia merely to reduce acute pain. In a recent update, the American Academy of Pediatrics (AAP) discusses the risks associated with their use in children, and states that their primary goal in using sedation or general anesthesia for diagnostic and therapeutic procedures is for controlling movement, to maintain patient safety and welfare by limiting physical discomfort, anxiety, pain, and injury due to movement. The AAP advocates the use of professional judgment to determine the appropriateness of using sedation or general anesthesia in a given patient. BMA is most often performed in investigations of systemic disease, or tumor staging, and canine patients subjected to this procedure are often old. Altogether, this emphasizes the need for careful consideration when using sedative analgesics or general anesthesia, and in some cases, both must be avoided (a situation that also occurs in pediatrics). Lastly, if a falsely elevated pain score was to lead to inappropriate use of an analgesic, the adverse effects could have severe and unwanted consequences.
We carefully considered our choice of deracoxib as a preemptive analgesic. We are well aware of its decreased bioavailability in fasted animals, but administration schedule was an experimental constraint and the product's monograph reports adequate bioavailability even in fasted dogs. Moreover, as stated in the introduction to our article, the pain associated with BMA arises during needle insertion and positioning through the periosteum, and aspiration. The amplification and duration of pain in this case is by nature due to inflammatory pain that occurs secondary to tissue trauma caused by the puncture. Characterization of pain by humans is not related to the mechanism of the particular pain, but is a description of the unpleasant sensory and emotional experience associated with the procedure. Only an understanding of basic pain mechanisms permits evaluation of the provided treatment's efficacy.
Our study protocol was overseen by the IACUC, which implies that a rescue analgesia protocol was in place. We did not describe it because rescue analgesia was not used due to the low pain levels assessed using the reference pain scale (SF-Glasgow). It is important to consider that the maximum value of this scale is 24, and its grading is not linear (ie, a score of 20 is not twice as painful as a score of 10). The unsedated animals in our study were not heavily restrained; they were held lightly in a sitting position, by one veterinarian (as they would be restrained for blood sample collection). Moreover, as indicated in our paper, dogs were accustomed to being handled by humans, having been acclimated to once daily handling for one week prior to the study, to decrease handling-related stress. We also monitored for signs of discomfort, fear, and for any pain-related behavior during the BMA.
With respect to Dr. Mitek's concerns regarding our statistical analyses, we would like to emphasize that 1) an expert in biostatistics was involved both in study design and in the analysis of data; 2) the statistics used were reviewed during the publishing process and no concerns were raised.
First, we did in fact test the data to confirm a Poisson distribution. Even though distribution type is difficult to verify with small numbers of data, pain scale data by nature have a Poisson distribution. Moreover, analyses of model residuals indicated that a Poisson distribution most accurately described our data.
Dr. Mitek did not indicate a reference to support her statement that mixed models analyses was an inappropriate method for our data. On the contrary, mixed models analyses for repeated data is best fit to handle repeated measures over time. This type of analysis 1) accounts for numerous correlations between data; 2) provides appropriate estimates of treatment effects and their standard errors; 3) does not require complete data from all subjects (ie, supports missing data).[6, 7]
We elected to use the median of the two observers’ scores as the outcome in our analyses because the data varied similarly over time for both observers, even though consistency between the two was not always excellent. The exploratory nature of our work did not justify independent presentation of the results for each observer.
Lastly, Dr. Mitek's final comment regarding our statistical analyses appears to indicate some confusion. The Poisson distribution is a discrete distribution that ranges from 0 to infinity and is especially suitable for counts. Its mean (E(x)) and variance (E(x-E(x))2 are indeed equal (λ > 0), but the skewness of a distribution refers to its asymmetry. A distribution is symmetric if its mean equals its median (skewness = 0; eg, normal distribution). However, for a Poisson distribution, which by definition is bounded only at its minimum (ie, negative counts are impossible), the median is ≈ (λ + 1/3 - 0.02 / λ), and the skewness is E(x-E(x))3/E(x-E(x))2, which obviously will be a positive value dependent on λ (=λ−1/2). Therefore, the conclusion that the presence of skewed data negates the possibility of a Poisson distribution is invalid.
In conclusion, the concerns expressed by Dr. Mitek over our work's validity are, in our opinion,without foundation. In a field with no preexisting record of empirical support for pain perception during BMA in dogs, the original findings of our study were BMA was demonstrated as moderately painful in dogs, and the SF-Glasgow pain scale displayed better psychometric properties, but an increased propensity to be affected by sedation, versus other measures. This has real pertinence to canine veterinary practice.