Lameness assessment using force plate gait analysis (FPGA) and owner assessment of chronic pain using the Canine Brief Pain Inventory (CBPI) are valid and reliable methods of evaluating canine osteoarthritis. There are no studies comparing these 2 outcome measures.
Evaluate the relationship between CBPI pain severity (PS) and interference (PI) scores with the vertical forces of FPGA as efficacy measures in canine osteoarthritis.
Sixty-eight client-owned dogs with osteoarthritis (50 hind limb and 18 forelimb).
Double-blind, randomized. Owners completed the CBPI, and dogs underwent FPGA on days 0 and 14. Dogs received carprofen or placebo on days 1 through 14. The change in PS and PI scores from day 0 to 14 were compared to the change in peak vertical force (PVF) and vertical impulse (VI).
PS and PI scores significantly decreased in carprofen- compared with placebo-treated dogs (P =.002 and P =.03, respectively). PVF and VI significantly increased in carprofen- compared with placebo-treated dogs (P =.006 and P =.02, respectively). There was no correlation or concordance between the PS or PI score changes and change in PVF or VI.
Conclusions and Clinical Importance
In these dogs with hind limb or forelimb osteoarthritis, owner assessment of chronic pain using the CBPI and assessment of lameness using FPGA detected significant improvement in dogs treated with carprofen. The lack of correlation or concordance between the change in owner scores and vertical forces suggests that owners were focused on behaviors other than lameness when making efficacy evaluations in their dogs.
The availability of quantitative measures of chronic pain that are valid and reliable in clinical patients is crucial for the development and testing of interventions (eg, drugs or surgical procedures) designed to decrease such pain. Studies designed to test the efficacy of interventions intended to decrease chronic pain in companion dogs with osteoarthritis have relied heavily on the assessment of lameness through the use of force plate gait analysis (FPGA).[1-8] Recently, increased attention has been given to the development and validation of outcome measures that can reliably quantify owners behavior-based assessment of chronic pain in their pets.[9-14]
The Canine Brief Pain Inventory (CBPI) was developed as an owner-completed questionnaire designed to quantify the owner's assessment of the severity and impact of chronic pain in their dogs with osteoarthritis. The CBPI contains 4 questions pertaining to the severity of pain, the responses for which can be used to calculate a mean value known as the pain severity (PS) score, and 6 questions pertaining to how the pain interferes with the dog's typical activities, the responses for which can be used to calculate a mean value that can provide a pain interference (PI) score. The CBPI was developed and validated by use of standard methods for the stepwise development of a health-assessment questionnaire.[15-17] Much like peak vertical force (PVF) and vertical impulse (VI), which are measured by FPGA and used to determine efficacy in canine osteoarthritis studies, the PS and PI scores from the CBPI detect changes in dogs with chronic pain that are the result of an intervention or disease progression.[1-11] To date, however, there are no studies investigating the relationship between these outcome measures.
In separate studies, dogs with osteoarthritis treated with NSAIDs have been shown to have increases in PVF and VI on FPGA, and decreases in PS and PI scores on the CBPI. Assuming owners use lameness as a primary behavioral assessment in the severity of pain associated with their dogs osteoarthritis, one would expect that if both outcome measures were used in the same study, as PVF and VI increase (ie, weight bearing lameness decreases), pain scores would decrease. The purpose of this study was to evaluate the relationships among PVF, VI, PS, and PI in the same cohort of dogs with osteoarthritis. The hypothesis was that within the context of a double-blinded randomized controlled clinical trial, PS and PI scores from the CBPI, as well as PVF and VI from FPGA, would detect substantial improvements in osteoarthritic dogs treated with an FDA-approved NSAID and there would be a significant negative correlation between PVF/VI and PS, as well as PVF/VI and PI scores. In addition, we compared the braking and propulsion forces that are measured with FPGA, but not typically used as outcome measures in efficacy studies, with the PS and PI scores of the CBPI to further elucidate the relationship between these 2 outcome measures.
Materials and Methods
For this study, we conducted further analyses of additional data from a previously published study that demonstrated the ability of the CBPI to detect a positive effect of carprofen over placebo. The primary study methods are repeated here as background for the new analyses evaluating the correlation and concordance between the CBPI pain scores and FPGA vertical forces described below. The study was approved by the Institutional Animal Care and Use Committee and dogs were enrolled in the study only after the owner gave written informed consent.
The study was conducted as a single-center, double-blinded, randomized, placebo-controlled clinical trial. All owners received a detailed written description of the protocol and provided written informed consent before their dogs were evaluated for inclusion in the study. The protocol was approved by an institutional animal care and use committee.
Inclusion criteria required that all dogs weighed ≥8 kg and had a medical history, clinical signs, physical examination findings, and radiographic findings consistent with osteoarthritis. Only dogs with newly diagnosed osteoarthritis and in which medical management had not yet been initiated, or in which a diagnosis of osteoarthritis had been made sometime in the past but the owners had opted not to medicate their dogs on a regular basis, were recruited. Dogs were excluded from the study if they had received NSAIDs within 2 weeks before the initial evaluation, or glucocorticoids or opioids within 4 weeks before the initial evaluation; had clinically important neurologic or orthopedic disease (other than osteoarthritis); had a history of coagulopathy, unexplained bleeding episodes, or hypersensitivity to NSAIDs; or had clinically relevant abnormalities detected during screening (CBC and serum biochemical profile). Thus, the study population consisted of medium- to large-breed dogs with no clinically important abnormalities (other than osteoarthritis) that were not currently receiving medications.
Estimation of Sample Size
At least 29 dogs would need to be included in each group (carprofen and control) to provide 80% power of detecting differences between the groups of ≥30% for changes in scores of the CBPI (SD, 40%; P =.05). To compensate for protocol deviations and losses to follow-up, it was determined that 70 dogs would be randomly assigned to each of the 2 groups (35 dogs/group). This is consistent with the sample size for another study in which investigators used a global owner assessment as well as FPGA to detect differences in outcome between carprofen- and placebo-treated dogs with osteoarthritis.
A simple randomization sequence with 2 potential treatment groups (carprofen or placebo) was generated at an off-site pharmacy. The sequence was concealed so that members of the research team were not aware of the group to which a dog would be allocated as it was evaluated in the screening process. Once screening was completed and a dog was considered eligible for inclusion in the study, a unique study number was assigned in sequence. The study number and body weight of each dog then were provided to the pharmacy. Pharmacy personnel matched the study number with their randomization sequence, formulated the appropriate treatment (carprofen or placebo), and packaged the pills into blister packs for use by the investigators.
Pharmacy personnel packaged carprofen or placebo for administration to each dog. Carprofen was administered at a dosage of 4.4 mg/kg. Pills containing the control substance were formulated to appear identical to the pills containing carprofen; packaging for both types of pills was identical. Thus, all study personnel and the owners were unaware of the treatment group to which each dog was assigned.
The CBPI was completed by an owner of each dog that was eligible for inclusion in the study. The 4 pain-severity questions were scored on a scale of 0 (no pain) to 10 (extreme pain). The responses for these questions were averaged to generate the PS score. The 6 PI questions (ie, how much the pain interfered with the dog's typical function) were scored on a scale of 0 (does not interfere) to 10 (completely interferes). The responses for these questions were averaged to generate the PI score. Also on day 0, dogs underwent gait analysis using a single piezoelectric force plate.1 Therefore, data for the right and left side were collected from separate passes across the plate. Dogs were acclimated to the force plate before data collection. Five valid trials for each dog and each limb were recorded. A trial was considered valid when the ipsilateral thoracic and pelvic limbs only fully contacted the force platform, and when velocity was controlled at 1.6–1.9 m/s with acceleration controlled at ±0.5 m/s2 measured by 3 photocells.2 Ground reaction forces were sampled at 200 Hz and recorded using software.3 All forces were normalized to body weight in kilograms. Data from the 5 valid trials for each limb were averaged to obtain a mean value for each gait parameter at each time point. Only the results for the most severely affected limb were included in the analysis.
Days 1 through 14
Owners administered the prescribed pills once daily. Three additional doses were included in the event that an owner encountered a delay in returning the dog for a follow-up appointment.
The same owner who completed the CBPI on day 0 completed a 2nd CBPI, and the dogs underwent FPGA again as described above. After collection of data, a 2-week supply of carprofen 4.4 mg/kg was dispensed to each owner for administration to their dogs so that all owners could determine the potential benefits of NSAIDs and could be counseled on continued treatment and follow-up evaluations with their veterinarians.
Descriptive statistics were calculated. Normally distributed continuous variables were expressed as mean and standard deviation, whereas nonnormally distributed continuous variables were expressed as median values and ranges. Categorical data were expressed as frequencies. To determine the efficacy of the intervention using the 2 outcome measures (FPGA and CBPI) in the context of the randomized, controlled clinical trial, the Mann-Whitney U-test was used to compare the change in PS and PI scores between the placebo- and carprofen-treated dogs, whereas the Student's t-test was used to compare the change in PVF and VI between the 2 groups. To determine the relationship between the vertical, braking, and propulsion forces from the FPGA and the PS and PI scores from the CBPI, Spearman rank correlations were used. In addition, the Somers D statistic, which is a more robust measure of whether 2 instruments are measuring the same thing, also was calculated. Two-tailed assessments were used and P values <.05 were considered significant. All analyses were performed by use of a statistical program.4
Seventy dogs meeting the inclusion and exclusion criteria were randomized. One dog in the carprofen group ruptured its cranial cruciate ligament during the 1st week of treatment and was referred to the orthopedic service at our university veterinary medical teaching hospital for appropriate care. Data on an additional dog in the carprofen group were lost because of photocell malfunction. Therefore, 68 dogs were included in the analysis, 35 in the placebo group, and 33 in the carprofen group. Twenty-five dogs in each group (placebo and carprofen) were most severely affected in a hind limb. Ten dogs in the placebo group and 8 dogs in the carprofen group were most severely affected in a forelimb. On the basis of information from the owners and review of the blister packs, all 68 dogs received the prescribed treatment for 14–16 days, all owners completed the CBPI on day 0 and the last day of treatment (days 14–16), and all dogs had complete force plate data collected on day 0 and the last day of treatment. Baseline characteristics of the 68 dogs are presented in Table 1.
Table 1. Baseline characteristics of 68 dogs with osteoarthritis of the forelimb or hind limb enrolled in a double-blind, randomized, controlled clinical trial of placebo versus carprofen
Placebo n = 35
Carprofen n = 33
1 or 2 of 10 other purebreds.
1 or 2 of 7 other purebreds.
12 mixed7 Rottweilers3 Doberman Pinschers10 other purebredsa
10 mixed11 Rottweilers3 Labrador Retrievers7 other purebredsb
Weight in kg median (range)
Age in years median (range)
11 Female, 24 Males
21 Females, 12 Males
Pain Severity Score median (range)
Pain Interference Score median (range)
Peak Vertical Force (% BW) of the most clinically affected limb mean ± sd
Ability of Vertical Forces and Pain Scores to Detect Response to NSAID Treatment
Dogs with forelimb or hind limb osteoarthritis that received carprofen had significant improvements in PS score (P =.002), PI score (P =.03), PVF (P =.006), and VI (P =.02) compared with the placebo-treated dogs (Table 2).
Table 2. Comparison of response to treatment outcomes using the owner-completed Canine Brief Pain Inventory (CBPI) and force plate gait analysis for assessment of treatment efficacy in a double-blind, randomized, controlled trial of carprofen versus placebo in dogs with osteoarthritis of the forelimb or hind limb
Change in CBPI Pain Severity Score
Change in CBPI Pain Interference Score
Change in Peak Vertical Force (%BW)
Change in Vertical Impulse (%BW)
Mean + sd
Mean + sd
Change was calculated between Day 0 (baseline) and Day 14 after randomization and change in placebo was compared with change in carprofen-treated dogs using the Mann-Whitney U-test*, ^ or the Student's t-test#,$.
Placebo-treated dogs (n = 35)
−1.3 ± 7.3
−0.09 ± 0.74
Carprofen-treated dogs (n = 33)
3.2 ± 0.8
0.32 + 0.61
Correlation and Concordance between FPGA Forces and CBPI Pain Scores
The relationship between the changes in PS and PI scores compared to the changes in forces on FPGA are presented in Figures 1-6. There was no significant correlation or concordance between the CBPI and FPGA forces in either the front- or hind limb with 1 exception. There was a mild correlation (r = 0.36, P =.01) between the peak propulsion force and the PI score of the CBPI for the hind limb. Results of the correlation and concordance analyses are presented in Tables 3 and 4.
Table 3. Correlation and concordance of forces measured on force plate gate analysis and Canine Brief Pain Inventory pain severity score
Front Leg (n = 18)
Hind Leg (n = 50)
BW,body weight; r, Spearman rho.
Peak vertical force (%BW)
r = −0.02 P =.93
r = 0.12 P =.40
Vertical impulse (%BW x s)
r = 0.04 P =.86
r = 0.10 P =.48
Peak braking force (%BW)
r = 0.18 P =.49
r = 0.15 P =.30
Braking impulse (%BW x s)
r = 0.20 P =.45
r = 0.17 P =.25
Peak propulsion force (%BW)
r = −0.34 P =.19
r = 0.15 P =.30
Propulsion impulse (%BW x s)
r = −0.20 P =.46
r = 0.04 P =.80
Table 4. Correlation and concordance of forces measured on force plate gate analysis and Canine Brief Pain Inventory pain interference score
Front Leg (n = 18)
Hind Leg (n = 50)
BW, body weight; r, Spearman rho.
Peak vertical force (%BW)
r = 0.30 P =.23
r = −0.05 P =.76
Vertical impulse (%BW x s)
r = −0.39 P =.11
r = 0.03 P =.83
Peak braking force (%BW)
r = 0.19P =.47
r = 0.22 P =.13
Braking impulse (%BW x s)
r = 0.26 P =.33
r = 0.23 P =.10
Peak propulsion force (%BW)
r = −0.29 P =.28
r = 0.36 P =.01
Propulsion impulse (%BW x s)
r = −0.16 P =.55
r = 0.28 P =.05
The CBPI detected changes in pain scores and FPGA detected changes in vertical forces that would be expected in the context of a randomized, controlled trial in which there was an active treatment (an NSAID) and a placebo. This was not unexpected in that both outcome measures have been proven valid and reliable in dogs with osteoarthritis, and both vertical forces and CBPI pain scores have been used as efficacy measures in clinical trials. What was unexpected, however, was the fact that there was no correlation or concordance between the pain scores and vertical force measures in this cohort of dogs with forelimb and hind limb osteoarthritis. This observation suggests that although both vertical forces and owner assessment using the CBPI are useful and appropriate tools to prove intervention efficacy in dogs with osteoarthritis, they are quantifying different things.
FPGA is viewed as the gold standard for the evaluation of lameness. Peak force (maximal force applied during stance phase) in the vertical axis is the most commonly used variable from gait analysis to assess degree of lameness. Although subjective visual analog, categorical, and numerical rating scales have been described for the visual assessment of lameness by owners or veterinarians, few have been adequately assessed for validity and reliability, and none have been shown to accurately reflect FPGA unless the lameness is severe. Although people may not be able to judge lameness with the precision and reliability of FPGA, we assumed that lameness would still play an important role in the owners assessment of chronic pain in their dogs with osteoarthritis, such that decreases in lameness (ie, increases in vertical forces) would be strongly associated with decreases in pain scores, but this was not the case.
The CBPI allows owners to reliably quantify the behavior-based assessment of chronic pain in their dogs in 2 ways. The PS score has a broad scope that allows owners to assimilate all of the pain behaviors that they believe to be important in their dog in a 0–10 rating of no pain to extreme pain, whereas the PI score relies on assessment of specific behaviors such as the dog's ability to climb up. Because the PI scoring questions do not directly ask for an assessment of lameness as 1 of the behaviors, it may be considered less likely to have correlated with vertical forces than the PS score. It was hypothesized, however, that many of the activities described in the PI scoring system, such as the ability to walk, could reasonably be expected to be affected by lameness, and therefore indirectly have a significant negative correlation with PVF. In the end, neither score change correlated with lameness change assessed by PVF. There could be a number of explanations for this lack of correlation or concordance.
The force plate may be detecting changes in levels of lameness that are not detectable by the human eye. There are no studies describing what kind of change in vertical forces must occur before one can visually identify a difference in the dog's gait. Therefore, there may have been no correlation or concordance between the change in vertical forces and the change in pain scores because the change in lameness was too small to be detectable to the owner's eye. Alternatively, owners may very well have noticed a change in lameness in their dog, but lameness in and of itself was not the overriding behavior that owners considered when assessing chronic pain in their dogs. There is some evidence that the latter explanation may be the case. In several studies, focus groups of owners of dogs with osteoarthritis describe many behaviors associated with their dogs condition that could be related to lameness such as “mobility” or “activity,” but owners do not describe discrete lameness in 1 leg as a specific behavior in their dog's condition.[10, 13, 14] These studies show that owners are much more focused on the dog as a whole, and its ability to perform its activities of daily living in its home environment, as opposed to increased or decreased use of a single limb at a walk or trot.
In focus groups, owners of dogs with osteoarthritis often complain about how their dog's pain interferes with its “ability to walk.”[10, 11] One might assume then that if the mechanics of the dog's gait are improved, owners will consider their dog's ability to walk improved. However, when the owners elaborate on their dog's “ability to walk,” they say that their dog “used to walk with them for an hour every night, but now just sits down after 3 blocks and can't go on.”  In their eyes, the dogs “ability to walk” is improved if it is able to resume its nightly hourly walks with them after treatment. Even if the dog's lameness is visibly improved, if its ability to go on its nightly walks has not improved, there is a good chance the owner will not consider the dog's “ability to walk” to be better. Similar arguments can be made for the other questions such as “ability to run” or “ability to climb up”. If a dog bears more weight on its osteoarthritic limb after treatment, but it is still reluctant to climb the stairs, the owner may not consider the dog to be better after treatment. The owners assessment of improvement in their dogs with osteoarthritis appears to have less to do with the mechanics associated with a single limb than with the dog's ability to perform its activities of daily living in its routine environment. It would be ideal to repeat the study described here using 150 dogs, 75 with a front limb most severely affected and 75 with a hind limb most severely affected. This would allow enough power to perform limb-specific efficacy analyses (ie, comparing response in placebo- versus carprofen-treated dogs) using the 2 outcome measures and further evaluate the pattern of associations between them. The pattern of association may be different when looking at front and hind limbs individually as opposed to as a group.
In addition to the efficacy analysis using the 2 outcome measures, the overall relationship between all forces measured in FPGA and the PS and PI scores was evaluated using correlation and concordance analyses. There was no significant correlation or concordance with FPGA forces and the PS score. One force, the peak propulsion force, had mild correlation and concordance with the PI score in the hind limb. The mild association could be explained by the fact that the rear limb provides the propulsion for the next step, and changes in this force could represent a gait abnormality that affects the activities described in the PI scoring system, such as the ability to run or the ability to jump up. Additional significant correlations with the PI might have been identified if the sample size for each leg were larger. Based on the rho values identified in this study, however, most were well below 0.3, and any correlation identified to be statistically significant would likely be mild. Based on the rho and P-values in the comparison of these gait parameters with the PS score, it is unlikely that a reasonable increase in sample size would uncover any significant correlations, and likely any identified would be mild, because the rho values generally remained <0.3.
If FPGA and owner assessment of chronic pain using the CBPI are quantifying different things (pain's impact on lameness versus pain's impact on activities of daily living), their preferred use as an outcome measure for a given study may depend on what question is being asked. If the study question is “Does the intervention improve lameness in dogs with osteoarthritis?”, FPGA would be the tool of choice. If the study question is “Does the intervention decrease chronic pain impact on activities of daily living in dogs with osteoarthritis?”, the CBPI might be more appropriate choice. In some cases, it may be ideal to ask both questions about the same intervention. However, sometimes the ideal is not feasible and the choice of question (and therefore the outcome measure used) may be dependent on the logistics and limitations of the outcome measure itself.
Under correctly controlled conditions, FPGA has the advantage of being an objective measure that is very reliable and sensitive to change in the dog. Logistically, however, it requires expensive, specialized equipment, and the data collection can be time consuming, requiring substantial expertise and effort from support personnel. In addition, gait analysis has the limitation of only evaluating an animal at 1 specific time point, outside of its normal environment, which may impact the generalizability of the results to the animal's waxing and waning condition in its home environment.
Use of the CBPI has the advantage of reliably quantifying the owners assessment of clinically relevant chronic pain-related behaviors with the dog in its routine environment. In addition, it asks the owner for an assessment over an extended period of time (ie, 7 days), as well as how their dog is doing “right now”, which allows the owner to take into account the waxing and waning aspects of the disease. The owner assessment is, however, a subjective outcome measure. Therefore, the same owner must complete the questionnaire throughout the entirety of a study, and it is best suited for studies in which there is a control group and blinding to treatment group is feasible.
In conclusion, in this cohort of dogs with osteoarthritis of the forelimb or hind limb, the CBPI detected changes in pain scores and FPGA detected changes in vertical forces that would be expected in the context of a randomized controlled trial in which there was an active treatment (an NSAID) and placebo. There was, however, no correlation or concordance between the pain scores and the vertical forces. This finding suggests that although both vertical forces using FPGA and owner assessments using the CBPI are useful and appropriate tools to prove intervention efficacy in dogs with osteoarthritis, they are quantifying different aspects of chronic pain in these animals. Whereas FPGA quantifies pain's impact on lameness, the CBPI appears to quantify pain's impact on activities of daily living. Ultimately, the choice of outcome assessment tool to be used in a given study depends on the study question to be answered and weighing the advantages and disadvantages of each valid and reliable outcome assessment instrument that is available.
We thank Molly Love and the staff in Veterinary Clinical Investigation Center for their assistance with this study. This was supported by the National Institutes of Health (grant No. 1-K08-DA-017720-02).Conflict of Interest: Authors disclose no conflict of interest.