Refinement and initial validation of a

multidimensional composite scale for use in

assessing acute postoperative pain in cats
Juliana Tabarelli Brondani, DVM, PhD; Stelio Pacca Loureiro Luna, DVM, PhD;
Carlos Roberto Padovani, BMath, PhD

Objective—To refine and test construct validity and reliability of a composite pain scale for
use in assessing acute postoperative pain in cats undergoing ovariohysterectomy.
Sample Population—40 cats that underwent ovariohysterectomy in a previous study.
Procedures—In a previous randomized, double-blind, placebo-controlled study, a compos-
ite pain scale was developed to assess postoperative pain in cats that received a placebo
or an analgesic (tramadol, vedaprofen, or tramadol-vedaprofen combination). In the present
study, the scale was refined via item analysis (distribution frequency and occurrence), a
nonparametric ANOVA, and item-to-total score correlation. Construct validity was assessed
via factor analysis and known-groups discrimination, and reliability was measured by as-
sessing internal consistency.
Results—Respiratory rate and respiratory pattern were rejected after item analysis. Fac-
tor analysis resulted in 5 dimensions (F1 [psychomotor change], posture, comfort, activity,
mental status, and miscellaneous behaviors; F2 [protection of wound area], reaction to
palpation of the surgical wound and palpation of the abdomen and flank; F3 [physiologic
variables], systolic arterial blood pressure and appetite; F4 [vocal expression of pain], vocal-
ization; and F5 [heart rate]). Internal consistency was excellent for the overall scale and for
F1, F2, and F3; very good for F4; and unacceptable for F5. Except for heart rate, the identi-
fied factors and scale total score could be used to detect differences between the analgesic
and placebo groups and differences among the analgesic treatments.
Conclusions and Clinical Relevance—Results provided initial evidence of construct valid-
ity and reliability of a multidimensional composite tool for use in assessing acute postopera-
tive pain in cats undergoing ovariohysterectomy. (Am J Vet Res 2011;72:174–183)

T reatment of pain in cats historically has been ne-

glected.1–5 However, there is evidence that this sce-
nario is changing.6 One of the primary causes for this
HR Heart rate
Abbreviations
RR Respiratory rate
problem may be the inherent difficulty in identifying SABP Systolic arterial blood pressure
and quantifying the intensity of pain in a cat. The lack
of available and validated pain assessment tools for this
species may result in clinicians providing inadequate
pain relief. Multiple instruments have been used for the as-
The cornerstone for adequate pain treatment is the sessment of pain in small animals; however, few have
availability of accurate pain assessment methods. Re- been validated for use in dogs,8–10 and to the authors’
liability and validity for pain assessment are vital for knowledge, none of these pain-scoring systems has
use in recognizing pain, quantifying pain intensity, and been tested for validity and reliability in cats. The vi-
evaluating treatment effectiveness.7 Therefore, it is nec- sual analogue scale is frequently used for pain assess-
essary to develop pain instruments with established ment in cats11–13; nevertheless, reliability and validity
reliability and validity for use in clinical practice and of this scale have not been assessed in this species. A
research. major difficulty with visual analogue scales is that they
are subject to several observer biases in the absence of
specific or objective criteria for pain rating.14,15
Received October 5, 2009. To investigate the validity and reliability of an
Accepted January 13, 2010. instrument requires knowledge of the various com-
From the Department of Veterinary Surgery and Anesthesiology, ponents and methods by which they may be tested.16
School of Veterinary Medicine and Animal Science, and the Validity is the effectiveness with which a test or scale
Department of Biostatistics, Institute of Biosciences, UNESP measures the property being investigated.10 However,
Universidade Estadual Paulista, 18618-970 Botucatu, SP, Brazil.
Supported by FAPESP – São Paulo Research Foundation.
it should be clear that validation is not performed for
Presented in part as a poster at the 10th World Congress of Veterinary the measuring instrument but instead for the measur-
Anaesthesia, Glasgow, Scotland, September 2009. ing instrument in relation to the purpose for which it
Address correspondence to Dr. Brondani ([email protected]). is used.17

174 AJVR, Vol 72, No. 2, February 2011

09-10-0348r.indd 174 1/20/2011 12:53:47 PM

 Validity may be estimated via several methods. Var- ent consent had been obtained for use of the cats in
ious terms have been used to describe the assessment of that study,23 and that study had been approved by an
validity, but it is traditionally subdivided into content, institutional animal research ethics committee (No.
construct, and criterion validity.10,18,19 However, it has 123/2003).
been recently proposed20 that validation is a process of
hypothesis testing and should not be constrained by the Study procedures—The present study had 2 phas-
aforementioned concepts. es. Phase 1 consisted of refinement of the composite
Content validity is the judgment as to whether an scale previously developed in the aforementioned
instrument samples all of the relevant or important con- study23 for use in assessing acute postoperative pain in
tent or domains.20 Criterion validity considers whether cats undergoing ovariohysterectomy. Phase 2 involved
scores on the instrument agree with a definitive crite- the initial assessment of construct validity and reliabil-
rion-referenced measurement of the same theme21; in ity of the refined composite pain scale.
other words, it compares the test with other measures
or outcomes already validated. Criterion validity may Development of composite pain scale—The com-
be divided into concurrent and predictive validity, de- posite pain scale for use in assessing acute postopera-
pending on whether the criterion refers to a current or tive pain in cats was developed in a previous study23
future state.21 It should be mentioned that both content for use in evaluating the analgesic efficacy of tramadol,
and criterion validity have limitations: the former be- vedaprofen, or the tramadol-vedaprofen combination
cause it is only based on judgment of experts regard- in 40 cats undergoing elective ovariohysterectomy. The
ing the content of the items, and the latter because of pool for the scale items was selected from instruments
a lack of relevant criterion variables for comparison. available in the literature and specifically designed for
On the other hand, construct validity has generalized use in the assessment of acute pain in dogs,8,9 cats,24,25
applicability because it focuses on the extent to which or both.26 After the initial selection of items from the
a measure performs in accordance with theoretical available scales, some items were modified and provid-
expectations.17 ed with better descriptions on the basis of the clinical
Construct validity refers to how well an instru- experience of the authors and pain indicators in cats
ment actually measures the construct it is intended to reported in the literature.27
measure.22 Construct validity is often provided in the The first version of the scale for the previous study23
form of factor analyses and evidence for the ability of consisted of 7 physiologic (HR, RR, respiratory pattern,
the measure to discriminate among groups.21 Construct SABP, rectal temperature, salivation, and pupil size)
validity can be evaluated in numerous ways because it and 10 behavioral (facial expression, posture, comfort,
is possible to perform several predictions on the basis of movement, reaction to palpation of the surgical wound,
a theory or construct.20 reaction to palpation of the abdomen and flank, vocal-
Reliability concerns the degree to which the results ization, behavior, mental status, and appetite) items.
are consistent across repeated measurements.17 In other Ordinal items were ranked numerically on the basis
words, reliability assesses whether a test is measuring of information obtained from the literature8,9,24–26 and
a factor in a reproducible manner. Reliability can be the investigators’ clinical experience, with 0 indicating
based on a single administration of the measure (inter- normal or no change and the highest value indicating
nal consistency) or by examining the reproducibility of the most marked change. Nominal items were assigned
a measure administered > 1 time (stability).20 a value of 0 or 1 for absence and presence of change,
The purpose of the study reported here was to re- respectively. Observers were instructed to choose the
fine and test construct validity and reliability of a com- single value within each item that best fit the condition
posite pain scale used for assessing acute postoperative of a cat. Scores for each item were summed to calculate
pain in cats undergoing ovariohysterectomy. Construct a total score.
validity was investigated via factor analysis and known- The first version of the scale was used in the pre-
groups discrimination. Exploratory factor analysis was vious study23 to assess acute postoperative pain in 40
used because no a priori hypothesis about the corre- cats that underwent ovariohysterectomy. Cats were al-
lations among the variables could be proposed. The located into 4 groups (10 cats/group) and were treated
hypothesis supporting the known-groups discrimina- with analgesics (tramadol, vedaprofen, or tramadol-
tion was that the factor scores and the total score for vedaprofen combination) or a placebo.23 Anesthesia
the scale could be used to distinguish cats with pain was induced with propofol and maintained with isoflu-
(placebo group) from cats treated with analgesics or to rane to perform the surgery. One hour after the end of
discriminate among the various analgesic treatments. surgery, all cats were completely recovered from anes-
Reliability was assessed via internal consistency of the thesia. One investigator who was unaware of the treat-
scale. It was assumed that although the scale was multi- ment group for each cat administered the drugs and as-
dimensional, the total score could be used if there was a sessed postoperative pain by use of the aforementioned
correlation between the factors and acceptable internal composite pain scale at regular intervals for 7 days.
consistency of the overall scale.
Refinement of the composite pain scale (phase
Materials and Methods 1)—On the basis of results of the previous study,23 the
first refinement of the scale in the present study was per-
Sample population—Data for 40 client-owned formed to modify the structure and content. Items that
cats that underwent ovariohysterectomy in a previ- did not have variability (salivation and facial expres-
ous study23 were used in the study reported here. Cli- sion), that had interference from the analgesics (rec-

AJVR, Vol 72, No. 2, February 2011 175

09-10-0348r.indd 175 1/20/2011 12:53:47 PM

 tal temperature and pupil size), or that were dependent with vedaprofen (with or without tramadol) was in-
on other variables (behavior) were removed from the vestigated by use of nonparametric repeated-measures
composite pain scale. Because squinted eyes was the ANOVA for 2 factors (group and time). The ability of each
only facial expression modified by pain, it was includ- item to differentiate the time of assessment (first hour af-
ed in the item miscellaneous behaviors; miscellaneous ter the end of surgery and the second day after surgery)
behaviors also included other behavioral changes (eg, tail was also investigated. Significance was set at values of P
wagging, contracting and extending the pelvic limbs, con- < 0.05.
tracting the abdominal muscles, and licking or chewing
on the wound area) observed during the study. The item Item-to-total score correlation coefficient
identified as movement was replaced by activity, and cat- Item-to-total score correlations were evaluated by
egories were discriminated better. The item mental status calculating the nonparametric Spearman rank correlation
was changed from a nominal item to an ordinal item. coefficients between each item and the sum of all items
After the first refinement, the instrument was for the scale. This correlation coefficient assessed the rele-
transformed to an ordinal scale, except for the nominal vance of each item in the instrument and identified items
item respiratory pattern. This transformation increased that contributed strongly to the score for the overall scale.
the possibilities for the use of data in statistical analysis. Items with a correlation coefficient < 0.4 were rejected.
The refined data were resubmitted to statistical analysis
to improve application of the data. A new detailed con- Validity and reliability testing (phase 2)—Con-
tent analysis was then performed. Some items and cat- struct validity and reliability testing of the refined com-
egories of the scale were specified better and clarified. posite pain scale were evaluated.

Item analysis Construct validity of factor structure

A series of statistical analyses were conducted to Factor analysis is a mathematical procedure used
investigate the relevance of the remaining items of the to reduce sets of interrelated variables to a small num-
scale. The statistical analysis was based on data col- ber of underlying factors that usually cannot be directly
lected during the immediate postoperative period (first observed. Factor analysis is able to determine the num-
hour after the end of surgery) and at the second day ber of factors (dimensions or domains) that are repre-
after surgery (mean for evaluations conducted 24, 28, sented by different variables.29 Factor analysis provides
and 32 hours after surgery) for the previous study.23 All a determination of the dimensionality of the scale.19
of the cats were completely recovered from anesthesia Principal-components analysis with varimax rota-
by 1 hour after the end of surgery. The analysis was tion was performed to examine the underlying factor
performed for these 2 time frames because the behav- structure. The varimax rotation method allowed us to
ioral and physiologic changes were modified during the obtain factors with the greatest potential of interpret-
postoperative period by a reduction in pain intensity ability. The identification of factors was based on the
or as a result of adaptation to pain. Therefore, relevant Kaiser criterion,30 which suggests retaining all com-
items were discriminated to assess pain in the immedi- ponents with an eigenvalue > 1. The factor structure
ate postoperative period and at the second day after sur- was determined by attributing any items that had fac-
gery. Items that were rejected in ≥ 3 statistical analyses tor loadings > 0.5 and communality > 0.7 to a specific
were removed from the composite pain scale. factor. Factor-to-total score and interfactor correlations
were analyzed by use of the Spearman rank correlation.
Frequency distribution
Variation of the response within each item and Construct validity of known-groups discrimination
during the interval between the first hour after the end Known-groups discrimination involves determin-
of surgery and the second day after surgery was veri- ing whether there can be significant discrimination of
fied via the frequency distribution of the scores of each pain scores among groups of subjects with specific char-
item. Items that had low variability and that did not acteristics who are expected to differ in their amount of
change over time were rejected. pain. A nonparametric ANOVA for 2 factors (group and
factor) was performed to enable us to detect whether
Occurrence distribution the factors identified in the factor analysis were able to
Data were initially transformed for dichotomous distinguish pain of differing severity. It was determined
items, where 0 was absence of pain and any value other whether the factors could discriminate cats with pain
than 0 was considered presence of pain. Items that could (placebo treatment) from cats treated with analgesics
not be used to identify the presence of pain and distin- (tramadol, vedaprofen, or tramadol-vedaprofen combi-
guish groups that received analgesic treatments (trama- nation) or distinguish among cats receiving the various
dol, vedaprofen, or tramadol-vedaprofen combination) analgesic treatments. Additionally, because the scale
from the placebo group were rejected. The Goodman can be used with a total score instead of a factor score,
test28 was used to analyze the frequency and occur- the Kruskal-Wallis test followed by use of the Dunn
rence distributions. Values of P < 0.05 were considered method was applied to investigate whether the total
significant. score was able to distinguish between or among groups.

Nonparametric ANOVA Reliability of internal consistency

The ability of an item to identify treatment with Internal consistency refers to the degree of in-
tramadol (with or without vedaprofen) or treatment terrelation among the items in the measure.31 It was

176 AJVR, Vol 72, No. 2, February 2011

09-10-0348r.indd 176 1/20/2011 12:53:48 PM

 assessed by calculating the Cronbach α coefficient, day after surgery. Respiratory rate was rejected at the
which measures reliability in terms of the ratio of first hour after the end of surgery and at the second day
true score variance to observed score variance.32 Al- after surgery because it had low variability and did not
though a high degree of internal consistency is as- change over time.
sociated with unidimensionality, a set of items can
be somewhat interrelated and multidimensional.33 Occurrence distribution
Because the factor analysis identified > 1 dimension, At the first hour after the end of surgery, RR and
the internal consistency of each factor was initially vocalization could be used to discriminate the placebo
calculated by use of the following modified equation group from the analgesic-treated group, but it could
for the Cronbach α coefficient: not be used to identify the presence of pain. Reaction

Σ
κ
to palpation of the surgical wound could be used to
α F = (k/[k – 1])•(1 – [{ •Ci2•si2}/s2TF]) identify pain but not to distinguish the placebo group
i=1 from the analgesic-treated group. Respiratory pattern
was the only item rejected at the first hour after the
where k is the number of items of the factor (although end of surgery because it could not be used to identify
each factor was represented only by items with factor the presence of pain nor distinguish the placebo group
loading > 0.5, the factor included all items of the scale), from the analgesic-treated group.
Ci2 is the square of the factor loading of i items, si2 is the
variance of i items, and s2TF is the total variance of the fac- Table 1—Results of item analysis of a composite pain scale
used for assessing acute postoperative pain in cats undergoing
tor. Factors that did not have minimally acceptable inter- ovariohysterectomy for evaluations conducted at the first hour
nal consistency were deleted from the scale, and the classi- after the end of surgery and at the second day after surgery in a
cal raw Cronbach α coefficient32 based on item correlation previous study.23
was calculated for the overall scale. Values for the α coeffi- First hour after Second day after
cient were categorized as follows: 0.60 to 0.64, minimally Item the end of surgery surgery
acceptable; 0.65 to 0.69, acceptable; 0.70 to 0.74, good;
HR CD B C D*
0.75 to 0.80, very good; and > 0.80, excellent.34 RR A C D* A B C D*
Respiratory pattern B C D* B C D*
Results SABP — B C D*
Posture — BD
Refinement of the composite pain scale (phase Comfort — B C D*
Activity — —
1)—Results were determined for phase 1.
Reaction to palpation of the — —
surgical wound
Item analysis Reaction to palpation of the — C
Items rejected in ≥ 3 of the statistical analyses were abdomen and flank
deleted from the scale (Table 1). Vocalization C B C D*
Mental status — C
Appetite C B C D*
Frequency distribution Miscellaneous behaviors — B C D*
Reaction to palpation of the surgical wound did Exclusion criteria were as follows: A = Frequency distribution,
not change over time. Low variability was observed for B = Occurrence distribution, C = Nonparametric ANOVA, and D =
respiratory pattern, vocalization, and miscellaneous be- Item-to-total score correlation coefficient.
haviors at the first hour after the end of surgery and for *Variable was excluded from the final version of the composite
pain scale because it was rejected in $ 3 statistical analyses.
respiratory pattern, SABP, posture, comfort, activity, vo- — = Not rejected in any statistical analysis.
calization, and miscellaneous behaviors at the second
Table 2—Spearman rank correlation coefficient for item score to total score of a composite pain scale for
use in assessing acute postoperative pain in cats undergoing ovariohysterectomy for evaluations conducted
at the first hour after the end of surgery and at the second day after surgery in a previous study.23

First hour after Second day

the end of surgery after surgery
Item r P value r P value
HR 0.280* 0.080 0.245* 0.127
RR 0.346* 0.029 0.238* 0.138
Respiratory pattern 0.331* 0.036 –0.224* 0.163
SABP 0.684 , 0.001 0.301* 0.058
Posture 0.915 , 0.001 0.362* 0.021
Comfort 0.885 , 0.001 0.342* 0.031
Activity 0.942 , 0.001 0.546 ,0.001
Reaction to palpation of the surgical wound 0.834 , 0.001 0.892 , 0.001
Reaction to palpation of the abdomen and flank 0.700 , 0.001 0.788 , 0.001
Vocalization 0.483 0.001 0.267* 0.095
Mental status 0.849 , 0.001 0.554 , 0.001
Appetite 0.607 , 0.001 0.208* 0.198
Miscellaneous behaviors 0.709 , 0.001 0.074* 0.644

*Item was rejected on the basis of r , 0.4.

AJVR, Vol 72, No. 2, February 2011 177

09-10-0348r.indd 177 1/20/2011 12:53:48 PM

 Table 3—Results of analysis via a rotated component matrix by use of the principal component
extraction method and varimax rotation with the Kaiser criterion of a composite pain scale used for
assessing acute postoperative pain in cats undergoing ovariohysterectomy in a previous study.23

Factor loading*
Item Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Communality†
HR 0.056 0.005 0.021 –0.033 0.968‡ 0.942
SABP 0.179 –0.472 0.574‡ 0.333 0.278 0.773
Posture 0.848‡ –0.282 0.385 0.114 0.046 0.962
Comfort 0.860‡ –0.258 0.349 0.077 –0.008 0.934
Activity 0.815‡ –0.321 0.390 0.193 0.049 0.959
Reaction to palpation of 0.577 –0.629‡ 0.063 0.317 0.074 0.839
the surgical wound

Reaction to palpation of the 0.344 –0.868‡ 0.176 0.028 –0.056 0.906

abdomen and flank
Vocalization 0.221 –0.092 0.050 0.916‡ –0.075 0.905
Mental status 0.832‡ –0.219 0.153 0.279 0.041 0.842
Appetite 0.395 –0.096 0.837‡ –0.085 –0.064 0.877
Miscellaneous behaviors 0.653‡ –0.193 –0.208 0.507 0.296 0.851

Eigenvalue 3.932 1.723 1.556 1.450 1.126 NA
Variance (%) 35.8 15.7 14.1 13.2 10.2 NA
Accumulated variance (%) 35.8 51.5 65.6 78.8 89.0 NA

*Factor loading represents correlations between the variables and factors. †Communality represents the
proportion of the variance for each item that can be explained by the factor. ‡Item was substantially loaded
for the factor.
NA = Not applicable.

At the second day after surgery, activ-

ity, reaction to palpation of the abdomen
and flank, and mental status could be used
to discriminate the placebo group from the
analgesic-treated group; however, it could
not be used to identify pain. Reaction to
palpation of the surgical wound could be
used to identify the presence of pain and
to distinguish the placebo group from the
analgesic-treated group. Heart rate, RR, re-
spiratory pattern, SABP, posture, comfort,
vocalization, appetite, and miscellaneous
behaviors were rejected at the second day
after surgery because they could not be
used to identify the presence of pain nor
to distinguish the placebo group from the
analgesic-treated group.

Nonparametric ANOVA Figure 1—Schematic depiction of rank Spearman correlation coefficients for interfac-
Heart rate, RR, respiratory pattern, tors, factor-to-total score, and item-to-total score for a 4-factor model of a composite
vocalization, and appetite were rejected at pain scale for use in assessing acute postoperative pain in cats undergoing ovario-
hysterectomy. All coefficients reported were significantly (P < 0.001) different from
the evaluation conducted 1 hour after the 0. *Total score did not include the excluded items HR, RR, and respiratory pattern.
end of surgery. Comfort, activity, reaction
to palpation of the surgical wound, reaction to pal- ble 2). Heart rate, RR, and respiratory pattern were re-
pation of the abdomen and flank, mental status, and jected because their correlation coefficients were < 0.4.
miscellaneous behaviors could not be used to identify Item-to-total score correlation at the second day after
treatment with vedaprofen. At the evaluation on the surgery ranged from 0.074 to 0.892. Heart rate, RR, re-
second day after surgery, only reaction to palpation spiratory pattern, SABP, posture, comfort, vocalization,
of the surgical wound could be used to discriminate appetite, and miscellaneous behaviors were rejected.
treatment with tramadol or vedaprofen. Posture and
activity remained in the composite pain scale because Validity and reliability testing (phase 2)—Results
they could be used to distinguish between the times of were determined for phase 2.
assessment.
Construct validity of factor structure
Item-to-total score correlation coefficient Inappropriate items (RR and respiratory pattern)
Item-to-total score correlation at the first hour after were removed on the basis of the previously defined
the end of surgery ranged between 0.280 and 0.942 (Ta- criteria. Exploratory factor analysis was then performed

178 AJVR, Vol 72, No. 2, February 2011

09-10-0348r.indd 178 1/20/2011 12:53:48 PM

 Table 4—Median (minimum and maximum) values of factor scores and total score of a composite pain scale for use in assessing acute
postoperative pain in cats undergoing ovariohysterectomy and receiving various analgesic treatments (10 cats/group) in a previous
study.23

Factor Placebo Tramadol Vedaprofen Tramadol-vedaprofen P value*

1 12.041 (6.361, 15.344)
a
1.852 (0, 11.473)
b
9.738 (4.061, 12.347)
c
1.067 (0, 2.512)
b
, 0.001
2 –6.393a (–8.620, –2.776) –2.122b (–6.623, 0.005) –5.386a (–6.869, –2.919) –1.150c (–2.065, 0.010) , 0.001
3 5.178 (1.693, 6.650)
a
1.014 (0, 6.176)
b
2.949 (1.215, 6.516)
c
0.100d (–0.082, 2.570) , 0.001
4 3.747a (2.540, 7.992) 0.723b (–0.033, 3.303) 3.164a (1.178, 6.716) 0.455c (–0.066, 0.987) , 0.001
5 3.085 (1.668, 4.607) 1.778 (0, 4.329) 1.510 (0.697, 3.810) 1.600 (0, 3.274) . 0.05
Total score† 18.500a (9.000, 25.000) 4.000b,c (0, 18.000) 15.000 a,c (6.000, 18.000) 2.000b (0, 5.000) , 0.001

*Values were considered significant at P , 0.05. †The total score did not include HR, RR, and respiratory pattern.
a–d
Within a row, values with different superscript letters differ significantly (P , 0.05).

with the 11 remaining items for the evaluation at the pain) to 28 (maximum pain). Alternatively, the total fi-
first hour after the end of surgery. nal score could be adjusted to a scale of 0 to 100 by
Three components attained eigenvalues > 1.0 on multiplying the total score by 3.47 (ie, total score X
the unrotated factor solution; however, the interpre- [100/28] = adjusted score). This procedure can facili-
tation of the pattern matrix (with varimax rotation) tate the ranking of pain intensity because it is generally
resulted in a 5-factor solution with eigenvalue magni- accepted that when a scale of 0 to 100 is used, scores
tudes of 3.93, 1.72, 1.55, 1.45, and 1.12. All items had < 30, from 30 to 70, and > 70 are considered indicative
satisfactory factor loadings and communality (Table 3). of mild, moderate, and severe pain, respectively.
Factor-to-total score and interfactors correlations were The short form of the scale designed for use in eval-
high (Figure 1). uating acute pain during the late postoperative period
Dimensions were labeled F1 (psychomotor change (> 24 hours after surgery) was composed of 5 items
[posture, comfort, activity, mental status, and miscel- (posture, activity, mental status, reaction to palpation
laneous behaviors]), which accounted for 35.8% of the of the surgical wound, and reaction to palpation of the
total variance; F2 (protection of wound area [reaction abdomen and flank). The total score of this short form
to palpation of the surgical wound and reaction to pal- ranged from 0 (no pain) to 15 (maximum pain).
pation of the abdomen and flank]), which accounted
for 15.7% of the total variance; F3 (physiologic vari- Discussion
ables [SABP and appetite]), which accounted for 14.1%
of the total variance; F4 (vocal expression of pain [vo- The study reported here was focused on the refine-
calization]), which accounted for 13.2% of the total ment and initial psychometric evaluation of a com-
variance; and F5 (HR), which accounted for 10.2% of posite postoperative pain scale for cats. Reliability and
the total variance. validity of pain assessment tools are important criteria
when selecting instruments for use in clinical practice
and for research purposes.16 The results provided ini-
Construct validity of known-groups discrimination tial evidence of construct validity and internal consis-
The total score and scores for F1, F2, F3, and F4 tency of a multidimensional composite tool for use in
could be used to distinguish the placebo group from the assessing acute postoperative pain in cats undergoing
analgesic-treated group as well as to distinguish among ovariohysterectomy.
groups receiving the various analgesic treatments. Fac- Behavioral and physiologic measures have been
tor 5 (HR) could not be used to distinguish the pla- used to assess pain in cats in clinical and research set-
cebo group from the analgesic-treated group or to dis- tings.24,25,35 Although behavioral observation is the pri-
tinguish among groups receiving the various analgesic mary approach for the assessment of pain in animals,
treatments (Table 4). On the basis of this result, the physiologic indicators (such as increases in HR and
total score did not include values for HR or items re- SABP and changes in RR and respiratory pattern) can
jected in the item analysis (RR and respiratory pattern). be used to provide information about an animal’s re-
sponse to noxious stimuli.36 Therefore, construction of
Reliability of internal consistency the scale included the selection of behavioral and phys-
Internal consistency of each domain (subscale) iologic items supported by the relevant literature8,9,24–27
as expressed by the modified Cronbach α coefficient as well as pain behaviors observed in the previous
was excellent for F1 (0.8985), F2 (0.8841), and F3 study23 conducted to assess postoperative pain in cats.
(0.8025); very good for F4 (0.7662); and unacceptable Refinement of the pain scale by use of item analy-
for F5 (0.2835). On the basis of these results, HR was sis allowed the selection of the most appropriate items
removed from the composite pain scale. The Cronbach to measure pain in cats in the immediate and late (>
α coefficient for the overall scale with the remaining 24 hours after surgery) postoperative period. Resira-
items was 0.8672, which indicated excellent internal tory rate and respiratory pattern had poor performance
consistency. and were deleted from the scale. Although physiologic
The final version of the composite pain scale rep- measures are considered to be more objective than are
resented 10 items scored categorically and distributed behavioral measures, they generally have been incon-
in 4 domains (subscales; Table 5). The total score was clusive for use in determining the presence of pain in
based on a sum of 4 subscales and ranged from 0 (no animals37–40 and children.41 Hence, emphasis has fo-

AJVR, Vol 72, No. 2, February 2011 179

09-10-0348r.indd 179 1/20/2011 12:53:48 PM

 Table 5—Final version of a composite pain scale for assessing acute postoperative pain in cats undergoing an ovariohysterectomy.

Factor Item Description Score

Psychomotor Posture Cat is in any usual posture for the species; cat looks comfortable and relaxed. 0
change Cat is lying in lateral recumbency; pelvic limbs are extended or partially extended 1
and muscles are tense.
Cat is in sternal recumbency with arched back and head held low; cat may be in 2
any position, but it has tense muscles and is reluctant to move.
Cat adopts different postures in an attempt to find a comfortable position. 3
Comfort Cat is awake or asleep; when stimulated, it is interested in its surroundings. 0
Cat is quiet and dissociated of its environment; when stimulated, it is not interested 1
in its surroundings. Cat may be sitting in the back of the cage or facing the back of the cage.
Cat is uncomfortable, agitated, and restless; it lies down and stands up continuously. 2
Activity Cat moves in a normal manner. 0
Cat is quieter than usual. 1
Cat is reluctant to move. 2
Cat frequently changes its body position. 3
Mental status* Observation of mental status A. 0
Observation of 1 of mental status B, C, D, or E. 1
Observation of 2 of mental status B, C, D, or E. 2
Observation of 3 or 4 of mental status B, C, D, or E. 3
Miscellaneous Observation of none of the listed behaviors. 0
behaviors† Observation of 1 of the listed behaviors. 1
Observation of 2 of the listed behaviors. 2
Observation of 3 or 4 of the listed behaviors. 3
Partial score __

Protection of Reaction to Cat does not react when the surgical wound is touched or pressed, or response 0
wound area palpation of the to palpation is similar to that during preoperative evaluation (if preoperative value
surgical wound was recorded).
Cat does not react when the surgical wound is touched but flinches and may 1
vocalize when surgical wound is pressed.
Cat flinches and may vocalize when the surgical wound is touched; cat withdraws 2
and turns its head toward the wound and may vocalize or try to bite when surgical
wound is pressed.
Cat withdraws with fast and intense head movement toward the wound and may vocalize 3
or try to bite when surgical wound is touched or pressed.
Cat vocalizes or tries to bite when the observer approaches and does not permit touching 4
of the surgical wound.
Reaction to Cat does not react when abdomen and flank are palpated or there is no change from 0
palpation of the preoperative palpation response (if preoperative value was recorded); abdomen
abdomen and flank‡ is not tense.
Cat vocalizes or tries to bite when abdomen and flank are palpated; abdomen is tense. 1
Cat vocalizes or tries to bite when the observer approaches; cat does not permit any 2
touching of the abdomen or flank.
Partial score __

Physiologic SABP§ 0% to 15% above preoperative value. 0
variables 16% to 29% above preoperative value. 1
30% to 45% above preoperative value. 2
. 45% above preoperative value. 3
Appetite Cat eats the usual amount of food. 0
Cat eats less than the usual amount of food. 1
Cat is not interested in food. 2
Partial score __

Vocal Vocalization Cat purrs when touched or meows and interacts with observer, but it does not groan, 0
expression hiss, or growl.
of pain Cat vocalizes (groan, hiss, or growl) when observer approaches, but it calms down when 1
touched.
Cat vocalizes (groan, hiss, or growl) when observer approaches, and it does not calm 2
down when touched.
Cat vocalizes (groan, hiss, or growl) spontaneously. 3
Partial score __
Total score __

Guidelines for use of the composite pain scale—The observer should choose only 1 value within each item that best fits the cat’s condition. Initially,
the cat’s behavior may be assessed from outside the cage. Observe whether the cat is at rest or moving, interested or not interested in its surroundings,
and silent or vocalizing. Then assess specific behaviors (miscellaneous behaviors). Open the door of the cage and call the cat’s name to encourage it
to approach to allow assessment of evoked behaviors and interaction with the observer. Assess its reaction (friendly, aggressive, scared, indifferent,
or vocal). Touch the cat and interact with it while observing whether it is receptive. If the cat is resting, encourage it to move by placing it in a standing
position. Observe whether the cat moves spontaneously or is reluctant to move. When possible, the cat should be removed from the cage to assess
activity. Observe the cat’s response after stimulating its interest in its surroundings by playing with it, offering it food, or talking to it. Gently place the cat in
lateral or ventral recumbency and measure SABP. Finally, assess the cat’s reaction when the area around the surgical wound is touched and then gently
pressed; perform the same procedure with the abdomen and flank, if appropriate.
*Categories were as follows: A = Satisfied (cat is alert and interested in its surroundings; it is friendly and interacts with the observer), B = Uninterested
(cat does not interact with the observer), C = Indifferent (cat is not interested in its surroundings), D = Anxious (cat is nervous or frightened and attempts to
hide or escape), and E = Aggressive (cat is aggressive and tries to bite or scratch the observer after the slightest manipulation). †The 4 behaviors were as
follows: the cat wags its tail excessively, contracts and extends its pelvic limbs or contracts the abdominal muscles, has partially closed eyes (ie, squinted
eyes), or frequently licks or chews the wound area. ‡Only for abdominal surgery. §It is necessary to record a preoperative (baseline) value for comparison.

180 AJVR, Vol 72, No. 2, February 2011

09-10-0348r.indd 180 1/20/2011 12:53:49 PM

 cused on the observation of changes in behavior when to indicate pain intensity; however, identification of the
developing scales for use in cats. subscales via factor analysis allowed each domain to be
Results of the item analysis for the late postopera- evaluated separately. This approach may be of interest
tive period indicated that a smaller number of items in both clinical and research settings.
were relevant for evaluating pain when compared with The advantage of assessing each domain separately
results for the immediate postoperative period. In the in clinical practice is that when there is any technical
late postoperative period, only posture, activity, mental difficulty for assessment of a determined dimension,
status, reaction to palpation of the surgical wound, and that dimension may be omitted and not considered in
reaction to palpation of the abdomen and flank were the total score. For example, an arterial blood pressure
not rejected in the item analysis. The reduction in the monitor may not be available or a cat may be agitated
number of relevant items to evaluate pain at > 24 hours when personnel are attempting to measure blood pres-
after surgery may be a reflection of an animal’s adap- sure. For those circumstances, F3 (physiologic vari-
tation to pain (with a reduction in the range of pain ables), which includes SABP and appetite, can be re-
behavior expression) or a reduction in the severity of moved from the scale, and the clinicians may evaluate
pain during the postoperative period because some be- pain in cats on the basis of behavioral changes alone.
havioral changes may be evident only in animals with In analgesic research, evaluation of the scores for
great pain intensity. Discrimination between the rele- the factors (considering the factor loading for each
vant items to assess pain in the immediate postopera- item) increases the quality of information because a
tive period versus at > 24 hours after surgery provided certain analgesic treatment may differentially affect the
the development of a short form specifically for use in factors. For example, one analgesic may be more effi-
evaluating pain in cats at > 24 hours after surgery. cient for control of pain at rest or during movement
Initially, construct validity of the scale was inves- (F1 [psychomotor change]), whereas another one may
tigated by use of a commonly applied method (explor- be more effective for reducing local hyperalgesia (F2
atory factor analysis that determined the dimensionality [protection of wound area]). Therefore, measuring the
of the scale). Factor analysis is a statistical method used factor score may assist clinicians and researchers in
for scale development to evaluate whether all variables identifying treatments that are efficacious for certain
measure a single dimension or > 1 dimension or fac- aspects of pain.
tor.19 Pain expression is considered to be a multidimen- On the other hand, in addition to the score for
sional phenomenon,42 and instruments that evaluate each subscale, multidimensional scales with correlated
multiple domains are more accurate than are scales that dimensions may be used to generate a total score for the
measure only 1 dimension of pain, namely its intensity scale.46 Use of an overall (global) score is possible even
(eg, simple descriptive scale, numeric rating scale, and though the scale was multidimensional because there
visual analogue scale).43 In humans, the McGill Pain was a high correlation between dimensions and the to-
Questionnaire is a self-assessment instrument used to tal pain score. In addition, internal consistency of the
evaluate sensory and affective pain qualities.44 overall scale was excellent. It is useful to create an over-
Factor analysis revealed that the pain scale items all pain score when the goal of assessment is to deter-
loaded on 5 important and interpretable factors, which mine the overall severity of pain. Because of the psycho-
indicated the multidimensionality of the instrument. metric properties of the scale reported here, either the
Five dimensions were highlighted (psychomotor total score or the score for each factor may be used.
change, protection of wound area, physiologic vari- Construct validity also was assessed via known-
ables, vocal expression of pain, and HR). These 5 fac- groups discrimination. This kind of validity determines
tors accounted for 89% of the total variance. However, the extent to which the measure differs between groups.19
because HR did not have acceptable internal consis- The instrument should be used to distinguish groups that
tency and could not be used to discriminate between would be expected to have differences in scores.45 Except
groups, it was deleted from the final version of the for F5 (HR), the identified factors had reasonable discrim-
scale. The 4 remaining factors contributed to 78.8% of inating properties and appeared to be sensitive enough
the total variance. A useful factor model should capture that they could be used to distinguish pain of different se-
a good amount of the total variance in the data,45 as was verities. The factors could be used to detect differences be-
observed in the present study. tween the analgesic-treated and placebo-treated group as
Factors F3 (physiologic variables) and F4 (vocal well as among the groups receiving the various analgesics.
expression of pain) included only 2 items and 1 item, This suggested that the scale may have value as a discrimi-
respectively. Although factors defined by only 1 or 2 native instrument, which is an essential requirement for
items are not considered desirable,45 these 2 factors re- clinical43 and research purposes.
mained in the scale because they were not rejected in Factor F5 (HR) had limited validity as a measure of
the item analysis, had acceptable internal consistency, acute postoperative pain in cats. This factor was exclud-
and could be used to discriminate between groups. ed from the scale because it was unreliable and did not
A test’s dimensionality has important implications have discriminating power. It has been reported23,35,37,40
for the scoring evaluation and use of the test. Multi- that HR is a poor indicator of pain in cats. This may be
dimensional tests with correlated dimensions have a explained by the fact that HR may be influenced by a
score for each subscale and may have a total score for number of factors other than pain, such as stress and
the scale. Both subscale score and total score for the the use of analgesic drugs.47
scale require psychometric evaluation.46 The scale re- The SABP was the only physiologic measure that
ported here was designed to be used with the total score remained in the final version of the scale. In contrast

AJVR, Vol 72, No. 2, February 2011 181

09-10-0348r.indd 181 1/20/2011 12:53:49 PM

 to the other physiologic variables, SABP has been cor- Thus, it is only possible to evaluate whether a cat has
related with acute postoperative pain in cats.23,37 Results more or less pain, but it is not possible to quantify how
of item analysis revealed that SABP was relevant for as- much more or how much less, such as for an interval
sessing pain only during the early postoperative period scale. Although the ordinal instrument has sufficient
but not > 24 hours after surgery. These results indicated precision for clinical purposes, interval measurements
that the value of physiologic measures as indicators of provide a more precise assessment, which is desired for
postoperative pain is limited48 because the autonomic research purposes.52
response appears to adapt with time despite the persis- Results of the study reported here provided initial
tence of pain.47 evidence of the construct validity and reliability of a
In addition to construct validity, the scale was test- multidimensional composite pain scale for use in as-
ed for reliability. Internal consistency approaches can be sessing acute postoperative pain in cats undergoing
used to estimate the reliability of multiple-item tests.46 ovariohysterectomy. Further studies are needed to con-
Analysis of the modified Cronbach α coefficient indi- firm these results because construct validity is an ongo-
cated that internal consistency was excellent for F1, F2, ing process. Because this was an initial study, the next
and F3 and very good for F4. The internal consistency step is to collect additional validity data (including re-
for the overall scale was also excellent. This degree of sponsiveness to analgesic effects and the assessment of
internal consistency allowed items in each factor to be the degree of agreement between 2 or more observers
assessed as a group to provide a score and also to sum (interrater reliabilibilty) and stability over time (intra-
the score for the items to generate a total score for the rater reliability).
scale. On the other hand, F5 had unacceptable internal
consistency and was deleted from the instrument. a. Chen HC, Radzi R, Rahman NA, et al. Analgesic effect of trama-
The need for rescue analgesia may be used in ad- dol combined with tolfenamic acid in cats after ovariohyster-
dition to pain intensity scores to evaluate the efficacy ectomy (abstr), in Proceedings. Annu Meet Am Coll Vet Anesth
2007;27.
of analgesic drugs. In the previous study,23 interven- b. Horspool LJ, Hoeijmakers M, van Laar P, et al. The efficacy and
tion analgesia was administrated to cats with a score safety of vedaprofen oral gel in postoperative pain management
> 33% of the total score for the scale. This value was in cats (abstr), in Proceedings. Voorjaarsdagen Cong 2001;161.
empirically selected on the basis of scores that would
indicate moderate pain and require intervention anal-
References
gesia. However, additional studies would be necessary
to investigate the cutoff point for the provision of relief 1. Hansen B, Hardie E. Prescription and use of analgesics in dogs
from pain. and cats in a veterinary teaching hospital: 258 cases (1983–
1989). J Am Vet Med Assoc 1993;202:1485–1494.
The study reported here had several limitations. 2. Dohoo SE, Dohoo IR. Postoperative use of analgesics in dogs
The sample size of 40 cats may be considered small. and cats by Canadian veterinarians. Can Vet J 1996;37:546–
The scale was refined, and the psychometric properties 551.
were tested by use of the Portuguese language; thus, it 3. Lascelles BDX, Capner CA, Waterman-Pearson AE. Current
is necessary to conduct further studies to assess valid- British veterinary attitudes to perioperative analgesia for cats
ity of the scale when it is translated into English. The and small mammals. Vet Rec 1999;145:601–604.
4. Raekallio M, Heinonen KM, Kuussaari J, et al. Pain alleviation
psychometric properties of the total scale and subscales in animals: attitudes and practices of Finnish veterinarians. Vet
were evaluated in cats undergoing ovariohysterectomy, J 2003;165:131–135.
which is the most commonly performed surgery in 5. Williams VM, Lascelles BDX, Robson MC. Current attitudes to,
research studies11–13,23,25,49 conducted to assess the effi- and use of, peri-operative analgesia in dogs and cats by veteri-
cacy of analgesics in cats. Therefore, this initial valida- narians in New Zealand. N Z Vet J 2005;53:193–202.
tion of the scale would be applicable only for research 6. Hewson CJ, Dohoo IR, Lemke KA. Perioperative use of analge-
sics in dogs and cats by Canadian veterinarians in 2001. Can J
studies. The psychometric properties of the total scale Vet 2006;47:352–359.
and subscales should also be tested in clinical settings 7. Manworren RCB, Hynan LS. Clinical validation of FLACC: pre-
with animals undergoing various surgeries to confirm verbal patient pain scale. Pediatr Nurs 2003;29:140–146.
its validity under those circumstances. Few studies 8. Firth AM, Haldane SL. Development of a scale to evaluate post-
have been published on the use of vedaprofen and tra- operative pain in dogs. J Am Vet Med Assoc 1999;214:651–659.
9. Holton L, Dawson P, Nolan A, et al. Development of a be-
madol in cats; however, the analgesic efficacy of these haviour-based scale to measure acute pain in dogs. Vet Rec
drugs (especially tramadol) has been indicated in the 2001;148:525–531.
literature.23,49–51,a,b Although the statistical analysis de- 10. Morton CM, Reid J, Scott ME, et al. Application of a scaling model
tected relevant differences among items, none of the to establish and validate an interval level pain scale for assessment
scaling methods was used to define weights for the of acute pain in dogs. Am J Vet Res 2005;66:2154–2166.
items and categories of the scale. In the final version of 11. Slingsby LS, Waterman-Pearson AE. Postoperative analgesia
in the cat after ovariohysterectomy by use of carprofen, ke-
the scale, all items were weighted equally. In the same toprofen, meloxicam or tolfenamic acid. J Small Anim Pract
manner, the order of the categories was selected arbi- 2000;41:447–450.
trarily on the basis of results of a literature review and 12. Gassel AD, Tobias KM, Egger CM, et al. Comparison of oral and
the investigators’ experiences. subcutaneous administration of buprenorphine and meloxicam
Another limitation of the final version of the instru- for preemptive analgesia in cats undergoing ovariohysterecto-
my. J Am Vet Med Assoc 2005;227:1937–1944.
ment is that it is an ordinal scale in terms of measure- 13. Tobias KM, Harvey RC, Byarlay JM. A comparison of four meth-
ment properties. In this type of scale, it is assumed that ods of analgesia in cats following ovariohysterectomy. Vet An-
the intensity of pain is reflected in the total pain score aesth Analg 2006;33:390–398.
(defined as the sum of the scores for each of the items). 14. Roughan JV, Flecknell PA. Evaluation of a short duration behav-

182 AJVR, Vol 72, No. 2, February 2011

09-10-0348r.indd 182 1/20/2011 12:53:49 PM

 iour-based post-operative pain scoring system in rats. Eur J Pain of a German version of the neck pain and disability scale. Eur
2003;7:397–406. Spine J 2008;17:922–929.
15. von Baeyer CL, Spagrud LJ. Systematic review of observational 35. Mollenhoff A, Nolte I, Kramer S. Anti-nociceptive efficacy of
(behavioral) measures of pain for children and adolescents aged carprofen, levomethadone and buprenorphine for pain relief in
3 to 18 years. Pain 2007;127:140–150. cats following major orthopaedic surgery. J Vet Med A Physiol
16. Crellin D, Sullivan TP, Babl FE, et al. Analysis of the validation Pathol Clin Med 2005;52:186–198.
of existing behavioral pain and distress scales for use in the pro- 36. Anil SS, Anil L, Deen J. Challenges of pain assessment in domes-
cedural setting. Paediatr Anaesth 2007;17:720–733. tic animals. J Am Vet Med Assoc 2002;220:313–319.
17. Carmines EG, Zeller RA. Reliability and validity assessment. Bev- 37. Smith JD, Allen SW, Quandt JE, et al. Indicators of postoperative
erly Hills, Calif: Sage Publications, 1979;9–69. pain in cats and correlation with clinical criteria. Am J Vet Res
18. Beyer JE, Wells N. The assessment of pain in children. Pediatr 1996;57:1674–1678.
Clin North Am 1989;36:837–854. 38. Conzemius MG, Hill CM, Sammarco JL, et al. Correlation be-
19. Jensen MP. Questionnaire validation: a brief guide for readers of tween subjective and objective measures used to determine
the research literature. Clin J Pain 2003;19:345–352. severity of postoperative pain in dogs. J Am Vet Med Assoc
20. Streiner DL, Norman GR. Health measurement scales: a practical 1997;210:1619–1622.
guide to their development and use. 4th ed. New York: Oxford 39. Holton LL, Scott EM, Nolan AM, et al. Relationship between
University Press, 2008;167–276. physiological factors and clinical pain in dogs scored using a
21. McDowell I. The theoretical and technical foundations of health numerical rating scale. J Small Anim Pract 1998;39:469–474.
measurement. Measuring health: a guide to rating scales an question- 40. Cambridge AJ, Tobias KM, Newberry RC, et al. Subjective and
naires. 3rd ed. New York: Oxford University Press, 2006;10–54. objective measurements of postoperative pain in cats. J Am Vet
22. Netemeyer RG, Bearden WO, Subhash S. Scaling procedures: Med Assoc 2000;217:685–690.
issues and application. Beverly Hills, Calif: Sage Publications, 41. Herr K, Coyne PJ, Key T, et al. Pain assessment in the nonverbal
2003;1–17. patient: position statement with clinical practice recommenda-
23. Brondani JT, Luna SP, Beier SL, et al. Analgesic efficacy of tions. Pain Manag Nurs 2006;7:44–52.
perioperative use of vedaprofen, tramadol or their combina- 42. Melzack R, Torgerson WS. On the language of pain. Anesthesiol-
tion in cats undergoing ovariohysterectomy. J Feline Med Surg ogy 1971;34:50–59.
2009;11:420–429. 43. Murrell JC, Psatha EP, Scott EM, et al. Application of a modified
24. Dobbins S, Brown NO, Shofer FS. Comparison of the effects of form of the Glasgow pain scale in a veterinary teaching centre
buprenorphine, oxymorphone hydrochloride, and ketoprofen for in the Netherlands. Vet Rec 2008;162:403–408.
postoperative analgesia after onychectomy or onychectomy and 44. Melzack R. The McGill Pain Questionnaire: major properties
sterilization in cats. J Am Anim Hosp Assoc 2002;38:507–514. and scoring methods. Pain 1975;1:277–299.
25. Al-Gizawiy MM, Rudé EP. Comparison of preoperative car- 45. Wiseman-Orr ML, Scott EM, Reid J, et al. Validation of a struc-
profen and postoperative butorphanol as postsurgical analge- tured questionnaire as an instrument to measure chronic pain in
sics in cats undergoing ovariohysterectomy. Vet Anaesth Analg dogs on the basis of effects on health-related quality of life. Am
2004;31:164–174. J Vet Res 2006;67:1826–1836.
26. Hellyer PW, Gaynor JS. How I treat acute postsurgical pain in 46. Furr RM, Bacharach VR. Test dimensionality and factor analy-
dogs and cats. Compend Contin Educ Pract Vet 1998;20:140–153. sis. Psychometrics: an introduction. Beverly Hills, Calif: Sage Pub-
27. Muir WW, Gaynor JS. Pain behaviors. In: Gaynor JS, Muir WW, lications, 2008;61–77.
eds. Handbook of veterinary pain management. St Louis: Mosby 47. Katz J, Melzack R. Measurement of pain. Surg Clin North Am
Elsevier Publishing, 2002;65–81. 1999;79:231–252.
28. Goodman LA. Simultaneous confidence intervals for contrasts 48. Hellyer P, Rodan I, Brunt J, et al. AAHA/AAFP Pain management
among multinomial populations. Ann Math Statist 1964;35:716– guidelines for dogs & cats. J Am Anim Hosp Assoc 2007;43:235–
725. 248.
29. Büttner W, Finke W. Analysis of behavioural and physiological pa- 49. Lopez S, Pertuy S, Horspool L, et al. Vedaprofen therapy in cats
rameters for the assessment of postoperative analgesic demand in with upper respiratory tract infection or following ovariohyster-
newborns, infants and young children: a comprehensive report on ectomy. J Small Anim Pract 2007;48:70–75.
seven consecutive studies. Paediatr Anaesth 2000;10:303–318. 50. Castro DS, Silva MF, Shih A, et al. Comparison between the an-
30. Kaiser HF. The varimax criterion for analytic rotation in factor algesic effects of morphine and tramadol delivered epidurally in
analysis. Psychometrika 1958;23:187–200. cats receiving a standardized noxious stimulation. J Feline Med
31. Green SB, Lissitz RW, Mulaik SA. Limitations of coefficient al- Surg 2009;11:948–953.
pha as an index of test unidimensionality. Educ Psychol Meas 51. Pypendop BH, Siao KT, Ilkiw JE. Effects of tramadol hy-
1977;37:827–838. drochloride on the thermal threshold in cats. Am J Vet Res
32. Cronbach LJ. Cofficient alpha and the internal structure of tests. 2009;70:1465–1470.
Psychometrika 1951;16:297–333. 52. Reid J, Nolan AM, Hughes JML, et al. Development of the
33. Cortina JM. What is coefficient alpha? An examination of theo- short-form Glasgow Composite Measure Pain Scale (CMPS-SF)
ry and applications. J Appl Psychol 1993;78:98–104. and derivation of an analgesic intervention score. Anim Welf
34. Scherer M, Blozik E, Himmel W, et al. Psychometric properties 2007;16:97–104.

AJVR, Vol 72, No. 2, February 2011 183

09-10-0348r.indd 183 1/20/2011 12:53:49 PM