Risk factors for tail injuries in dogs in Great Britain

Abstract

The aim of the current study was to quantify the risk of tail injury, to evaluate the extent to which tail docking reduces this risk, and to identify other major risk factors for tail injury in dogs in Great Britain. A nested case-control study was conducted during 2008 and 2009. Data were obtained from a stratified random sample of veterinary practices throughout Great Britain, and questionnaires were sent to owners of dogs with tail injuries and owners of a randomly selected sample of dogs without tail injuries. The risks of injury were reported adjusting for the sampling approach, and mixed effects logistic regression was used to develop a multivariable model for risk factors associated with tail injury. Two hundred and eighty-one tail injuries were recorded from a population of 138,212 dogs attending 52 participating practices. The weighted risk of tail injuries was 0.23 per cent (95 per cent confidence interval 0.20 to 0.25 per cent). Thirty-six per cent of injuries were reportedly related to injuries sustained in the home, 17.5 per cent were outdoor-related injuries, 14.4 per cent were due to the tail being caught in a door, for 16.5 per cent the cause was unknown and the remainder were due to other causes. Dogs with a wide angle of wag and dogs kept in kennels were at significantly higher risk of sustaining a tail injury. Dogs with docked tails were significantly less likely to sustain a tail injury; however, approximately 500 dogs would need to be docked in order to prevent one tail injury. English springer spaniels, cocker spaniels, greyhounds, lurchers and whippets were all at significantly higher risk when compared to labradors and other retrievers. Differences between countries (England, Scotland and Wales) and between rural and urban environments were not significant.

Full text

May xx, 2010 | Veterinary Record | 1
Papers
Papers
Risk factors for tail injuries in dogs in
Great Britain
G. Diesel, D. Pfeiffer, S. Crispin, D. Brodbelt
The aim of the current study was to quantify the risk of tail injury, to evaluate the extent to
which tail docking reduces this risk, and to identify other major risk factors for tail injury in
dogs in Great Britain (GB). A nested case-control study was conducted during 2008 and 2009.
Data were obtained from a stratified random sample of veterinary practices throughout
Great Britain, and questionnaires were sent to owners of dogs with tail injuries and owners
of a randomly selected sample of dogs without tail injuries. The risks of injury were reported
adjusting for the sampling approach, and mixed effects logistic regression was used to
develop a multivariable model for risk factors associated with tail injury. Two hundred
and eighty-one tail injuries were recorded from a population of 138,212 dogs attending
52 participating practices. The weighted risk of tail injuries was 0.23 per cent (95 per cent
confidence interval 0.20 to 0.25 per cent). Thirty-six per cent of injuries were reportedly
related to injuries sustained in the home, 17.5 per cent were outdoor-related injuries,
14.4 per cent were due to the tail being caught in a door, for 16.5 per cent the cause was
unknown and the remainder were due to other causes. Dogs with a wide angle of wag and
dogs kept in kennels were at significantly higher risk of sustaining a tail injury. Dogs with
docked tails were significantly less likely to sustain a tail injury; however, approximately 500
dogs would need to be docked in order to prevent one tail injury. English springer spaniels,
cocker spaniels, greyhounds, lurchers and whippets were all at significantly higher risk
when compared to labradors and other retrievers. Differences between countries (England,
Scotland and Wales) and between rural and urban environments were not significant.
THE docking of dogs’ tails remains controversial and centres on
whether non-therapeutic docking reduces the risk of tail injury suf-
ficiently to justify the ethical concerns of a prophylactic intervention
(Orlans and others 1998, Bennett and Perini 2003). A ban on non-
therapeutic tail docking was introduced in Great Britain in early 2007.
In Scotland, a complete ban was introduced, in Wales the ban was
introduced with specific working breed exemptions, and in England
the ban was introduced with specific working breed-type exemptions
(Anon 2006, Defra 2007). The exemptions include dogs involved in
law enforcement, the armed forces, emergency rescue, lawful pest con-
trol and lawful shooting of animals. These variations in legislation
provided a unique opportunity to evaluate the association between
docking and tail injuries in a population of dogs including substantial
G. Diesel, BVSc, MSc, PhD, MRCVS,
D. Pfeiffer, DrMedVet, PhD, MACVSc,
DipECVPH,
D. Brodbelt, MA, VetMB, PhD, DVA,
DipECVAA, MRCVS,
Department of Veterinary Clinical
Sciences, Royal Veterinary College,
Hawkshead Lane, North Mymms,
Hertfordshire AL9 7TA
S. Crispin, MA, VetMB, BSc, PhD,
DVA, DVOphthal, DipECVO, FRCVS,
Department of Anatomy, University of
Bristol, University Walk, Clifton, Bristol
BS8 1TD
E-mail for correspondence:
gilldiesel@yahoo.co.uk
Provenance: not commissioned;
externally peer reviewed
Veterinary Record (2010) 166, xx-xx doi: 10.1136/vr.b4880
numbers of docked and undocked animals, and to assess whether
country (England, Scotland or Wales) and location (rural or urban) are
risk factors in themselves.
A previous study conducted in Edinburgh in 1985 showed that
tail injuries were rare, with the estimated prevalence being 0.39 per
cent (Darke and others 1985). That study estimated that not dock-
ing a dog’s tail increased the risk of a tail injury 1.28 times, but this
was found to be not significant (95 per cent confidence interval [CI]
0.61 to 2.69 per cent). A more recent survey, which recorded the types
of injuries and causes of lameness in dogs involved in game shoot-
ing, showed a highly significant association between tail injuries and
being undocked among springer spaniels (P=0.008) and cocker span-
iels (P=0.004) (Houlton 2008). Both these studies represented a sub-
set of the dog population in Great Britain and were conducted before
implementation of the restrictions on docking. Additionally, the study
by Houlton (2008) of working dogs relied on a convenience sample,
and the study by Darke and others (1985) is more than 20 years old;
therefore, further work to evaluate tail injuries in Great Britain was
considered necessary. The aim of this study was to quantify the risk of
tail injuries, to ascertain the extent to which docking reduces the risk
of tail injury, and to identify other major risk factors for tail injury in
dogs attending veterinary practices in Great Britain.
Materials and methods
Participants and procedure
A case-control study design was used nested within a cohort of dogs
attending veterinary practices between March 2008 and March 2009.
Power calculations carried out before the study estimated that approxi-
mately 250 dogs with tail injuries would be required. However, these
PE1230/H

2 | Veterinary Record | May xx, 2010
PapersPapers
PapersPapers
calculations were revised on the basis of preliminary estimates of the
prevalence of dogs with docked tails among the dogs recruited into the
study. The revised sample size calculations estimated that approxi-
mately 90 to 120 cases of tail injury would be required based on the
detection of an odds ratio of 0.3 to 0.5, assuming that the prevalence
of docking among dogs was approximately 12 to 14 per cent (95 per
cent confidence level, 80 per cent power, case:control ratio of 1:4)
(Win Episcope 2.0; CLIVE).
A list of mixed and companion animal veterinary practices was
taken from the Royal College of Veterinary Surgeons Practice Register
(RCVS 2008). This list was stratified by country (England, Scotland
or Wales,) and then the list for each country was stratified by location
(rural and urban) based on the postcode classification of the practice
location (Office for National Statistics 2006). A sample of veterinary
practices was then randomly selected, using random number gen-
eration, from each of these lists. The practices in the sample were
approached to determine whether they were using one of seven speci-
fied computerised practice management systems (RoboVet or PremVet
[Vet Solutions], Midshires or Ventana [Cosulsoft], Teleos [Teleos
Systems], Vet-one [Gemhader Software] and RxWorks [RX Works]),
and whether they were willing to participate in the study. Data were
extracted from the practice database of all participating practices, to
obtain a list of all dogs that had attended the veterinary practice in
the previous 12-month period and their clinical histories. A free-text
search was used to identify all dogs that had sustained a tail injury by
searching for the word ‘tail’. The search detected all words containing
‘tail’ whether there was a space or not before or after the word.
Cases were defined as any dog presented to the veterinary practice
within the previous 12 months for treatment of a tail injury, includ-
ing fractures, dislocations, lacerations, contusions, self-trauma and
neoplasia. Tail problems relating to neoplasia and self-trauma were
included as it has been reported anecdotally that in some of these cases
there is an underlying chronic traumatic injury that eventually leads
to the development of a tumour or a self-traumatic injury. A list of all
dogs that had attended each of the participating veterinary practices
during the same one-year period as the case dogs was obtained, and
control dogs were then randomly selected from this list by random
number allocation. For each case, approximately four control dogs
were randomly selected. Dogs selected as controls that had sustained
a tail injury within the past 12 months but had not been treated by
a veterinarian were excluded as controls. Dogs suffering from water
tail/limber tail were excluded from the study as these injuries are not
well understood and it is thought that they are due to muscle fatigue.
It was also thought that including these dogs as cases would result in a
weakening of the power of the study and the possibility of examining
associations between risk factors and typical tail injuries.
Questionnaire design
The owners of the selected cases and controls were sent a questionnaire
during 2008 and 2009. The questionnaire was designed and pretested
before the study. The questionnaire was reviewed by five epidemiolo-
gists and eight clinicians. It was then pretested on five dog owners to
ensure it was clear and easy to follow. The questionnaire was also
translated into Welsh. A prepaid reply envelope was supplied with the
questionnaire, in addition to a disposable tape measure to enable own-
ers to measure the length and height of their dog. The questionnaire
investigated aspects relating to the size, temperament (as perceived by
the owner) and breed of the dog, the home environment, whether
the dog was used as a working dog and the nature of any tail injuries
(Table 1) (questionnaire available on request from GD). Tail wag angle
was assessed by asking the owners to estimate how far the tail deviated
from the midline position by selecting one of three options provided
in the form of a diagram. Dog owners who returned their question-
naire were entered into a monthly prize draw in order to increase the
response rate. A second questionnaire and reminder letter were sent to
all owners if no response was received within four weeks.
Data analysis
All data were entered into a predesigned database with data entry
validation rules (Access 2003; Microsoft). The data were checked,
cleaned and then exported to Stata version 9 (Stata Corp) for analysis.
The weighted risk estimates were calculated accounting for the sam-
pling strategy by using the Stata ‘survey’ commands. Additional risk
approximations were calculated for working and non-working dogs,
for docked and non-docked dogs, and for individual breeds or breed
types based on estimated denominator data. This was calculated by
using the proportion calculated from the data relating to the control
dogs enrolled in the study. ‘Attributable risk’, ‘number needed to treat’
and ‘population attributable risk fraction’ were calculated where
appropriate.
The analysis assessing risk factors initially involved univariable
screening. This was done using chi-squared tests of association and
univariable logistic regression. The ‘xtlogit’ command (with country
and urban/rural as fixed effects and veterinary practice identity as a
random effect) was used in order to account for the clustering in the
dataset. All variables were assessed for collinearity using a correlation
matrix, and where two variables were found to be highly collinear
a decision was made to exclude one variable from the model based
on considerations including a priori importance of the risk factors,
strength of associations and missing values (Dohoo and others 2003).
All continuous variables were assessed graphically for normality. All
variables that had a P<0.2 on univariable screening were put forward
for mulitvariable analysis. Manual forward and backward stepwise
multivariable mixed-effects logistic regression models were developed
assessing the addition or removal of individual variables using the like-
lihood ratio test. Statistical significance was set at the 5 per cent level.
If the likelihood ratio test was not significant, it was also checked
whether the variable had a confounding effect by assessing changes in
the coefficients and significance of other variables in the model before
being removed. All final model variables were assessed for interac-
tions. The fit of the model was assessed using Hosmer-Lemeshow
goodness-of-fit test on the basic logistic regression model. As the ‘xt’
commands in Stata version 9 do not support goodness-of-fit tests, fur-
ther diagnostics, including the calculation of leverage and delta-betas,
were used to identify any outliers or highly influential observations.
The ‘quadchk’ command was used on the final ‘xtlogit’ model to
assess the sensitivity of the quadrature approximation. The change in
coefficients was less than 0.01 per cent and therefore it can be assumed
that the choice of quadrature did not significantly affect the results.
Due to the a priori interest in working dogs, the variable ‘work’ was
forced into all models to assess its significance. Several multivariable
models were developed in order to assess various aspects of the data.
A model was developed for all dogs in the study using different breed
classifications, for spaniels only and for working dogs only.
The breed, sex and age of the dogs owned by non-responders
among the cases and controls were compared with those that did
respond in order to assess the representativeness of cases and controls.
Additionally, the types of injuries recorded among the non-respond-
ing cases were compared to those of the cases whose owners did
respond.
Results
A total of 314 veterinary practices were contacted initially. Of these
practices, 198 either refused to participate or did not have a suitable
computer system to be eligible for inclusion in the study. The remain-
ing 116 practices were then sent a letter requesting their participation
in the study, and 52 agreed to participate. The practices that did not
TABLE 1: Risk factors evaluated in a case-control study of tail
injuries in dogs in Great Britain
Factor
Dog characteristics Age, sex, neuter status, breed, weight, height, tail
length, body length, coat length, coat type, body
condition, docked before injury, tail shape, tail hair,
temperament, tail wag angle, tail wag in circles,
bottom wag, style of tail wag
Owner details/type of activity Country, urban/rural, veterinary practice, uses dog
for work, shows dog, where is dog kept, type of
property, how many other dogs owned, frequency
of exercise, exercise hours, exercise environment,
type of work, frequency of work, work hours, work
environment
PE1230/H

PapersPapers
May xx, 2010 | Veterinary Record | 3
PapersPapers
agree to participate stated one of the following reasons: they did not
want to participate in a study looking at such a topical issue, they
did not have the time, or they were uncomfortable contacting their
clients with questionnaires. The 52 participating veterinary practices
provided clinical records for 138,212 dogs that had attended the prac-
tices within the previous 12-month period. A total of 281 cases were
identified among these clinical records, but questionnaires could not
be sent to all cases at the request of some practices. Three practices
withdrew from participating in the study after their database had
been queried, meaning that data were available on the number of
cases and number of dogs attending the practice but the owners of
these cases could not be sent questionnaires. Additionally, there were
some cases that had recently died or been euthanased; the veterinary
practice requested that a questionnaire not be sent to the owners of
these dogs. A total of 224 questionnaires were sent out to owners
of cases and 799 to owners of controls. Of all the cases, 97 owners
responded (response rate 43.3 per cent), and 227 of the owners of
controls responded (response rate 28.4 per cent). Five controls were
excluded because these dogs had sustained a tail injury in the previous
12-month period but had not been seen or treated by a veterinar-
ian. Among these five controls, two working dogs had sustained an
injury while working and the other three dogs had sustained a house-
hold injury. One of these dogs had a docked tail before sustaining an
injury. The proportion of male dogs among the controls was 48 per
cent and among the cases it was 53 per cent. The mean (sd) age of the
controls was 6.3 (4.2) years and of the cases it was 5.7 (3.8) years.
There was no significant difference between the proportions
of specific breeds among the cases that responded and the cases
that did not respond (P=0.351). Additionally, there was no sig-
nificant difference in age (P=0.985) or sex (P=0.686) between the
case responders and non-responders. Similar results were found
when comparing the responders and non-responders among the
controls (breeds P=0.974; age P=0.974; sex P=0.561). There was
no significant difference in the type of tail injuries recorded in the
clinical data between the case responders and case non-responders
(P=0.873).
Tables 2 and 3 show some descriptive results of the number of
dogs that were docked, the number used for work and the number of
dogs of specific breeds among the cases and controls enrolled in the
study. Among the 29 working dogs, all were used for game shooting
except for five dogs: one of these was a racing greyhound, one was a
German shepherd police dog and three were herding collies.
Risk of tail injury
The weighted risk of tail injuries seen by veterinarians across all regions
was 0.23 per cent/year (95 per cent CI 0.20 to 0.25 per cent). The risks
of tail injury in each country and location are given in Table 4.
Based on the proportion of working and non-working dogs
among the cases and controls, the approximated risk among
working dogs was 0.29 per cent (32 injuries among 10,974 dogs,
95 per cent CI 0.21 to 0.43 per cent) and the approximated risk
among non-working dogs was 0.19 per cent (249 injuries among
127,238 dogs, 95 per cent CI 0.17 to 0.22 per cent); 29 was the
number of working dogs among those that did respond, while 32 is
the approximated number of working dog injuries expected had all
the owners responded to the questionnaire, out of the total 10,974
clinical records. Working dogs had a statistically significantly higher
risk than non-working dogs (P=0.032). The approximated risk for
docked dogs was 0.03 per cent (six injuries among 21,285 dogs,
95 per cent CI 0.01 to 0.06 per cent) and for undocked dogs it was
0.23 per cent (275 injuries among 116,927 dogs, 95 per cent CI 0.21
to 0.27 per cent). Undocked dogs had a significantly higher risk than
docked dogs (P<0.001). The attributable risk was calculated from
these risk approximations and was found to be 0.20 per cent for
docking, and therefore the ‘number needed to treat’ to prevent one
tail injury was 500 dogs. The population attributable risk fraction for
docking was a decrease of 11.9 per cent. Risk approximations were
also calculated for breeds, and these results are given in Table 5.
Types of tail injury
Of the 97 cases for which a questionnaire was completed, 70.1 per cent
(68 cases) were reported to be lacerations and bleeding, 20.6 per cent
TABLE 2: Number of dogs that were tail docked and that were used
for work among the cases and controls in a study of the risk factors
of tail injury
Cases Controls
Working
Not
working Total Working
Not
working Total
Docked 0 2 2 9 26 35
Not docked 12 83 95 8177 185
Total 12 85 97 17 203 220*
* Two owners did not state whether or not their dog’s tail was docked
TABLE 3: Number of dogs of specific breeds/breed types, and
whether they were or were not used for work, among the cases
and controls
Cases Controls
Breed/breed type Working
Not
working Total Working
Not
working Total
Labradors and other
retrievers
316 19 434 38
English springer
spaniels
413 17 7 9 16
Cocker spaniels 1 3 4 1 4 5
Border collies, rough
collies
1 5 6 2 30 32
Jack Russell terriers 0 1 1 1 14 15
Lurchers, greyhounds,
whippets
214 16 0 6 6
Other 133 34 2108 110
Total 12 85 97 17 205 222
TABLE 4: Risk estimates for tail injury among dogs living in
different countries within Great Britain and locations (rural
or urban). No significant difference was found between any
countries or locations
Category
Number of
cases
Number of dogs
at risk
Risk
estimate (%) 95% CI
England 0.17 0.13-0.21
Urban 65 36,509 0.18 0.14-0.22
Rural 22 13,442 0.16 0.09-0.23
Scotland 0.22 0.18-0.26
Urban 48 25,816 0.19 0.14-0.24
Rural 72 29,679 0.24 0.18-0.30
Wales 0.23 0.18-0.28
Urban 72 31,646 0.23 0.18-0.28
Rural 21120 0.18 0.00- 0.43
Weighted risk for
Great Britain
281 138,212 0.23 0.20- 0.25
CI Confidence interval
TABLE 5: Risk approximations for tail injuries in dogs of different
breeds/breed types
Breed/breed type
Number
of
cases
Approximate number
of dogs at risk
Risk
estimate (%) 95% CI
Labradors and other
retrievers 56 23,911 0.23 0.18-0.30
English springer
spaniels 47 10,366 0.45 0.34-0.60
Cocker spaniels 12 3179 0.37 0.22=0.66
Border collies, rough
collies 18 20,732 0.08 0.06-0.14
Jack Russell terriers 39675 0.03 0.01-0.09
Lurchers, greyhounds,
whippets 47 3870 1.22 0.90-1.61
Other 98 66,479 0.15 0.12-0.18
CI Confidence interval
PE1230/H

4 | Veterinary Record | May xx, 2010
PapersPapers
PapersPapers
(20 cases) fractures or dislocations, and of the rest (9.3 per cent, nine
cases) six cases were self-trauma and three cases were neoplasia. The
questionnaires reported that 44.3 per cent (43 cases) were recurrent
tail injuries (based on the owners’ assessments) and 53.6 per cent (52
cases) were not recurrent; in two cases it was not stated whether the
injury was recurrent. According to the owners’ assessments, 36.1 per
cent (35 cases) of the injuries were caused by the dog knocking its tail
against a wall, kennel wall or another household object, 17.5 per cent
(17 cases) were injuries from undergrowth or fences during exercise
or work, 14.4 per cent (14 cases) were due to the tail being caught in
a door, 15.5 per cent (15 cases) were due to other various causes, and
in 16.5 per cent (16 cases) the cause was unknown. The majority
of injuries (57.7 per cent, 56 cases) were treated conservatively with
antibiotics, anti-inflammatories and dressings, 30.9 per cent (30 cases)
resulted in amputation of the tail, and 11.4 per cent (11 cases) did not
require any specific treatment.
Risk factors for tail injuries
The major risk factors for tail injuries identified in the final multi-
variable model are shown in Table 6. Breed was an important factor:
English springer spaniels had 5.97 times the odds of sustaining an
injury compared with labradors and other retrievers, and greyhounds,
lurchers and whippets had 6.85 times the odds. Dogs with docked
tails had 0.03 times the odds of an injury compared with the dogs
that were undocked. Dogs kept in kennels during the day, at night
or both had 3.60 times the odds of sustaining a tail injury compared
with those that were not kept in a kennel. Also, dogs that wagged
their tails in a very wide angle had 3.72 times the odds, and those that
wagged their tail in a moderately wide angle had 2.91 times the odds,
of sustaining an injury compared with the dogs that wagged their tails
in only a narrow angle.
Other factors (the height and weight of the dog, body length, coat
type and type of tail hair) were also shown to be significant factors
(results not shown). However, these factors were not included in the
final model as there was strong collinearity with the variable breed,
which increased the standard errors of the estimates for breed and
made the model unstable.
The variable ‘work’ was forced into the model due to the a priori
interest in work as a risk factor, despite this variable not being signifi-
cant. A variable classifying dogs into ‘game shooting’, ‘other type of
work’ or ‘no work’ was also assessed and found to be not significant.
There were no interactions found and the fit
of the model was good (Hosmer-Lemeshow
model fit statistic P=0.733). The area under
the receiver operating characteristic (ROC)
curve for the logistic regression model was
0.7854 and there were no particularly high
leverage or delta-beta values (defined as
delta-beta >1.0, leverage >2 k/n, where k
is the number variables and n is the number
of observations) (Hosmer and Lemeshow
2000), which indicated no highly influen-
tial observations and supported good model
fit.
Due to the high level of collinearity of
many variables with breed and the increased
odds in spaniels, the model was repeated
restricting the analysis to only spaniels
(cocker and English springer spaniels). The
results of this model are shown in Table 7.
This shows that whether a dog’s tail was
docked or not was the most important fac-
tor, with docked dogs having 0.008 times
the odds of sustaining a tail injury compared
to dogs with undocked tails. The dog’s sex
was included in the model as it had a con-
founding effect on docking. ‘Work’ was
forced into the model but was found to be
non-significant. The fit of the model was
good and the area under the ROC curve
was 0.930. The model development was
repeated using the different classifications
of breeds according to the current English
and Welsh legislation for tail docking (results not shown). The results
of these models showed similar results to the model shown in Table
6. A model was also developed restricting the analysis to just work-
ing dogs. In this case, docked dogs were at significantly lower risk of
sustaining a tail injury compared with those with undocked tails, and
dogs kept in kennels were at a significantly higher risk (results not
shown).
Discussion
This study has been able to estimate the risk of tail injuries in Great
Britain and identify major factors associated with a tail injury occurring
in a large population of dogs attending a veterianry practice. The overall
risk of injury was low, and trauma not associated with work accounted
for the majority of injuries seen by participating veterinary practices.
Work in itself was not a major risk factor, and characteristics such as
the dogs’ breed, tail wag angle and docking status were more important
factors associated with tail injury in practice-attending dogs.
The overall weighted risk of tail injuries in dogs in Great Britain
was estimated to be 0.23 per cent per year, which was lower than the
prevalence (0.39 per cent) found by Darke and others (1985). This
suggests that tail injuries requiring treatment in the general dog popu-
lation of Great Britain could be rarer than previously thought. The dif-
ference in results between the studies may be due to differences in the
population studied. In the study by Darke and others (1985), the study
population was predominantly urban, and restricted to dogs attending
the University of Edinburgh’s small animal clinic. In the present study,
the dogs sampled were selected from veterinary practices throughout
Great Britain, in both urban and rural areas, and therefore were more
likely to be representative of the general dog population of Great
Britain. The study in Edinburgh included dogs with tail lacerations,
contusions, fractures, dislocations, self-trauma, neoplasia and derma-
toses among the cases. However, the present study included only dogs
with lacerations, contusions, fractures, dislocations, self-trauma and
neoplasia as cases. Dogs with tail dermatoses were not considered
as cases for the present study as there are many potential causes of
this condition, such as allergies, flea infestation or even impacted anal
glands. In addition, the risk estimate in the present study is based on
a population of 138,212 dogs, whereas the study by Darke and others
(1985) based the risk estimate on a population of 12,129 dogs.
TABLE 6: Results of a multivariable mixed-effects logistic regression model of risk factors
associated with tail injuries in dogs in Great Britain (the number of observations used in the
final model was 309 out of 319)
Variable category
Number of
cases
Number of
controls β (se) Odds ratio 95% CI P
Breed
Labradors and other retrievers 19 37 1.00
English springer spaniels 16 16 1.786 (0.655) 5.97 1.65-21.52 0.006
Cocker spaniels 4 5 1.558 (0.989) 4.75 0.68-33.03 0.115
Border collies/rough collies 632 –0.753 (0.546) 0.47 0.16-1.37 0.168
Jack Russell terriers 115 –1.492 (1.096) 0.22 0.03-1.93 0.173
Greyhounds, lurchers, whippets 16 61.924 (0.604) 6.85 2.10-22.39 0.001
Other breeds 33 103 –0.152 (0.365) 0.86 0.42-1.76 0.677
Missing 2 8
Tail docked before injury
No 93 181 1.00
Yes 233 –3.467 (0.913) 0.03 0.01-0.19 <0.001
Tail wag angle
Narrow 10 61 1.00
Moderately wide 28 62 1.066 (0.464) 2.91 1.17-7.21 0.021
Very wide 57 91 1.315 (0.433) 3.72 1.59-8.70 0.002
Dog kept in kennels (during night, day or both)
No 78 201 1
Yes 17 13 1.281 (0.508) 3.60 1.33-9.75 0.012
Work use*
No 84 197 1
Yes 11 17 –0.339 (0.656) 0.71 0.20-2.58 0.605
Intercept – – –1.906 (0.493) – – –
Random effect of practice identity (ρ) – – 0.009 (0.013) – – 0.350
* Forced into model due to a priori interest in working dogs
CI Confidence interval
PE1230/H

PapersPapers
May xx, 2010 | Veterinary Record | 5
PapersPapers
The risk of tail injuries found in the present study indicates that
tail injuries are very rare, and the approximated risk of tail injuries in
working dogs was only slightly higher at 0.29 per cent. In the study
by Houlton (2008), 21 of 668 (3.14 per cent) working dogs studied
sustained injuries including articular pathology, fractures and muscu-
lar injuries, among which tail injuries were included. However, direct
comparison of these risks cannot be made due to the differences in the
populations of dogs studied: the study by Houlton (2008) focused only
on working dogs but the present study included all practice-attending
dogs, of which working dogs represented only a small proportion
(9.1 per cent). In addition, the risk estimated by Houlton (2008) related
to many different types of injury, not just tail injury.
The present study found no significant difference in risk between
England, Scotland and Wales, or between urban and rural areas. This
could indicate that there are no differences at all and the rate of tail
injury is so low that minor policy differences between the countries
have no practical consequences, or that these differences have yet to
have a significant impact on the likelihood of tail injuries. This study
was started approximately one year after the introduction of the new
legislation, and therefore it may be too soon to detect differences in the
risks of tail injury due to the differences in legislation. Dogs born after
the ban on tail docking would have been at most 18 to 24 months of
age at the time of the study. Additionally, the current legislation does
not prevent docked or undocked dogs from being moved between
countries.
The most common type of tail injury reported in the present
study was lacerations and bleeding. This is similar to the findings
of Houlton (2008), where tail tip injuries were the most frequently
reported tail injury. It was also interesting to note that 44.3 per cent
of the tail injuries were reported to be recurrent injuries. This shows
an agreement with anecdotal evidence that suggests that tail injuries
are very difficult to treat, often resulting in many treatment attempts
before finally having to amputate the tail. In the present study, almost
one-third of tail injuries requiring veterinary treatment resulted in
amputation.
The risk factor analysis identified several important risk factors.
English springer spaniels and cocker spaniels were both at much
higher risk compared with labradors and other retrievers. This finding
supports that of Houlton (2008), who found that tail injuries were
much more common among these breeds than labradors or pointers.
Additionally, it was found that greyhounds, lurchers and whippets
were at a significantly higher risk than labradors, and also higher than
English springer and cocker spaniels. It has been anecdotally reported
that the high risk among greyhounds, lurchers and whippets may be
due to their long, whip-like tails, which have very little hair cover for
protection (Anon 2008). However, it is important to keep in mind
that, despite these breeds being shown to be the highest risk groups,
the overall risk of tail injuries was still low.
Factors such as height, weight, body length, coat type and tail hair
were found to be significant factors on univariable analysis. However,
these factors could not be included in the
final model because they were highly col-
linear with breed.
In the final model, tail wag angle was
found to be a risk factor, with dogs that
reportedly wagged their tails over a very
wide angle being at greater risk. This intui-
tively makes sense, as the wider a dog wags
its tail, the more likely it is to knock the tail
against objects in its surroundings compared
with dogs that wag their tails in a narrow
angle; in addition, the force with which
dogs wag their tails may be greater over a
wide angle. A dog being kept in kennels
was found to be an important risk factor
for a tail injury. This could possibly be due
to the size of the kennels being too small
in relation to the size of the dogs, thereby
increasing the chances of the dog knocking
its tail against the kennel wall. It could also
be closely linked to working dogs (58.6 per cent of working dogs lived
in kennels, while only 5.2 per cent of non-working dogs lived in ken-
nels). However, the variable ‘work’ was found to be non-significant
regardless of whether the kennel variable was included in the model.
This suggests that work itself was not a major risk factor after adjust-
ing for other major factors. ‘Work’ was highly collinear with breed,
and it could be argued that breed was masking the effect of work.
However, in the model examining only spaniels, work was still non-
significant. The present study had only low power to evaluate work
as a risk factor based on the prestudy power calculations (8 per cent
of the control population were working dogs), and further work on
working dogs may be merited.
The present study suggests that dogs that are docked are less likely
to sustain a tail injury. This supports the findings of the study con-
ducted by Houlton (2008), which showed that there was a strong asso-
ciation between tail injuries and undocked English springer and cocker
spaniels. In contrast, Darke and others (1985) found no significant
association. The difference in findings from the latter study may have
been related to that study assessing the customary/traditional docking
status of breeds and not the actual docking status of individual dogs,
the predominately urban clientele, the lack of adjustment for con-
founding factors, and the small sample size. In the present study, the
results of the additional models for spaniels only and for working dogs
only also showed tail docking to be an important factor in reducing
the likelihood of a dog sustaining a tail injury. This is to be expected,
as if a dog does not have a tail, it has no tail to injure, or if it has a tail of
reduced length, it is less likely to injure the shorter tail. The important
factor to examine is the level of protection that docking provides and
how much more likely an undocked dog is to sustain a tail injury.
The population attributable risk fraction estimate indicates, assuming
a causal association, that tail docking in the dog population studied
is responsible for a 12 per cent reduction in tail injuries, which could
be considered to be a large and notable decrease. However, in absolute
terms, the attributable risk was small at 0.20 per cent, and the number
of dogs that would need to be treated (docked) in order to prevent one
tail injury was very large, at 500 dogs. Additionally, when considering
these results, due to the low number of docked dogs among the cases,
extrapolation of the results to the general dog population in Great
Britain should be interpreted cautiously. One of the factors of interest
at the start of the study was the length of the dog’s tail, and not just
whether or not it had been docked. Some breeds of dog have their tails
docked to two-thirds the normal length (for example, Weimaraner,
Hungarian vizsla), others to half the length (for example, miniature
poodle), and other breeds have most of the tail removed (for example,
rottweiler, Welsh corgi). Unfortunately, due to the small number of
docked dogs among the cases, it was not possible to categorise dogs
into different docking lengths in this study.
Tail docking remains a controversial issue, as evidenced by recent
correspondence (Davidson 2006, King 2007, Penny 2007) and the
number of submissions received by Parliament in the drafting of the
Animal Welfare Bill (Defra 2002). The debate is centred on whether
TABLE 7: Results of multivariable mixed-effects logistic regression model of risk factors
associated with tail injuries in spaniels in Great Britain (the number of observation used in
the final model was 41)
Variable category Number of cases Number of controls β ( se) Odds ratio 95% CI P
Tail docked before injury
No 19 4 1
Yes 117 –4.885 (1.390) 0.008 0.0004-0.12 <0.001
Sex
Male 14 8 1
Female 613 –2.108 (1.214) 0.121 0.01-1.31 0.082
Work use*
No 15 13 1
Yes 5 8 –0.068 (1.144) 0.934 0.10-8.81 0.953
Intercept – – 2.758 (1.073) – – –
Random effect of
practice identity (ρ)
– – 0.012 (0.030) – – 0.426
* Forced into model due to a priori interest in working dogs
CI Confidence interval
7
PE1230/H

6 | Veterinary Record | May xx, 2010
PapersPapers
non-therapeutic tail docking reduces the risk of tail injuries sufficiently
to justify the ethical concerns regarding this prophylactic intervention
(Bower and Anderson 1992, Morton 1992, Bennett and Perini 2003).
A study conducted in Sweden reported that, after a tail docking ban
was put in place, the incidence of tail injuries in German shorthaired
pointers had increased (Strejffert 1992). However, that study also had
several weaknesses: it followed a limited number of litters (53), did not
make comparisons between docked and undocked dogs, did not com-
pare animals before and after the ban, and did not make any statistical
comparisons to support the conclusions. Therefore, conclusions based
on the study should be examined cautiously.
It is important to be aware of the limitations of the present study.
Due to the random sampling and selection of veterinary practices,
only a small number of working dogs were included in the study.
This could potentially decrease the chance of finding any significant
association between work and tail injuries. Additionally, many of the
variables in this dataset were highly collinear, forcing decisions to be
made as to which variables to include and which to exclude from the
final model. This, too, may have resulted in the presence of residual
confounding, thereby weakening any associations or potentially
masking others. One of the potential biases could be the representa-
tiveness of the sample selected. The numbers of veterinary practices
selected in each region were not sampled by probability proportional
to size. This is because there is a very high proportion of practices in
England, such that if this approach had been used, almost no practices
would have been selected in Wales and Scotland, making it impossible
to estimate the risk of tail injuries in these regions with any confi-
dence. Additionally, only practices using specific software packages
were included in the study, and it could be argued that this makes the
sample unrepresentative of the general population of dogs in Great
Britain. However, the cooperation of the some of the biggest software
companies was obtained and seven different practice management
systems were included. As mentioned previously, the sample may be
unrepresentative because not all injuries would have been seen by a
veterinarian. This bias was also highlighted by Houlton (2008). Some
dogs that had sustained a tail injury may not have been examined by
a veterinarian. It is likely that the present study was biased towards
evaluation of major injuries, as more minor injuries may be less likely
to be examined and/or treated by a veterinarian. Five control dogs had
to be excluded because they had sustained a tail injury in the previ-
ous 12-month period but not been seen by a veterinarian. This may
indicate that the prevalence of all tail injuries could be higher than
estimated in this study; however, these injuries were likely to be less
severe, as they had not been seen by a veterinarian, and therefore less
likely to raise welfare concerns. Additionally, the number of untreated
injuries among the controls was based on a relatively small sample
(5 of 227 controls, 2.20 per cent) and the likely range in the true value
would be great (95 per cent CI 0.94 to 5.35 per cent).
The response rate of practices was low, and the average response
rate of dog owners (cases and controls) was 35 per cent. This may
be due to the controversial nature of tail docking, with some people
unwilling to participate. Comparison of a number of key character-
istics available suggested that responders were representative of the
target population.
This study is the largest study to date and the first study to assess
the risk of tail injury and risk factors for dogs from all parts of Great
Britain allowing objective assessment of the frequency of injuries and
risk factors associated with them. The present study has suggested that
the overall risk of tail injuries is low, although specific breeds including
spaniels, greyhounds and lurchers were at substantially higher odds
of injury. The final multivariable risk factor model showed that being
a working dog was not a major risk factor for tail injury, and other
factors, including breed characteristics and levels of activity of dogs,
were more important than work itself in the practice-attending popu-
lation. Docking appeared to have a protective effect against injury, as
expected; however, it was calculated that 500 dogs would need to be
docked in order to prevent one tail injury. Further studies focusing on
what appear to be the highest-risk groups of dogs would be valuable.
Acknowledgements
The authors thank the Welsh Assembly, the Scottish Government and
Defra for funding this study. They would particularly like to thank
the practice management companies, Midshires, RxWorks, Teleos,
Vet-one and Vetsolutions, for assisting with the data querying. They
also thank all the veterinary practices and owners who participated in
the current study, without whose cooperation this study would not
have been possible.
References
ANON (2006) Animal Welfare Act 2006. Chapter 45. Stationery Office. Accessed May
12, 2010
ANON (2008) Why the tail-docking of dogs should be prohibited. www.onekind.org/
info-hub/reports/108-tail-docking. Accessed June 3, 2010
BENNETT, P. C. & PERINI, E. (2003) Tail docking in dogs: a review of the issues.
Australian Veterinary Journal 81, 208-218
BOWER, J. & ANDERSON, H. P. (1992) Confronting cosmetic surgery: cropping and
docking not clear-cut issues. Journal of the American Veterinary Medical Association 200,
1595-1597
DARKE, P. G., THRUSFIELD, M. V. & AITKEN, C. G. (1985) Association between tail
injuries and docking in dogs. Veterinary Record 116, 409
DAVIDSON, D. (2006) Non-therapeutic docking of dogs’ tails. Veterinary Record 158, 70
DEFRA (2007) Regulatory Impact Assessment: Draft of the Docking of Working Dogs’
Tails (England) Regulations 2007. www.opsi.gov.uk/si/si2007/em/uksiem_20071120_
en.pdf. Accessed March 26, 2009
DEFRA (2002) The consultation on an Animal Welfare Bill: an analysis of the replies.
www.defra.gov.uk/wildlife-pets/pets/cruelty/documents/awbillconsultanalysis.pdf.
Accessed March 26, 2009
DOHOO, I., STRYHN, H. & MARTIN, S. W. (2003) Model building strategies. In
Veterinary Epidemiologic Research. University of Prince Edward Island. pp 317-334
HOSMER, D. W. & LEMESHOW, S. (2000) Assessing the fit of the model. In Applied
Logistic Regression. John Wiley & Sons. pp 167-187
HOULTON, J. E. (2008) A survey of gundog lameness and injuries in Great Britain in
the shooting seasons 2005/2006 and 2006/2007. Veterinary and Comparative Orthopaedics
and Traumatology 21, 231-237
KING, R. (2007) Docking of dogs’ tails. Veterinary Record 160, 99
MORTON, D. (1992) Docking of dogs: practical and ethical aspects. Veterinary Record
131, 301-306
OFFICE FOR NATIONAL STATISTICS (2006) Postcode Directories. Census Geography
Data Unit/EDINA
ORLANS, F. B., BEAUCHAMP, T. L., DRESSER, R., MORTON, D. B. & GLUCK, J.
P. (1998) Should the tail wag the dog? In The Human Use of Animals: Case Studies in
Ethical Choice. Oxford University Press. pp 273-287
PENNY, R. H. C. (2007) Docking of dogs’ tails. Veterinary Record 160, 99-100
RCVS (2008) Directory of Veterinary Practices 2008. RCVS
STREJFFERT, G. (1992) Tail injuries of Shorthaired German Pointer dogs born in Sweden
1989. www.cdb.org/countries/sweden.htm. Accessed March 25, 2009
PE1230/H