Conformational risk factors of brachycephalic obstructive airway syndrome (BOAS) in pugs, French bulldogs, and bulldogs

Abstract

Extremely brachycephalic, or short-muzzled, dog breeds such as pugs, French bulldogs, and bulldogs are prone to the conformation-related respiratory disorder—brachycephalic obstructive airway syndrome (BOAS). Affected dogs present with a wide range of clinical signs from snoring and exercise intolerance, to life-threatening events such as syncope. In this study, conformational risk factors for BOAS that could potentially aid in breeding away from BOAS were sought. Six hundred and four pugs, French bulldogs, and bulldogs were included in the study. Soft tape measurements of the head and body were used and the inter-observer reproducibility was evaluated. Breed-specific models were developed to assess the associations between the conformational factors and BOAS status based on functional grading. The models were further validated by means of a BOAS index, which is an objective measurement of respiratory function using whole-body barometric plethysmography. The final models have good predictive power for discriminating BOAS (-) and BOAS (+) phenotypes indicated by the area under the curve values of >80% on the receiver operating curves. When other factors were controlled, stenotic nostrils were associated with BOAS in all three breeds; pugs and bulldogs with higher body condition scores (BCS) had a higher risk of developing BOAS. Among the standardized conformational measurements (i.e. craniofacial ratio (CFR), eye width ratio (EWR), skull index (SI), neck girth ratio (NGR), and neck length ratio (NLR)), for pugs EWR and SI, for French bulldogs NGR and NLR, and for bulldogs SI and NGR showed significant associations with BOAS status. However, the NGR in bulldogs was the only significant predictor that also had satisfactory inter-observer reproducibility. A NGR higher than 0.71 in male bulldogs was predictive of BOAS with approximately 70% sensitivity and specificity. In conclusion, stenotic nostrils, BCS, and NGR were found to be valid, easily applicable predictors for BOAS (+).

Full text

RESEARCH ARTICLE
Conformational risk factors of brachycephalic
obstructive airway syndrome (BOAS) in pugs,
French bulldogs, and bulldogs
Nai-Chieh Liu
1
, Eileen L. Troconis
1
, Lajos Kalmar
1
, David J. Price
1
, Hattie E. Wright
1
, Vicki
J. Adams
2
, David R. Sargan
1☯
, Jane F. Ladlow
1☯
*
1Department of Veterinary Medicine, University of Cambridge, Cambridge, Cambridgeshire,United
Kingdom, 2Vet Epi, Mildenhall, Suffolk, United Kingdom
☯These authors contributed equally to this work.
*jfl1001@cam.ac.uk
Abstract
Extremely brachycephalic, or short-muzzled, dog breeds such as pugs, French bulldogs,
and bulldogs are prone to the conformation-related respiratory disorder—brachycephalic
obstructive airway syndrome (BOAS). Affected dogs present with a wide range of clinical
signs from snoring and exercise intolerance, to life-threatening events such as syncope. In
this study, conformational risk factors for BOAS that could potentially aid in breeding away
from BOAS were sought. Six hundred and four pugs, French bulldogs, and bulldogs were
included in the study. Soft tape measurements of the head and body were used and the
inter-observer reproducibility was evaluated. Breed-specific models were developed to
assess the associations between the conformational factors and BOAS status based on
functional grading. The models were further validated by means of a BOAS index, which
is an objective measurement of respiratory function using whole-body barometric plethys-
mography. The final models have good predictive power for discriminating BOAS (-) and
BOAS (+) phenotypes indicated by the area under the curve values of >80% on the receiver
operating curves. When other factors were controlled, stenotic nostrils were associated with
BOAS in all three breeds; pugs and bulldogs with higher body condition scores (BCS) had a
higher risk of developing BOAS. Among the standardized conformational measurements
(i.e. craniofacial ratio (CFR), eye width ratio (EWR), skull index (SI), neck girth ratio (NGR),
and neck length ratio (NLR)), for pugs EWR and SI, for French bulldogs NGR and NLR, and
for bulldogs SI and NGR showed significant associations with BOAS status. However, the
NGR in bulldogs was the only significant predictor that also had satisfactory inter-observer
reproducibility. A NGR higher than 0.71 in male bulldogs was predictive of BOAS with
approximately 70% sensitivity and specificity. In conclusion, stenotic nostrils, BCS, and
NGR were found to be valid, easily applicable predictors for BOAS (+).
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 1 / 24
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Liu N-C, Troconis EL, Kalmar L, Price DJ,
Wright HE, Adams VJ, et al. (2017) Conformational
risk factors of brachycephalic obstructive airway
syndrome (BOAS) in pugs, French bulldogs, and
bulldogs. PLoS ONE 12(8): e0181928. https://doi.
org/10.1371/journal.pone.0181928
Editor: Francesco Staffieri, University of Bari,
ITALY
Received: March 8, 2017
Accepted: July 10, 2017
Published: August 1, 2017
Copyright: ©2017 Liu et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: The Kennel Club Charitable Trust
(RG71960, https://kccharitabletrust.org.uk)
provided financial support and sources for data
collection to DRS, LK. NCL, and JFL. Cambridge
Overseas Trust (https://www.cambridgetrust.org/
about/cambridge-overseas-trust/) supported NCL
for her PhD. The current study was conducted
during her PhD but was not included in her thesis.

Introduction
Brachycephalic obstructive airway syndrome (BOAS) is a conformation-related respiratory
disorder of dog breeds with shortened skulls and muzzles, such as the pug, the French bulldog,
the bulldog, and others [1]. Breeding selection for extreme brachycephalia has resulted in
deformation of the upper airway tract leading to obstruction, as the soft tissues have not
reduced proportionately with the length of the skull [2]. Affected dogs show noisy and
laboured breathing with exercise and heat intolerance, often accompanied by sleep disturbed
breathing, gastrointestinal disorders such as regurgitation and vomiting, and in the worst
cases, cyanosis, collapse and death [3–6]. Although clinical signs of BOAS can, at an individual
level, be improved by surgery, severely affected dogs are at a higher anaesthetic risk. Moreover,
for dogs that have developed secondary lesions such as collapse of the laryngeal cartilages
(Grade II-III laryngeal collapse), the prognosis may be guarded.
BOAS is a serious welfare issue [7,8]. The average lifespan of brachycephalic breeds is
reduced by approximately three years when compared to that of mesaticephalic and dolicho-
cephalic breeds of similar body size [9–12], with much of this difference likely due to BOAS
and its syndromic effects. The problems caused by BOAS have been compounded by the
increased popularity of the three extreme brachycephalic breeds named above over the
last two decades in the UK and elsewhere. Moreover, the drivers that may have caused the
increasing numbers of these dogs, such as celebrity endorsement and widespread adoption
by advertisers, remain in place. All three breeds are now amongst the top ten breeds in popu-
larity in the UK [13].
In the past century, the skull shape in the extreme brachycephalic breeds has gradually
decreased in facial length and increased in skull width proportionally [14], and it has often
been suggested that this is associated with an increase in both severity and prevalence of BOAS
[15]. In our previous study, we found that approximately 50% of our study dogs in the three
extreme brachycephalic breeds were BOAS-affected according to whole-body barometric
plethysmography (WBBP), an objective respiratory function test [16]. Others have shown that
during a two-year period in the UK, about 20% of pugs have respiratory disease at a level
which triggers a veterinary consultation [17]. Given that an estimated 60% of owners do not
recognise the clinical signs of BOAS [18,19], the true prevalence may be much higher. How-
ever, it is also true that the high-risk breeds include elderly dogs that have normal respiratory
traces and have not suffered from BOAS during their lifespan.
Recently, there has been a growing awareness of BOAS-related welfare issues amongst dog
owners and other stakeholders. It has been suggested that reformation of the breed standards
of brachycephalic breeds could reduce the prevalence of BOAS in extreme brachycephalic
breeds. Several studies have identified and quantified the causal anatomical lesions along the
upper airway using advanced diagnostic images such as computed tomography (CT) and
endoscopy. Quantitative measurements were described in order to look for potential biomark-
ers for BOAS. These include the length and thickness of the soft palate [20], the tracheal diam-
eter [21], nasopharyngeal dimension [22], glottis dimension [23], and mucosal contact points
of the nasal turbinates [24]. However, these measurements required sedation or general anaes-
thesia, which is impractical for screening the pet population—particularly in dogs that have no
clinical signs of BOAS. Alternatively, soft tape measurements have been proposed to quantify
morphology in dogs [25] and were further applied to study BOAS [26]. The Packer et al.
(2015) study reported that craniofacial ratio (CFR) and neck girth are conformation-related
risk factors for BOAS when comparing data across breeds. These measurements are non-inva-
sive and easily accessible. The practical implications of these measurements in specific breeds
is still uncertain, and their reproducibility is also uncertain. Hence there is an urgent need for
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 2 / 24
VJA runs her own commercial consultancy
company, Vet Epi. VJA was involved in data
acquisition and manuscript preparation and was
remunerated for her work on the project from the
research grants. Vet Epi did not have any additional
role in the study design, data collection and
analysis, decision to publish, or preparation of the
manuscript. The specific roles of these authors are
articulated in the ‘author contributions’ section.
Competing interests: VJA runs her own
commercial consultancy company, Vet Epi. There
are no patents, products in development, or
marketed products to declare. This does not alter
our adherence to PLOS ONE policies on sharing
data and materials.
Abbreviations: AIC, Akaike’s Information Criterion;
AUC, area under the curve; BCS, body condition
score; BL, body length; BOAS, brachycephalic
obstructive airway syndrome; BW, body weight;
CFR, craniofacial ratio; CG, chest girth; CI,
confidence interval; CL, cranial length; EMMS,
electromedical measurement systems; ETT,
exercise tolerance test; EW, eye width; EWR, eye
width ratio; ICC, intra-class correlation coefficient;
MV, minute volume; NG, neck girth; NGR, neck
girth ratio; NL, neck length; NLR, neck length ratio;
OR, odds ratio; PEF, peak expiratory flow rate; PIF,
peak inspiratory flow rate; QDA, quadratic
discriminant analysis; ROC, receiver operating
characteristic; RR, respiratory rate; SD, standard
deviation; SI, skull index; SL, skull length; SnL,
snout length; SW, skull width; TBFVL, tidal
breathing flow volume loops; Te, expiratory time;
Ti, inspiratory time; TV, tidal volume; WBBP,
whole-body barometric plethysmography.

anatomical or conformational markers that can allow breeders of these short-faced dogs to
select away from BOAS.
The objective of this study was to identify breed specific external conformational character-
istics that are associated with BOAS. Soft tape measurements were used to quantify conforma-
tional features. Inter-observer reproducibility of the measurements was examined to see
whether these measurements could form a secure basis for breeder decisions. The conforma-
tional measurements and other potential factors such as body condition score (BCS) were then
compared with BOAS Functional Grade (i.e. clinical assessment of respiratory signs before
and after an exercise tolerance test) and the associations were further validated using the
BOAS index (i.e. objective score of respiratory obstruction measured from WBBP parameters)
developed using breed-specific computational models (16).
Materials & methods
Animals
Six hundreds and four dogs (189 pugs, 214 French bulldogs, 201 bulldogs) were included in
this study. The subjects were either: referred for BOAS consultation at the Queen’s Veterinary
School Hospital (QVSH), University of Cambridge; were pet dogs volunteered by UK owners
and breeders; were show dogs exhibited at regional dog shows between September 2013 and
September 2016. Dogs were excluded if: they were aged <1 year of age as they were considered
immature in head and body dimension; they had had previous upper airway surgery; they had
history/clinical findings of lower airway disease. Dogs that were on medications that may
change respiration (e.g. prednisolone and anti-inflammatory drugs) were also excluded from
the study. A detailed history of each dog was taken from owners including type, severity, fre-
quency and circumstances of occurrence of respiratory signs. Work was performed under
informed ethical consents CR62 and CR63 from the Department of Veterinary Medicine, Uni-
versity of Cambridge.
Respiratory grading and BOAS Index
All participating dogs were graded for functional severity of BOAS using a previously estab-
lished four-point functional grading system based on clinical evaluation before and after an
exercise tolerance test [16,19]. BOAS functional Grade 0 dogs (asymptomatic, BOAS free) and
Grade I dogs (mild BOAS, the dog shows mild respiratory noise but exercise tolerance is unaf-
fected) were considered not to have clinically significant BOAS, namely ‘BOAS (-)’ in this
study; Grade II dogs (moderate BOAS, the dog requires medical attention such as weight con-
trol and/or surgical intervention) and Grade III dogs (severe BOAS, the dog requires immedi-
ate surgical intervention) were considered clinically affected, namely ‘BOAS (+)’ in this study.
The BOAS index (a numeric scale of 0–100%, with 0% being entirely asymptomatic and
100% being the most severe status of BOAS), derived from previously established breed-spe-
cific computational models [16] using quadratic discriminant analysis, computed from WBBP
waveforms of some participating dogs, was used to validate the risk predictive models (for
details see the section ‘statistical analysis’). Not all participants had available WBBP data due to
an insufficient testing period, a distracting testing environment, and/or test intolerance of the
animal. The detailed protocols of the WBBP test were reported previously [16].
Conformational classifications and measurements
For all dogs, a body weight measurement, and a standard assessment of the body fat, the body
condition score (BCS) on a 1–9 point scale [27], was performed. The degree of nostril stenosis
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 3 / 24

was examined and classified using a previously established grading system [16]. “Open or
mild” stenosis was considered acceptable for these breeds, and “moderate or severe” stenosis
was defined as ‘stenotic nostrils’ in this study [28]. Examples of the nostril grading for each
breed are shown in Fig 1.
Nine measurements of the skull and body of each dog were taken using a standard one-
meter soft tape measure (millimeter scales): skull length (SL), cranial length (CL), snout
length (SnL), skull width (SW), eye width (EW), neck length (NL), neck girth (NG),
chest girth (CG), and body length (BL). The definitions of each measurement are listed in
Table 1, while Fig 2 illustrates the method of measurement. Some of these measurements
were described previously [25]. SL and SW were used to estimate the skull index, which was
used as an indicator of different skull shapes: brachycephalic, mesaticephalic, or dolichoce-
phalic [29].
Inter-observer reproducibility of the conformational soft tape
measurements
Inter-observer reproducibility of each measurement was tested for each breed separately due
to their different facial and body conformations. Sixty dogs (20 dogs for each breed) were
included in the inter-observer reproducibility test. The bulldogs were from a single dog show,
while the pugs and French bulldogs were from volunteered dogs and referral dogs measured
at the QVSH. For each dog, soft tape measurements were repeated by two investigators who
had been trained to perform the measurements. We randomly combine the two observers
from a group of four investigators to avoid the inclusion of systematic errors and better model
a potential pair of breeders.
Statistical analysis
All Statistical analyses were performed using R (version 3.3.0 for Mac, https://www.r-project.
org). Significance level was set at 0.05 in all tests, unless otherwise indicated.
Comparisons of numeric variables between breeds were performed using analysis of vari-
ance (ANOVA) with a post-hoc test for pairwise comparison. Categorical variables were
compared using Chi-square tests, followed by multiple pairwise comparisons. Bonferroni cor-
rections were used in each case.
Due to facial and body differences between breeds, all of the following analyses were com-
pleted within each breed.
(1) Inter-observer reproducibility of soft tape measurements. To evaluate the inter-
observer agreement, the intra-class correlation coefficient (ICC) [30] for each measurement
of each breed was calculated using R’s package “irr” [31]. Two-way models were used and
the reproducibility between the observers were estimated. The test of agreement was chosen
instead of the test consistency (raw R function code can be found in https://rdrr.io/cran/irr/
src/R/icc.R). This decision was made as in our case the mean values were of interest when
assessing the difference. At least 18 subjects were required to assess ICC. This sample size was
calculated using R’s package “ICC.Sample.Size” and function “calculateIccSampleSize” with
two raters, at 5% significance, 80% power and with preliminary estimate of the CFR’s ICC
at 0.6 across breeds. To evaluate the variation in measurement difference between the two
observers, the estimated mean of random measurement error (i.e. the ratio of measurement
difference between the two observations to the mean of the observations, eME) of each mea-
surement was calculated for each breed. The inter-observer agreement of the measurements
was judged based on the ICC value with the criteria as follows [32]:
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 4 / 24

Fig 1. Examples of different degrees of nostril stenosis in pugs, French bulldogs, and bulldogs. The following
descriptions were adapted from a previously published figure by the authors (Fig 1 in Liu et al. 2016): “Open nostrils: nostrils are
wide open; mildly stenotic nostrils: slightly narrowed nostrils where the lateral nostril wall does not touch the medial nostril wall.
Immediately after the exercise tolerance test (ETT), the nostril wings should move dorsolaterally to open on inspiration;
moderately stenotic nostrils: the lateral nostril wall touches the medial nostril wall at the dorsal part of the nostrils and the nostrils
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 5 / 24

Excellent inter-observer agreement: ICC greater than 0.9
Good inter-observer agreement: ICC between 0.75 and 0.9
Moderate inter-observer agreement: ICC between 0.5 and 0.75
Poor inter-observer agreement: ICC less than 0.5
(2) Predictive models of conformational risk factors for BOAS. Principle component
analysis was used initially to group soft tape measures as a means of reducing the number of
variables, but this did not improve the fit significantly. For further modelling, five ratios of the
measurements were calculated that partially compensate for body size within breeds:
Craniofacial ratio (CFR): snout length (SnL) / cranial length (CL)
Eye width ratio (EWR): eye width (EW) / skull width (SW)
Skull index (SI): skull width (SW)/ skull length (SL)
Neck girth ratio (NGR): neck girth (NG) / chest girth (CG)
Neck length ratio (NLR): neck length (NL) / body length (BL)
Eleven predictive variables of interest were included in breed-specific multivariate logistic
regression models initially, namely: age (months; numeric variable), gender (male/female,
are only open at the bottom. Immediately after the ETT, the nostril wings are not able to move dorsolaterally and there may be
nasal flaring (ie, muscle contraction around the nose trying to enlarge the nostrils; severely stenotic nostrils: nostrils are almost
closed. The dog may switch to oral breathing from nasal breathing with stress or very gentle exercise such as playing.” (Liu et al.
2016).
https://doi.org/10.1371/journal.pone.0181928.g001
Table 1. Descriptions of the soft tape measurements.
Descriptions
Cranial length
(CL)
The distance along the surface of the head at the skull midline from the external
occipital protuberance to the point between the medial canthus of the left eye and the
medial canthus of the right eye.
Snout length
(SnL)
The distance along the surface of the head at the skull midline from the stop to the
rostral end of the nasal planum.
Skull length
(SL) *
The distance along the surface of the head at the skull midline from the external
occipital protuberance to the rostral end of the nasal planum.
Skull width
(SW)
The linear distance (widest distance) between the left external zygomatic arch and the
right external zygomatic arch.
Eye width (EW) The linear distance between the medial canthus of the left eye and the medial canthus
of the right eye.
Neck length
(NL)
The distance along the dorsal body midline from the external occipital protuberance to
the point between the cranial angle of the left scapula and the cranial angle of the right
scapula.
Neck girth (NG) The circumference of the neck at the median distance between the external occipital
protuberance and the point between the cranial angle of the left scapula and the cranial
angle of the right scapula.
Chest girth (CG) The circumference of the deepest part of the thoracic cavity.
Body length
(BL)
The distance along the dorsal body midline from the the point between the cranial angle
of the left scapula and the cranial angle of the right scapula to the root of the tail.
*Skull length measurement in this study was the sum of CL measurement and SnL measurement
https://doi.org/10.1371/journal.pone.0181928.t001
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 6 / 24

binary variable), neuter status (yes/no; binary variable), body weight (kilograms; numeric vari-
able), BCS (scaled 1–9; numeric variable), stenotic nostrils (yes/no; binary), CFR (in propor-
tion; numeric variable), EWR (in proportion; numeric variable), SI (in proportion; numeric
variable), NGR (in proportion; numeric variable), and NLR (in proportion; numeric variable)
where the BOAS status (BOAS (-) and BOAS (+); binary variable) was the outcome variable.
Backwards stepwise model-selection based on Akaike’s Information Criterion (AIC) [33] was
used to obtain the best-fit model. Gender was a variable of interest; even if it did not improve
the final model significantly, it was retained in the model. The estimated probability of BOAS
(+) was computed for each dog from the models. Receiver operating characteristic (ROC)
curves were used to evaluate the predictive performance of the breed-specific models based on
the area under the ROC curve (AUC).
Conformational predictors that were retained in the final model and had satisfactory inter-
observer agreement (good or excellent ICC) were further investigated to establish a threshold
for predicting BOAS that can easily be implemented by breeders. Threshold values for the con-
formational measurements were calculated for male and female separately (due to the signifi-
cant difference between gender) based on maximizing the sum of sensitivity and specificity
from ROC curves.
(3) Validation of the conformational risk factors. A validation test was performed to
test whether the predictive variables that were retained in the logistic regression models (see
previous section) predict BOAS index. BOAS indices were available for only about half of the
dogs (115/189 pugs, 100/214 French bulldogs, and 79/201 bulldogs). The BOAS (+) preva-
lence was compared between the total study population (i.e. all study dogs) and the valida-
tion population (i.e. the ones that had a BOAS index available) for each breed using a Chi-
squared test. The variables were fitted into a multivariate linear regression where BOAS
index (%) was the outcome variable. The adjusted R-squared was calculated for each breed-
specific model.
Fig 2. Demonstration of the soft tape measurements. Nine measurements were made with a firmly held soft tape measure with the dog standing and at
rest: Skull length (SL), cranial length (CL), snout length (SnL), neck length (NL), body length (BL), eye width (EW), and skull width (SW). The detailed
definitions of the measurements are shown in Table 1. The photos of the dogs were taken from two of the study dogs by the authors.
https://doi.org/10.1371/journal.pone.0181928.g002
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 7 / 24

Results
Subjects
Details of the signalment are shown in Table 2. Pugs had significantly higher BCS (median
BCS = 7) than French bulldogs (median BCS = 5, p<0.0001) and bulldogs (median, BCS = 6,
p<0.0001). There was no significant difference in gender distribution between breeds
(p = 0.736). The bulldog group had a significantly lower proportion of dogs that were neutered
when compared to the pug (χ
2
= 40.41, p<0.0001) and the French bulldog groups (χ
2
= 32.12,
p<0.0001). The distribution of the degree of nostril stenosis was significantly different between
the three breeds (χ
2
= 43.28, p<0.0001). 75.4% of French bulldogs had moderately to severely
stenotic nostrils, while prevalence amongst pugs (65.3%) and bulldogs (44.2%) was lower. The
proportion of BOAS (+) pugs (64.6%) was higher than the French bulldog (58.9%) and the
bulldog (51.2%) groups. However, the distribution of the BOAS functional grades did not dif-
fer significantly between breeds.
As all analyses were carried out for each individual breed, results are presented separately
for each breed.
Pugs
(1) Inter-observer reproducibility of soft tape measurements. The inter-observer varia-
tions in soft tape measurements and the results of inter-observer reproducibility are shown in
S1 and S2 Tables (raw data can be found in S1 Data). Overall, most of the measurements and
their ratios had poor inter-observer agreement (ICC<0.5) between the two observers, except
for SnL (ICC = 0.83; 95%CI: 0.62 to 0.93) and CG (ICC = 0.83; 95%CI: 0.62 to 0.93) (Fig 3).
Table 2. Signalment and the proportions of the subjects with stenotic nostrils and different functional grades.
Pug French bulldog Bulldog
Number 189 214 201
Study population Clinical: 14.8% Clinical: 17.3% Clinical: 4.5%
Volunteered: 85.2% Volunteered: 82.7% Volunteered: 95.5%
Age (months, median with range) 36 (12–147) 28 (12–126) 26 (12–178)
Gender Male: 40.7% Male: 44.4% Male: 43.8%
Female: 59.3% Females: 55.6% Female: 56.2%
Neuter status Intact: 64.6% Intact: 68.7% Intact: 91.5%
Neutered: 35.4% Neutered: 31.3% Neutered: 8.5%
Body weight (kg, mean ±SD) 8.53 ±1.47 12.12 ±2.05 25.47 ±3.11
BCS (median with range; obesity *) 7 (4–9); 60.8% 5 (3–9); 8.4% 6 (4–8); 35.3%
Degree of nostril stenosis Open: 9.5% Open: 10.8% Open: 26.9%
Mild: 21.2% Mild: 13.6% Mild: 28.4%
Moderate: 38.1% Moderate: 29.0% Moderate: 34.3%
Severe: 19.6% Severe: 45.33% Severe: 9.5%
NA: 11.6% NA: 1.4% NA: 1.0%
BOAS Functional Grade Grade 0: 4.8% Grade 0: 10.7% Grade 0: 10.9%
Grade I: 30.7% Grade I: 30.4% Grade I: 37.8%
Grade II: 44.9% Grade II: 43.5% Grade II: 38.8%
Grade III: 19.6% Grade III: 15.4% Grade III: 12.4%
SD, standard deviation; BCS, body condition score; BOAS, brachycephalic obstructive airway syndrome; NA, data not available.
*Obesity was defined as BCS 7 on a 9-point scale.
https://doi.org/10.1371/journal.pone.0181928.t002
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 8 / 24

Fig 3. Scatter plots show the inter-observer reproducibility of the conformational soft tape measures.
The diagonal red line indicates perfect agreement. ICC, intra-class correlation coefficient; CFR, craniofacial
ratio; EWR, eye width ratio; SI, skull index; NGR, neck girth ratio; NLR, neck length ratio.
https://doi.org/10.1371/journal.pone.0181928.g003
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 9 / 24

Among the ratios, only CFR had good inter-observer agreement with the estimated ICC = 0.84
(95%CI: 0.64 to 0.93). However, it is worth noting that although CFR had good measurement
agreement, its eME was high at 18.9 ±17.7%. CFR was the ratio of SnL to CL, where CL had
poor inter-observer agreement (ICC = 0.4, 95%CI: -0.05 to 0.71), while SnL had relatively
good inter-observer agreement but with high eME (likely due to the relatively large difference
between the two measurements to its relatively small absolute value: mean = 1.16cm). Conse-
quently, the inter-observer agreement of CFR measurement was questioned.
(2) Soft tape measurement of conformation in relation to BOAS. The summary of each
soft tape measurement with BOAS functional grade can be seen in S3 Table. Substantial over-
laps are present in all ratios among different functional grades (S1 Fig). Among the direct
measurements, the BOAS (+) dogs had significantly greater EW than the BOAS (-) pugs
(p = 0.009). There is no significant difference in any ratios between BOAS (-) and BOAS (+).
(3) Full model of conformation, nostril stenosis and BCS predicting BOAS. The results
of the multivariate logistic regression models are shown in Table 3 (raw data can be found in
S2 Data). The final model for pugs contained seven variables that accounts for 48% of the total
variation in predicting BOAS (+). Female pugs had 5.35 (95%CI: 2.2 to 13.9) times greater
odds than males of being BOAS (+), after adjusting for the other factors. Pugs with moder-
ately/severely stenotic nostrils had 4.58 (95%CI: 2.11 to 10.4) times greater odds of being
BOAS (+) than those with open/mildly stenotic nostrils. Fig 4 shows a clear trend that the
higher the functional grade the higher the proportion of dogs with moderate/severe nostril
stenosis.
With regards to the obesity-related variables, BCS was retained in the final model although
it was not significant (p = 0.064). Nevertheless, the estimated odds ratio was 1.44 with the
upper limit of its 95%CI at 2.13, thus the impact of BCS on BOAS could still be clinically signif-
icant. There was a clear trend that the higher the functional grade, the higher the proportion of
high-BCS (Fig 4).
Conformational factors EWR (p = 0.032) and SI (p = 0.001) were significantly associated
with BOAS after adjusting for other factors. There was a tendency for dogs with higher EWR
and SI to be more likely BOAS (+) (Table 3). However, we note that both of these measure-
ments showed poor inter-observer agreement that question their validity as predictors (S1
and S2 Tables). Fig 5 gives an indication of the effect of each ratio on the probability of BOAS;
depicting the univariate logistic curve fit to each ratio individually.
Fig 6 shows the predictive performance of the final model. The classification was based on a
cut-off value of 0.5 (Fig 6A; that is, a predicted probability of disease >0.5 meant the dog was
classified as positive, otherwise, negative). This could be adjusted depending on the users’ pref-
erence for sensitivity or specificity. The ROC curve had an AUC of 80% (95%CI: 74% to 87%);
considered as good accuracy in classifying BOAS (+) and BOAS (-) dogs.
(4) Validation of the model using BOAS index. The BOAS (+) prevalence (functional
grade II/III) was not significantly different between the total study pugs (i.e. all study pugs)
and the validation population (i.e. the ones that had BOAS index available) (χ
2
= 0.183,
p = 0.669). The results of the multivariate linear regression model can be found in Table 4
(raw data can be found in S3 Data). The adjusted R-squared value of the multivariate linear
regression model was 0.2, indicating the proportion of variability in BOAS index that can be
explained by the models. The predictive factors that were significant in the validation model
were similar to the full model (Table 3), where female pugs (p = 0.03), pugs with moderate/
severe stenotic nares (p = 0.007), and higher SI (p = 0.044) are more likely to have higher
BOAS index. Interestingly, while body weight was not significant in the validation model
(p = 0.36), BCS was significant (p = 0.007) in predicting BOAS index: with one unit increase in
BCS corresponding to an increased BOAS index of 6.4%, on average (all else constant).
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 10 / 24

French bulldogs
(1) Inter-observer reproducibility of soft tape measurements. The inter-observer varia-
tions in soft tape measurements are shown in S1 and S2 Tables (raw data can be found in
Table 3. The breed-specific models predicting the probability of having brachycephalic obstructive airway syndrome (BOAS).
B (SE) z value Odd ratio (95%CI) p value
Model (Pug):pseudo-R
2
=0.48
(Intercept) -12.959 (3.749) - - -
Gender -1.676 (0.468) -3.584 0.187 (0.072 to 0.454) <0.001 ***
BCS 0.363 (0.196) 1.849 1.437 (0.983 to 2.131) 0.064
Stenotic nostrils 1.521 (0.405) 3.759 4.579 (2.110 to 10.396) <0.001 ***
Body weight (kg) 0.460 (0.176) 2.611 1.584 (1.135 to 2.272) 0.009 **
EWR 0.081 (0.038) 2.147 1.084 (1.009 to 1.171) 0.032 *
SI 0.068 (0.021) 3.238 1.070 (1.028 to 1.117) 0.001 **
NLR -0.070 (0.037) -1.869 0.933 (0.866 to 1.003) 0.062
Model (French bulldog):pseudo-R
2
=0.37
(Intercept) -7.563 (2.70) - - -
Age (month) 0.011 (0.007) 1.547 1.011 (0.997 to 1.026) 0.122
Gender 0.757 (0.340) 2.227 2.132 (1.102 to 4.198) 0.026 *
BCS 0.256 (0.177) 1.447 1.292 (0.919 to 1.846) 0.148
Stenotic nostrils 1.731 (0.397) 4.360 5.645 (2.649 to 12.676) <0.0001 ***
CFR -0.065 (0.046) -1.429 0.937 (0.855 to 1.023) 0.153
NGR 0.115 (0.035) 3.265 1.122 (1.049 to 1.206) 0.001 **
NLR -0.071 (0.029) -2.431 0.932 (0.879 to 0.986) 0.015 *
Model (Bulldog):pseudo-R
2
=0.37
(Intercept) -24.119 (4.224) - - -
Gender 0.120 (0.386) 0.311 1.127 (0.524 to 2.390) 0.756
Neuter status 2.091 (0.727) 2.876 8.093 (2.142 to 38.94) 0.004 **
BCS 0.441 (0.193) 2.287 1.555 (1.073 to 2.295) 0.022 *
Stenotic nostrils 0.463 (0.344) 1.349 1.590 (0.811 to 3.132) 0.177
SI 0.045 (0.019) 2.403 1.046 (1.009 to 1.086) 0.016 *
NGR 0.255 (0.050) 5.075 1.290 (1.175 to 1.431) <0.0001 ***
Multivariate logistic regression was used with BOAS (+) and BOAS (-) as the binary outcome variable.
B, regression coefficient; SE, standard error; CI, confidence interval; BCS, body condition score; EWR, eye width ratio = eye width /skull width; SI, skull
index = skull width / skull length; NLR, neck length ratio = neck length / body length; CFR, craniofacial ratio = snout length / cranial length; NGR, neck girth
ratio = neck girth / chest girth.
When including the ratios (EWR, SI, NLR, CFR, NGR), in the above models, the values were converted from a proportion/ratio to a percentage (i.e., the
ratio times 100) so that the coefficients were easier to read.
The binary outcome variable used in the multivariate logistic regression was based on BOAS functional grades: 0 = function grade 0/I; 1 = functional
grade II/III
For the binary variable the coding was defined as follows:
Gender (0 = female; 1 = male)
Stenotic nostrils (0 = open or mildly stenotic nostrils; 1 = moderately or severely stenotic nostrils)
Neuter status (0 = intact; 1 = neutered)
*The significance level was set at p<0.05;
** p<0.01,
*** p<0.001.
https://doi.org/10.1371/journal.pone.0181928.t003
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 11 / 24

S1 data). Overall, most of the measurements and their ratios had poor to moderate agreement
between the two observers. NG (ICC = 0.89, 95%CI: 0.74 to 0.95) and CG (ICC = 0.91, 95%CI:
0.78 to 0.96) measurements had good to excellent inter-observer agreement, but their ratio,
NGR, had poor inter-observer agreement (ICC = 0.41, 95%CI: 0 to 0.71), which is likely con-
tributed by the combinations of variances from NG and CG (Fig 3).
(2) Soft tape measurement of conformation in relation to BOAS. The summary of each
soft tape measurement with BOAS functional grade can be seen in S3 Table. Overall, the
BOAS (+) dogs had significantly shorter SnL (p = 0.035), wider SW (p<0.001), wider EW
(p<0.001), larger NG (p<0.0001), and longer BL (p = 0.047) than BOAS (-) dogs. In terms of
Fig 4. Stacked bar charts demonstrate the relationship between BOAS functional grade and body condition score (BCS), and the degree of
nostril stenosis. The y-axis shows the percentage of dogs of that functional grade showing each degree of BCS or nostril stenosis. Note the clear trends
of an increased proportion of obese dogs and dogs with moderately or severely stenotic nostrils with anincrease in functional grade. In French bulldogs,
there were over 10% of Grade III dogs that had BCS of 3. The underweight condition was attributed to BOAS-related frequent regurgitation.
https://doi.org/10.1371/journal.pone.0181928.g004
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 12 / 24

Fig 5. Univariate logistic regression plots demonstrate the trend of the five conformational ratios againstthe estimated BOAS
probability. The x-axis is the numeric data of each ratio in percentage; the y-axis is the estimated probability of BOAS. CFR, craniofacial ratio;
EWR, eye width ratio; SI, skull index; NGR, neck girth ratio; NLR, neck length ratio.
https://doi.org/10.1371/journal.pone.0181928.g005
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 13 / 24

ratios, the BOAS (+) French bulldog had significantly lower CFR (p = 0.012) and NLR
(p = 0.034), and higher values of SI (p = 0.042) and NGR (p<0.0001). However, the distribu-
tion of ratios was similar to the pugs in that there are substantial overlaps in ratios among dif-
ferent functional grades (S1 Fig). Nonetheless, it is worth noting that the four French bulldogs
that had CFR >0.3 were all Grade 0.
(3) Full model of conformation, nostril stenosis and BCS predicting BOAS. The results
of the multivariate logistic regression model are shown in Table 3 (Raw data can be found in S2
Data). The final model for French bulldogs contained seven variables that accounted for 37% of
the total variation in predicting BOAS (+). Male French bulldogs had 2.13 (95%CI: 1.1 to 4.2)
times greater odds of being BOAS (+), than females. French bulldogs that had moderately/
severely stenotic nostrils had 5.65 (95%CI: 2.65 to 12.68) times greater odds of being BOAS (+)
than those with open/mildly stenotic nostrils. Fig 4 shows a clear trend that the higher the func-
tional grade the higher the proportion of dogs with moderate/severe nostril stenosis.
Fig 6. The predictive performance of the breed specific models. (A) boxplots show the distributions of the estimated probability of being BOAS-
affected. The dotted line at 50% of the probability represents the raw predictive cut-off. For pugs and French bulldogs, the cut-off values canbe adjusted to
improve the specificity of the models; (B) receiver operating characteristic (ROC) curves show the predictive performance of the breedspecific models.
Area under the curve (AUC) was computed with its 95% confidence interval in the bracket.
https://doi.org/10.1371/journal.pone.0181928.g006
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 14 / 24

Fig 5 gives an indication of the effect of each conformational ratio on the probability of
BOAS; depicting the univariate logistic curve fit to each ratio individually. NGR was signifi-
cantly associated with BOAS in French bulldogs, with odds ratios of 1.12 (95%CI: 1.05 to 1.21)
for an increase of 0.01 in NGR, respectively. NLR was significantly associated with BOAS, with
Table 4. Validation of the predictive models using BOAS index.
B (SE) 95%CI t value p value
Validation_Model (Pug): R
2
= 0.20
(Intercept) -27.968 (37.925) - - -
Gender -11.128 (5.046) -21.137 to -1.119 -2.205 0.030 *
BCS 6.417 (2.203) 2.047 to 10.786 2.913 0.004 **
Stenotic nostrils 12.921 (4.681) 3.637 to 22.205 2.760 0.007 **
Body weight (kg) 1.717 (1.867) -1.987 to 5.421 0.919 0.360
EWR 0.249 (0.423) -0.591 to 1.088 0.588 0.558
SI 0.442 (0.217) 0.011 to 0.873 2.035 0.044 *
NLR -0.623 (0.411) -1.437 to 0.192 -1.517 0.132
Validation_Model (French bulldog): R
2
= 0.20
(Intercept) -37.271 (39.269) - - -
Age (month) 0.071 (0.10) -0.128 to 0.271 0.710 0.479
Gender -5.460 (5.395) -16.175 to 5.255 -1.012 0.314
BCS 0.524 (2.650) -4.740 to 5.788 0.198 0.844
Stenotic nostrils 16.490 (5.775) 5.020 to 27.960 2.855 0.005 **
CFR -0.633 (0.666) -1.955 to 0.690 -0.950 0.345
NGR 1.674 (0.530) 0.622 to 2.726 3.160 0.002 **
NLR -0.848 (0.433) -1.707 to 0.012 -1.959 0.053
Validation_Model (Bulldog): R
2
= 0.26
(Intercept) -180.142 (53.314) - - -
Gender -4.725 (6.765) -18.211 to 8.760 -0.699 0.487
Neuter status 23.737 (11.561) 0.690 to 46.784 2.053 0.044 *
BCS 4.427 (2.849) -1.251 to 10.105 1.554 0.125
Stenotic nostrils 16.457 (5.423) 5.646 to 27.267 3.034 0.003 **
SI 0.402 (0.260) -0.115 to 0.919 1.549 0.126
NGR 2.339 (0.698) 0.948 to 3.731 3.351 0.001 **
Multivariate regression was used with BOAS index as the outcome variable.
To validate the predictive models (Table 3) that the outcome variable was based on a subjective clinical assessment (functional grading system), the
objective respiratory function severity score, BOAS index (a numeric score computed from flow waveforms obtained from whole-body barometric
plethysmography), was further used on a smaller study population (pug = 115; FB = 100; bulldog = 79). Variables that were included in the predictive
models in Table 3 were input into the multivariate regression models.
BOAS, brachycephalic obstructive airway syndrome; B, regression coefficient; SE, standard error; CI, confidence interval; BCS, body condition score;
EWR, eye width ratio = eye width/ skull width; SI, skull index = skull width / skull length; NLR, neck length ratio = neck length / body length; CFR, craniofacial
ratio; NGR, neck girth ratio = neck girth / chest girth;
When including the ratios (EWR, SI, NLR, CFR, NGR), in the above models, the values were converted from a proportion/ratio to a percentage (i.e., the
ratio times 100) so that the coefficients were easier to read.
For the binary variable the coding was defined as follows:
Gender (0 = female; 1 = male)
Stenotic nostrils (0 = open or mildly stenotic nostrils; 1 = moderately or severely stenotic nostrils)
Neuter status (0 = intact; 1 = neutered)
*The significance level was set at p<0.05
** p<0.01
https://doi.org/10.1371/journal.pone.0181928.t004
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 15 / 24

an odds ratio of 1.07 (95% CI: 1.01 to 1.14) for a decrease of 0.01 in NLR. In addition, there
was a tendency for a reduction in CFR to be associated with BOAS, although this was not
significant (p = 0.153): the estimated odds ratio was 1.07 (95% CI: 0.98 to 1.17) for a decrease
of 0.01 in CFR. BCS was retained in the best-fit model, however, it was not significant
(p = 0.148). There was no clear relationship between BCS and BOAS functional grade (Fig 4).
Note that, in French bulldogs, 15% of Grade III dogs had BCS of 3, considered as underweight,
which accounts for the majority of underweight French bulldogs among the population.
Fig 6 shows the predictive performance of the final model. The classification was based on a
cut-off value of 0.5 as before (Fig 6A). The model had AUC of ROC at 80% (95%CI: 74% to
86%), considered as good accuracy in classifying BOAS (+) and BOAS (-) dogs.
(4) Validation of the model using BOAS index. The BOAS (+) prevalence (functional
grade II/III) was not significantly different between the total study French bulldogs (i.e. all
study French bulldogs) and the French bulldogs in the validation population (i.e. the ones
that had BOAS index available) (χ
2
= 0.022, p = 0.883). The results of the multivariate linear
regression model can be found in Table 4 (Raw data can be found in S3 Data). The adjusted R-
squared value of the multivariate linear regression model was 0.2, indicating the proportion of
variability in BOAS index that can be explained by the models. The degree of stenotic nostrils
was a significant predictor (p = 0.005). Dogs with moderately/severely stenotic nostrils (com-
pared to open/mild) had a mean increase in BOAS index of 16%. The NGR was significantly
associated with BOAS index (p = 0.002); a 0.01 increase in NGR increases BOAS index by
1.67% on average.
Bulldogs
(1) Inter-observer reproducibility of soft tape measurements. The inter-observer varia-
tions in soft tape measurements are shown in S1 and S2 Tables (raw data can be found in S1
Data). Overall, most of the direct measurements and their ratios had poor to moderate agree-
ment between the two observers except for CG. Among other ratios, only NGR had good
inter-observer agreement with the estimated ICC = 0.81 (95%CI: 0.58 to 0.92) (Fig 3) and its
eME was only a mere 3.7% (S1 Table).
(2) Soft tape measurement of conformation in relation to BOAS. The summary of each
soft tape measurement with BOAS functional grade can be seen in S3 Table. SW (p<0.001),
EW (p = 0.014), and NG (p<0.0001) were significantly greater in BOAS (+) bulldogs, who also
had significantly higher SI (p<0.001) and NGR (p<0.0001). Similar to the other two breeds,
there are substantial overlaps in all ratios among different functional grades (S1 Fig).
(3) Full model of conformation, nostril stenosis and BCS predicting BOAS. The results
of the multivariate logistic regression model are shown in Table 3 (Raw data can be found in
S2 Data). The final model for bulldogs contained six variables that accounted for 37% of the
total variation in predicting BOAS (+). Bulldogs that were neutered had 8.1 times greater the
odds of being BOAS (+). However, the 95% CI was wide (2.14 to 38.94), which is likely due
to the higher proportion of dogs that were not neutered (91.5%). With regards to the obesity-
related variables, BCS was significant in the model (p = 0.022) with odds ratio at 1.56 (95%CI:
1.07 to 2.3). Fig 4 demonstrates the distribution of BCS and degree of nostril stenosis against
BOAS functional grade.
Fig 5 gives an indication of the effect of each conformational ratio on the probability of
BOAS; depicting the univariate logistic curve fit to each ratio individually. In the final model
(Table 3), SI (p = 0.016) and NGR (p<0.0001) were significantly associated with BOAS, with
odds ratios of 1.05 (95%CI: 1.01 to 1.09) and 1.29 (95%CI: 1.18 to 1.43) for an increase of 0.01
in SI and NGR, respectively.
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 16 / 24

Fig 6 shows the predictive performance of the final model. The classification was based on a
cut-off value of 0.5 (Fig 6A), which showed good accuracy in classifying BOAS (+) and BOAS
(-) dogs with AUC of the ROC at 81% (95%CI: 75% to 87%).
In the final model, NGR was retained as a valid predictor in the model (p<0.0001) with
good inter-observer agreement of the measurement (ICC = 0.81), as well as a significant pre-
dictor for BOAS index (p = 0.001). The mean NGR in male bulldogs (mean = 0.71) was signifi-
cantly higher than that of female bulldogs (mean = 0.66, t = -7.40, df = 170.9, p<0.0001). The
AUC of the ROC for NGR alone was 73% (95%CI: 61–84%) for male bulldogs, which indicates
moderate accuracy in classifying BOAS (+) and BOAS (-) dogs. The cut-off NGR value was
0.71 with sensitivity of 71% and a specificity of 69%. NGR was less sensitive in female bulldogs
than male bulldogs: the AUC was slightly lower in female at 70% (95%CI: 61–80%). With a
cut-off NGR value at 0.66, the sensitivity was 71% and the specificity was lower at 61%.
(4) Validation of the model using BOAS index. The BOAS (+) prevalence (functional
grade II/III) was not significantly different between all study bulldogs and the bulldogs in the
validation group (i.e. the ones that had BOAS index available) (χ
2
= 1.093, p = 0.296). The
results of the multivariate linear regression models can be found in Table 4 (raw data can be
found in S3 Data). The adjusted R-squared value of the multivariate linear regression model
was 0.26.
Neuter status (p = 0.044), stenotic nostrils (p = 0.003), and NGR (p = 0.001) were signifi-
cantly associated with BOAS index. Compared to the intact bulldogs, the neutered bulldogs
had a mean increase in BOAS index of 24%, although again the 95%CI was wide (1% to 47%).
Dogs with moderately/severely stenotic nostrils (compared to open/mild) had a mean increase
in BOAS index of 16% (95%CI: 6% to 27%). A 0.01 increase in NGR is associated with a mean
increase of 2% (95%CI: 1% to 4%) in BOAS index.
Discussion
This study describes breed-specific models using several conformational factors to predict the
probability of being BOAS-affected. The large sample size in the study supports that each ratio
is reliable in population terms, and could guide the writing of breed standards, although the
likelihood of inaccuracy in many of the individual measurement as shown by ICC and eME
means that most of the individual limits cannot be set. The reliable measures, such as the NGR
in bulldogs, and other easily accessible factors, such as nostril stenosis and BCS, may be of use
for breeding selection.
Inter-observer agreement of conformational soft tape measurements
Sutter et al. (2008) reported that approximately 0.13% of soft tape measurements used in
a multi-breed study were judged to be measurement errors, yet the actual inter-observer
agreement of the measurements was not tested. Neither did the Packer et al. (2015) study
investigating the conformational risk factors for BOAS describe measurement inter-observer
agreement. Unfortunately, most of the conformational soft tape measurements in the present
study had poor inter-observer agreement in all three breeds. The authors have found that per-
forming tape measurements may be challenging on unsedated dogs, as measurements can be
altered easily with slight changes in position (e.g. small changes in degree of the angle between
the neck and the back or in head carriage when standing). In dogs with loose and thick skin
and/or thick fat coverage it is particularly difficult to reproduce the measurements with good
accuracy. Moreover, some of the dogs objected to facial measurements such as SnL. The mea-
surements mentioned above showed large errors of up to 18.7% between two different trained
observers and had poor inter-observer agreement according to ICC. These measurement
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 17 / 24

errors directly affect the inter-observer agreement of the respective ratios. The inter-observer
reproducibility of the CFR was worst in French bulldogs with the mean measurement errors
over 22%. French bulldogs have highly variable over-nose skin fold patterns that affect this
measure, whereas in the other two breeds, the type of fold was more uniform (a single fold).
Among all the ratios, only CFR in pugs and NGR in bulldogs had reasonably good inter-
observer agreement.
Breed-specific predictors for brachycephalic obstructive airway
syndrome
In pugs, female dogs have a higher risk of developing BOAS than male dogs. Interestingly, in
French bulldogs, the trend was the opposite. While male dogs are often used as stud dogs at an
early age, postponing the decision to breed until the dog is older is recommended, as affected
dogs may only show clinical signs in adulthood.
Stenotic nostrils were a significant predictor for BOAS in all three breeds, consistent with
our previous findings [16]. Stenosis of the nostrils is the only BOAS airway lesion that can be
easily diagnosed without sedation and/or specific equipment, such as an endoscope. The grad-
ing system proposed by the authors is straightforward and easily applicable by dog owners.
Importantly, nostril stenosis may play a significant role in the severity of BOAS. Nasal breath-
ing is predominant in dogs, even when the dog is panting, the majority of airflow passes
through the nasal cavity during inspiration [34,35]. Commonly, dogs with moderately/
severely stenotic nostrils have immobile nostril wings during exercise. Whereas dogs with
open/mildly stenotic nostrils usually have mobile nostril wings that can abduct further when
needed [15]. Due to the restriction of airflow at the entrance of the airway, dogs with stenotic
nostrils can be prone to poor thermal regulation and may have an excessive increase in nega-
tive pressure within the airway. Stenotic nostrils are a particular issue in French bulldogs, in
the present study, 45% of French bulldogs had severe stenosis of the nostrils. Since the impact
of stenotic nostrils on BOAS is substantial, the responsible breeder should avoid using dogs
with moderate/severe stenotic nostrils.
Obesity, as quantified using BCS, was a robust risk factor for BOAS, and this result is con-
sistent with previous studies [16,26]. The impact of obesity on respiratory function includes a
decrease in minute volume with an increase in respiratory rate, exercise intolerance, and a
decrease in estimated arterial oxygen saturation [16,36,37]. Interestingly, while BCS was a sig-
nificant risk factor for BOAS in pugs and bulldogs, this was not the case in French bulldogs.
Only 8.4% of French bulldogs were obese compared to 60.8% and 35.3% of pugs and bulldogs,
respectively. About 15% of Grade III French bulldogs were underweight and all of these had
frequent regurgitation. In addition to the possible anatomical abnormalities (e.g. oesophageal
diverticula), the increase in thoracic negative pressure during respiratory distress could further
trigger gastrointestinal signs such as regurgitation and vomiting as a result of gastro-oesopha-
geal reflux and temporary hiatal hernia [38,39]. Nevertheless, the impact of obesity on BOAS
in French bulldogs should not be ignored. It was noted that the majority of the obese French
bulldogs were BOAS (+) (Fig 4A).
BOAS, in many ways, is similar to human obstructive sleep apnoea (OSA) [40–43]. The effect
of obesity on OSA has been investigated with different measurements. Recently, body mass
index (BMI) z-score and neck-to-waist ratio were recognized as independent predictors of OSA
[44]. Similar findings were shown in the current study for French bulldogs and bulldogs. Packer
et al. (2015) have previously reported that a greater neck girth increased the risk of BOAS, but
the NG to CG ratio was not significantly associated with BOAS. In the current study, both the
absolute measurement of NG and NGR were significantly higher in BOAS (+) French bulldogs
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 18 / 24

and bulldogs (S3 Table). The reason why NG and other direct measurements were not included
in the initial models was that the measurements are significantly affected by body size and gen-
der. Although gender was included in all models, the variation in size within the same gender
should not be ignored. As the CG was comparable between the BOAS (-) and BOAS (+) dogs, it
could be used as a reference to normalize NG. Neck fat accumulation is associated with severity
of OSA in humans, and could cause a reduction in pharyngeal lumen diameter, further trigger-
ing collapse of the airway [45–49]. Unlike pugs and bulldogs, French bulldogs in our study
tended to have an ideal BCS. High NGR in slim dogs could be caused by either fat or muscle.
The actual relationship between the NGR and the impact on upper airway obstruction is
unknown. Nevertheless, NGR is a strong and valid predictor of BOAS independently of other
factors in bulldogs and it may be assumed that selection away from this phenotype will reduce
the prevalence of BOAS. In French bulldogs, it was unexpected that the inter-observer agree-
ment of NGR was poor when the measures of NG and CG were both reasonably reproducible.
It was found that in some dogs, the observer-1 measured the NG longer but CG shorter than
observer-2. Although the differences were not considerable for both of the measurements, the
accumulated errors have a significant impact on NGR measurement inter-observer agreement.
Other conformational factors such as SI and EWR in pugs, and NLR in French bulldogs
were significantly associated with BOAS in the final models, after adjusting for other factors
such as gender. However, not only did these measurements have poor inter-observer agree-
ment, they were also not significantly associated or only marginally associated with the BOAS
index. Therefore, the measurements may not be valid for predicting individual BOAS-affected
dogs at this stage. However, it is possible, that by introducing more reliable measurement
methods these factors may be used as valid predictors for BOAS (+) in the future.
Craniofacial ratio and brachycephalic obstructive airway syndrome
A previous study suggested that BOAS risk increases in dogs with relatively shorter muzzles
(craniofacial ratio, CFR) and thicker necks, across different brachycephalic breeds: from
extreme brachycephalic breeds such as the pug (median CFR = 0.08) to moderate brachyce-
phalic breeds such as the Staffordshire Bull Terrier (median CFR = 0.5) [26]. In our study with
large numbers of dogs of the three breeds, we obtained supportive data on NGR, but only a
weak association of BOAS status with CFR in a single breed. Within breeds, the variations in
CFR were very limited. CFR overlapped considerably between the different BOAS functional
grades. Our findings on the reproducibility of these measures and the large differences in
detailed conformation between brachycephalic dog breeds suggest that the true associations
between CFR and BOAS for specific breeds may not be comparable to the findings in the Packer
et al. (2015) study that compared multiple breeds with, in most cases, relatively small numbers
of dogs. Anatomically, the CFR measurement cannot determine the main internal BOAS lesions
along the upper airway. Fig 7 illustrates the position at which CFR measurements are made, in
comparison to the position of the internal lesions of BOAS. For extreme brachycephalic dogs,
the SnL only includes the region of the nasal planum and nasal vestibule, while other common
BOAS lesions such as overcrowded and aberrant nasal turbinates, elongated soft palate, and
macroglossia underlie the CL. Therefore it is questionable that having higher CFR would effec-
tively decrease the risk of BOAS for all individuals in the current population. Instead, airway
crowding will occur both for individuals with a short facial length and for individuals with a
short cranium, so that the most severely affected CFR is not predictable. In dogs that had CFR
higher than the third upper-quartile (CFR >0.19), 46% were still BOAS (+). In contrast to this
result, it is more encouraging that among the dogs that had open/mildly stenotic nostrils, only
25% of them were BOAS (+) whereas among the dogs that had moderately/severely stenotic
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 19 / 24

nostrils, 70% of them were BOAS (+). When considering a more effective criterion to assist in
breeding away from BOAS, it is likely that the nostril grading would be superior to CFR.
Limitations of the study
There are several limitations to this study. Firstly, only about half of the study population had
an available BOAS index, which markedly reduces the power to validate the final model.
Fig 7. An illustration demonstrates the relationship between the external craniofacial ratio (CFR) measurement and the corresponding internal
anatomical structures of the upper airway. The realistic anatomical illustration was made according to a computed tomographic 3-dimensional
rendering image of a French bulldog. The illustration was reprinted from the Cambridge BOAS research group website (http://www.vet.cam.ac.uk/boas)
under a CC BY license, with permission from the group in the University of Cambridge, original copyright 2016.
https://doi.org/10.1371/journal.pone.0181928.g007
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 20 / 24

Secondly, the study was conducted over three years. It might be possible that the investigators
gradually gained more experience on measuring the dogs over time. It should be noted though
that the reproducibility measurements were performed towards the end of the study. It is also
possible, though unlikely, that the trend of the conformation changed over the study period.
Thirdly, as all study dogs were recruited from the UK, the results may have limited significance
on these breeds in other regions of the world.
Conclusion
Nostril stenosis is a strong predictor of BOAS for all three breeds. Dogs with moderate to
severe stenosis of the nostrils were at higher risk of developing BOAS. BCS is significantly asso-
ciated with BOAS in pugs and bulldogs with obese dogs having a higher risk of being BOAS
(+). Among the conformation measurements, NGR is a valid predictor of BOAS in male bull-
dogs and highly reliable between different observers, thus it could potentially be used for
breeding selection. EWR and SI in pugs, and NGR and NLR in French bulldogs, SI in bulldogs,
were associated with BOAS but had poor-moderate inter-observer reproducibility. Neverthe-
less, they may be of use for directing the reformation of breed standards.
Overall, the conformational and external factors as measured here contribute less than 50%
of the variance that is seen in BOAS. The authors strongly suggest using these in conjunction
with regular clinical assessment of respiratory signs before and after exercise (BOAS Func-
tional Grading). More importantly, breeding toward extreme brachycephalic features should
be strictly avoided.
Supporting information
S1 Table. The results of the inter-observer mean measurement errors of the conforma-
tional soft tape measurements.
(DOCX)
S2 Table. The results of the inter-observer agreement of the conformational soft tape mea-
surements.
(DOCX)
S3 Table. Comparison of conformational soft tape measures between BOAS (-) and BOAS
(+) dogs.
(DOCX)
S1 Fig. Boxplots show the distribution of the five conformation ratios against BOAS func-
tional grades. The x-axis is BOAS functional grade; the y-axis is the ratios in percentage. CFR,
craniofacial ratio; EWR, eye with ratio; SI, skull index; NGR, neck girth ratio; NLR, neck length
ratio.
(TIF)
S1 Data. Raw data for the reproducibility tests of soft tape measurements.
(XLSX)
S2 Data. Raw data for the full model.
(XLSX)
S3 Data. Raw data for the validation model.
(XLSX)
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 21 / 24

Acknowledgments
The French Bulldog Club of England, Pug Dog Club, The Bulldog Breed Council, Birming-
ham & Midland Counties Bulldog Club, Junior Bulldog Club, Midland & Northern Counties
French Bulldog Club, Northern Pug Dog Club, Pennine & Scottish French Bulldog Club,
Rochdale & District Bulldog Club, West Pennine Pug Dog Club, local breeders and volunteer
pet owners.
Author Contributions
Conceptualization: NCL LK DRS JFL.
Formal analysis: NCL LK DJP.
Funding acquisition: NCL LK DRS JFL.
Investigation: NCL ELT LK HEW VJA DRS JFL.
Methodology: NCL LK DJP DRS JFL.
Project administration: NCL LK DRS JFL.
Supervision: DRS JFL.
Validation: NCL ELT DRS JFL.
Visualization: NCL LK DJP.
Writing – original draft: NCL DRS.
Writing – review & editing: NCL ELT LK DJP VJA DRS JFL.
References
1. Knecht C. Upper airway obstruction in brachycephalic dogs. Comp Cont Educ Pract. 1979; 1:25–30.
2. Hendricks JC. Brachycephalic airway syndrome. Vet Clin North Am Small Anim Pract. 1992; 22
(5):1145–53. PMID: 1523786
3. Amis TC, Kurpershoek C. Pattern of breathing in brachycephalic dogs. Am J Vet Res. 1986; 47
(10):2200–4. PMID: 3777646
4. Hendricks JC, Kline LR, Kovalski RJ, O’Brien JA, Morrison AR, Pack AI. The English bulldog: a natural
model of sleep-disordered breathing. J Appl Physiol. 1987; 63(4):1344–50. PMID: 3693167
5. Hendricks JC. Brachycephalic airway syndrome. Vet Clin North Am Small Anim Pract. 1992; 22
(5):1145–53. PMID: 1523786
6. Meola SD. Brachycephalic airway syndrome. Top Companion Anim Med. 2013; 28(3):91–6. https://doi.
org/10.1053/j.tcam.2013.06.004 PMID: 24182996
7. Asher L, Diesel G, Summers JF, McGreevy PD, Collins LM. Inherited defects in pedigree dogs. Part 1:
disorders related to breed standards. Vet J. 2009; 182(3):402–11. https://doi.org/10.1016/j.tvjl.2009.08.
033 PMID: 19836981
8. Bateson P. Independent Inquiry into Dog Breeding. University of Cambridge. 2016. https://www.
ourdogs.co.uk/special/final-dog-inquiry-120110.pdf
9. O’Neill DG, Church DB, McGreevy PD, Thomson PC, Brodbelt DC. Longevity and mortality of owned
dogs in England. Vet J. 2013; 198(3):638–43. https://doi.org/10.1016/j.tvjl.2013.09.020 PMID:
24206631
10. Michell AR. Longevity of British breeds of dog and its relationships with sex, size, cardiovascular vari-
ables and disease. Vet Rec. 1999; 145(22):625–9. PMID: 10619607
11. Cassidy KM. Breed weight and lifespan. Dog longevity. 2007. http://users.pullman.com/lostriver/
longhome.htm.
12. O’Neill DG, Jackson C, Guy JH, Church DB, McGreevy PD, Thomson PC, et al. Epidemiological associ-
ations between brachycephaly and upper respiratory tract disorders in dogs attending veterinary
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 22 / 24

practices in England. Canine Genet Epidemiol. 2015; 2:10. https://doi.org/10.1186/s40575-015-0023-8
PMID: 26401338
13. The_Kennel_Club_UK. Dog breeds: registration statistics in the UK: The Kennel Club; 2016. http://
www.thekennelclub.org.uk/registration/breed-registration-statistics/
14. Koch D, Sturzenegger N. Vera
¨nderung des Scha
¨dels bei brachycephalen Hunden im Verlaufe der letz-
ten 100 Jahre. Kurzmitteilungen. 2015; 157(3):1–3.
15. Oechtering G. Brachycephalic syndrome—new information on an old congenital disease. Vet Rec.
2010; 20(2):2–9.
16. Liu N-C, Adams V, Kalmar L, Ladlow J, Sargan D. Whole-body barometric plethysmography character-
izes upper airway obstruction in 3 brachycephalic breeds dogs. J Vet Intern Med. 2016; 30(3):853–65.
https://doi.org/10.1111/jvim.13933 PMID: 27159898
17. O’Neill DG, Darwent EC, Church DB, Brodbelt DC. Demography and health of Pugs under primary vet-
erinary care in England. Canine Genet Epidemiol. 2016; 3:5. https://doi.org/10.1186/s40575-016-0035-
zPMID: 27293771
18. Packer R, Hendricks A, Burn C. Do dog owners perceive the clinical signs related to conformational
inherited disorders as ‘normal’ for the breed? A potential constraint to improving canine welfare. Anim
Welf. 2012; 21(1):81–93.
19. Liu N-C, Sargan D, Adams V, Ladlow J. Characterisation of Brachycephalic Obstructive Airway Syn-
drome in French Bulldogs Using Whole-Body Barometric Plethysmography. PLoS One. 2015; 10(6):
e0130741. https://doi.org/10.1371/journal.pone.0130741 PMID: 26079684
20. Grand J-GR, Bureau S. Structural characteristics of the soft palate and meatus nasopharyngeus in
brachycephalic and non-brachycephalic dogs analysed by CT. J Small Anim Pract. 2011; 52(5):232–9.
https://doi.org/10.1111/j.1748-5827.2011.01047.x PMID: 21539567
21. Kaye BM, Boroffka SA, Haagsman AN, Haar GT. Computed tomographic, radiographic, and endo-
scopic tracheal dimensions in English bulldogs with Grade 1 clinical signs of brachycephalic airway syn-
drome. Vet Radiol Ultrasound. 2015; 56(6):609–16. https://doi.org/10.1111/vru.12277 PMID: 26202379
22. Heidenreich D, Gradner G, Kneissl S, Dupre G. Nasopharyngeal dimensions From computed tomogra-
phy of pugs and French bulldogs with brachycephalic airway syndrome. Vet Surg. 2016; 45(1):83–90.
https://doi.org/10.1111/vsu.12418 PMID: 26731598
23. Caccamo R, Buracco P, La Rosa G, Cantatore M, Romussi S. Glottic and skull indices in canine brachy-
cephalic airway obstructive syndrome. BMC Vet Res. 2014; 10:12. https://doi.org/10.1186/1746-6148-
10-12 PMID: 24410902
24. Schuenemann R, Oechtering G. Inside the brachycephalic nose: intranasal mucosal contact points. J
Am Anim Hosp Assoc. 2014; 50:149–58. https://doi.org/10.5326/JAAHA-MS-5991 PMID: 24659729
25. Sutter NB, Mosher DS, Gray MM, Ostrander EA. Morphometrics within dog breeds are highly reproduc-
ible and dispute Rensch&apos;s rule. Mamm Genome. 2008; 19(10–12):713–23.
26. Packer R, Hendricks A, Tivers M, Burn C. Impact of facial conformation on canine health: brachyce-
phalic obstructive airway syndrome. PLoS ONE. 2015; 10(10):e0137496. https://doi.org/10.1371/
journal.pone.0137496 PMID: 26509577
27. Lafflamme DP. Development and validation of a body condition score system for dogs. Canine Pract.
1997; 22:10–5.
28. Liu NC, Adams VJ, Kalmar L, Ladlow JF, Sargan DR. Whole-body barometric plethysmography charac-
terizes upper airway obstruction in 3 brachycephalic breeds of dogs. J Vet Intern Med. 2016; 30(3):853–
65. https://doi.org/10.1111/jvim.13933 PMID: 27159898
29. Evans H. The Skeleton. Miller’s Anotomy of the Dog. 4th ed. Philadelphia, USA: Saunders; 1993. p.
122–218.
30. Shrout PF, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;
86:420–8. PMID: 18839484
31. Gamer M, Leom J, Singh I. irr: Various Coefficients of Interrater Reliability and Agreement. R package
version 0.84. 2012. https://CRAN.R-project.org/package=irr.
32. Portney L, Watkins M. Foundations of Clinical Research: Application to Practice. 3rd ed. Upper Saddle
River, NJ: Pretice Hall; 2007.
33. Akaike H. A new look at the statistical model identification. IEEE Trans Autom Control. 1974; 19
(6):716–23.
34. Schmidt-Nielsen K, Bretz WL, Taylor CR. Panting in dogs: unidirectional air flow over evaporative sur-
faces. Science. 1970; 169(3950):1102–4. PMID: 5449323
35. Goldberg MB, Langman VA, Taylor CR. Panting in dogs: paths of air flow in response to heat and exer-
cise. Resp Physiol. 1981; 43:327–38.
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 23 / 24

36. Manens J, Bolognin M, Bernaerts F, Diez M, Kirschvink N, Clercx C. Effects of obesity on lung function
and airway reactivity in healthy dogs. Vet J. 2012; 193(1):217–21. https://doi.org/10.1016/j.tvjl.2011.10.
013 PMID: 22099184
37. Manens J, Ricci R, Damoiseaux C, Gault S, Contiero B, Diez M, et al. Effect of body weight loss on car-
diopulmonary function assessed by 6-minute walk test and arterial blood gas analysis in obese dogs. J
Vet Intern Med. 2014; 28(2):371–8. https://doi.org/10.1111/jvim.12260 PMID: 24351032
38. Poncet C, Dupre
´G, Freiche V, Estrada M, Poubannet Y, Bouvy B. Prevalence of gastrointestinal tract
lesions in 73 brachycephalic dogs with upper respiratory syndrome. J Small Anim Pract. 2005; 46:273–
9. PMID: 15971897
39. Poncet C, Dupre G, Freiche V, Bouvy B. Long-term results of upper respiratory syndrome surgery and
gastrointestinal tract medical treatment in 51 brachycephalic dogs. J Small Anim Pract. 2006; 47:137–
42. https://doi.org/10.1111/j.1748-5827.2006.00057.x PMID: 16512845
40. Hendricks J, Kline LR, Kovalski RJ, O’Brien J, Morrison A, Pack AI. The English bulldog: a natural
model of sleep-disordered breathing. J Appl Physiol. 1987; 63:1344–50. PMID: 3693167
41. Hoareau G, Mellema M. Pro-coagulant thromboelastographic features in the bulldog. J Small Anim
Pract. 2015; 56(2):103–7. https://doi.org/10.1111/jsap.12299 PMID: 25482453
42. Hoareau GL, Jourdan G, Mellema M, Verwaerde P. Evaluation of arterial blood gases and arterial blood
pressures in brachycephalic dogs. J Vet Intern Med. 2012; 26(4):897–904. https://doi.org/10.1111/j.
1939-1676.2012.00941.x PMID: 22574946
43. Mellema MS, Hoareau GL. Hypomagnesemia in brachycephalic dogs. J Vet Intern Med. 2014; 28
(5):1418–23. https://doi.org/10.1111/jvim.12393 PMID: 24986211
44. Katz SL, Vaccani JP, Barrowman N, Momoli F, Bradbury CL, Murto K. Does neck-to-waist ratio predict
obstructive sleep apnea in children? J Clin Sleep Med. 2014; 10(12):1303–8. https://doi.org/10.5664/
jcsm.4284 PMID: 25325606
45. Mortimore I, Marshall I, Wraith P, Sellar R, Douglas N. Neck and total body fat deposition in nonobese
and obese patients with sleep apnea compared with that in control subjects. Am J Respir Crit Care
Med. 1998; 157(1):280–3. https://doi.org/10.1164/ajrccm.157.1.9703018 PMID: 9445310
46. Smith P, Wise RA, Gold AR, Schwartz AR, Permutt S. Upper airway pressure-flow relationships in
obstructive sleep apnea. J Appl Physiol. 1988; 64(2):789–95. PMID: 3372436
47. Onat A, Hergenc¸G, Yu¨ksel H, Can G, Ayhan E, Kaya Z, et al. Neck circumference as a measure of cen-
tral obesity: Associations with metabolic syndrome and obstructive sleep apnea syndrome beyond
waist circumference. Clin Nutr. 2009; 28:46–51. https://doi.org/10.1016/j.clnu.2008.10.006 PMID:
19010573
48. Hoffstein V, Mateik S. Differences in abdominal and neck circumferences in patients with and without
obstructive sleep apnea. Eur Respir J. 1992; 5(4):377–81. PMID: 1563498
49. Cizza G, de Jonge L, Piaggi P, Mattingly M, Zhao X, Lucassen E, et al. Neck circumference is a predic-
tor of metabolic syndrome and obstructive sleep apnea in short-sleeping obese men and women.
Metab Syndr Relat Disord. 2014; 12(4):231–41. https://doi.org/10.1089/met.2013.0093 PMID:
24571423
Conformational risk factors for BOAS in brachycephalic dogs
PLOS ONE | https://doi.org/10.1371/journal.pone.0181928 August 1, 2017 24 / 24