Research in Veterinary Science 93 (2012) 1006–1014

Contents lists available at SciVerse ScienceDirect

Research in Veterinary Science

journal homepage: www.elsevier.com/locate/rvsc

Correlation and discriminant analysis between clinical, endoscopic, thoracic

X-ray and bronchoalveolar lavage fluid cytology scores, for staging horses
with recurrent airway obstruction (RAO)
P. Tilley a,⇑, J.P. Sales Luis a, M. Branco Ferreira b
a
Department of Clinics, Faculdade de Medicina Veterinária, Universidade Técnica de Lisboa, Avenida da Universidade Técnica, 1300 Lisboa, Portugal
b
Medicina II, Faculdade de Medicina, Hospital Santa Maria, 1100 Lisboa, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: As recurrent airway obstruction (RAO) is progressive and as medical history is frequently unknown by
Received 13 July 2011 owners, it’s important to suggest a score model to characterize RAO stages for a more accurate diagnosis
Accepted 28 October 2011 and treatment. The authors correlated clinical (CS), endoscopic (ES), thoracic X-ray (XRS) and bronchoal-
veolar lavage fluid (BALFS) scores in horses with RAO, in an attempt to establish relevance of each factor’s
contribution for the characterization of RAO stages and to suggest a staging method. Thirty horses with
Keywords: RAO and ten healthy controls were studied. Pearson correlation coefficients were determined between
RAO
CS, ES, XRS and BALFS. Only significant correlation coefficients (>0.60) were considered. One way variance
Endoscopy
Correlation
analyses were used to compare the two groups. A discriminant analysis model was adjusted on the RAO
BAL staging method suggested. There was a significant correlation coefficient between the CS cough, nostril
flare and abdominal lift, all the mucus ES (0.61–0.84), the XRS interstitial pattern, bronchial radiopacity
and thickening and tracheal thickening (0.67–0.78) and the BALFS neutrophil percentages (0.63–0.84).
These variables (e.g., cough) which presented a significant correlation coefficient were considered rele-
vant and chosen for a score model to characterize RAO stages. The ten healthy controls were attributed
stage 0 and the 30 RAO horses were attributed stages 1 (4 horses), 2 (7 horses), 3 (10 horses) and 4 (9
horses). There was also a significant correlation coefficient between all the relevant variables and the
RAO stage (0.61–0.89). Furthermore, discriminant analysis of the RAO staging method showed 92.5% of
original grouped cases and 85.0% of cross-validated grouped cases correctly classified, having confirmed
major contribution of the same variables that had significant correlation coefficients. Even though further
confirmation by lung functional testing is desirable, the significant correlation between relevant variables
and RAO stage and the results of RAO staging discriminant analysis support the suggestion of our score
model for the characterization of RAO stages.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction allergens (Robinson, 1997; Künzle et al., 2007; Riihimäki, 2008).

Lower airway disease is reported to be second only to musculo-
skeletal disease as a cause of wastage among performance horses
(Rossdale et al., 1985). Bracher et al. (1991) and Dixon et al.
(1995) have reported over 50% prevalence of recurrent airway
obstruction (RAO) in horses referred for respiratory problems,
emphasizing the relevance of the disease.
RAO is a common condition in stabled horses, usually character-
ized by airway inflammation (with increased numbers of
bronchoalveolar lavage neutrophils and eosinophils) and obstruc-
tion following exposure of susceptible horses to environmental
⇑ Corresponding author. Tel.: +351 96 2812549; fax: +351 21 3652815.
E-mail addresses: paula_tilley@sapo.pt, paulatilley@fmv.utl.pt (P. Tilley).
Obstruction is the end result of the inflammatory process, associ-
ated with bronchospasm and with the accumulation of mucus
and exudates in the airway lumen. Recurrent bouts of inflamma-
tion lead to structural changes in the mucosa and in the smooth
muscle so that the airway wall is thickened (Robinson, 1997).
Being an obstructive disease, it is characterized by coughing and
pronounced abdominal lift at the end of exhalation (Costa et al.,
2000).
The International Workshop on Chronic Airway Disease held in
Michigan State University, USA in June 2000 (Robinson, 2001),
made as a first recommendation for research the development of
a standardised, internationally accepted clinical scoring system
for equine lower airway disease and Robinson et al. (2000) and
Couëtil et al. (2001) reported failure of clinical score alone to reflect

0034-5288/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.rvsc.2011.10.024
 P. Tilley et al. / Research in Veterinary Science 93 (2012) 1006–1014
low-grade airway obstruction, suggesting that airway disease is
under-diagnosed, especially as there can be considerable obstruc-
tion before the onset of clinical signs.
Because medical history or physical examination are not totally
reliable to exclude other diagnosis or characterize airway impair-
ment, more objective assessments are still necessary for a more
accurate diagnosis of RAO in horses. Respiratory endoscopy, tho-
racic radiography and scan and bronchoalveolar lavage fluid cytol-
ogy and total protein concentration can give relevant support to
the diagnosis.
Furthermore, the condition is progressive, with respiratory dis-
tress becoming more severe over the years. As medical history is
frequently unknown, as horses change owners and environment
various times through their lives, it is important to suggest a score
model for the characterization of different stages of RAO.
In this view, the first objective of this study was to suggest clin-
ical (CS), endoscopic (ES), thoracic X-ray (XRS) and bronchoalveolar
lavage fluid cytology scores (BALFS) for use in RAO diagnosis is
horses. The next purpose was to find out which are the best corre-
lations for the suggested CS, ES, XRS and BALFS in an attempt to
establish the relevance of each factor as a contribution for the char-
acterization of different stages of RAO. The final aim of our study
was to suggest a staging method of RAO horses.
2. Material and methods
2.1. Horses
Thirty horses with RAO and 10 healthy controls were selected
and studied according to the phenotypic description of RAO
(Robinson, 2001), taking into account medical history and physical
examination, additionally supported by respiratory endoscopy,
thoracic radiography and bronchoalveolar lavage fluid (BALF)
cytology and total protein concentration.
Taking into account the reversible airway obstruction induced
by exposure to organic dust which is characteristic of RAO (Robin-
son, 2001), all 30 affected horses included in this study had been
environmentally exposed before entering the Veterinary Teaching
Hospital by being exposed to a particulate rich environment which
triggered inflammation and airway obstruction (locally relevant
allergens are moulds – Alternaria alternata, Aspergillus fumigatus,
Cladosporium mix, Mucor mucedo, Penicillium notatum; mites – Der-
matophagoides (D.) pteronyssinus, D. farinae, Lepidoglyphus destruc-
tor, Tyrophagus putrescentiae; grasses – Dactylis glomerata, Phleum
pratense, Lolium perenne; cereals – Secale cereale; trees – Olea euro-
paea; weeds – Parietaria judaica; and dog dander). This is to say
that they were always examined in exacerbation.
Ramseyer et al. (2007) and Laumen et al. (2010) considered the
time when the horse exhibited the most severe disease manifesta-
tion for assessment of chronic lower airway disease and concluded
that this reliably identifies RAO-affected horses. All horses had to
have had a longer than 2-month history of hay feeding in their life.
The control horses in the present study came from the same sta-
bles as RAO horses and so were equally exposed, not having shown
any clinical respiratory signs.
All 40 horses were Lusitanos or cross-Lusitanos. Age was not
significantly different between the two groups (P = 0.25; t-test),
with a normal distribution (Shapiro–Wilk test). Age average and
standard deviation was 14.68 ± 3.98 (years) in the RAO group
and 16.50 ± 4.63 (years) in the Control group.
All horses were vaccinated regularly (ProteqFlu-TeÒ, Merial1)
and on a regular anthelmintic program and had all been treated with
1 7
Merial Portuguesa LDA, Av. Maria Lamas, Lte 19, Serra das minas, 2635-432 Rio de
Mouro, Portugal.
1007
ivermectin (EqvalanÒ, MSD-AgVet2) a month before being examined,
having shown no evidence of the equine lungworm Dyctiocaulus
arnfieldi L1 larvae in the BALF using the Baermann technique (Hare
and Viel, 1998).
No corticosteroids, anti-histaminics or bronchodilators were
given to any of the horses up to a minimum of two weeks before
being examined and they had not previously been subjected to
immunotherapy.
2.2. Horses – further characterization (variables not used for RAO
staging)
Although not taken into account for the RAO staging method
proposed, the authors also performed skin prick tests (SPT) to 16
locally relevant allergens on the 40 horses included in the present
study. All the horses belonged to a very valuable autochtonous
breed and were all privately owned which posed an obstacle to
the use of inhalation challenge tests to clarify the relevance of
SPT identified allergens. Therefore, following SPT identification of
relevant aeroallergens for each RAO horse, owners were instructed
to take measures to remove the selected aeroallergens from the
horse’s environment (Tilley et al., 2010).
Electrocardiograms (ECG) and echocardiograms (EcoCG) were
also carried out on all horses to evaluate possible cardiovascular ef-
fects of RAO.
2.3. Clinical exam
All horses were examined at the equine hospital of Lisbońs Vet-
erinary College by two independent clinicians and relevant clinical
variables were classified according to Tables 1 and 2. These vari-
ables were cough, nostril flare, abdominal lift, exercise intolerance,
nasal discharge, respiratory rate and pulmonary auscultation.
2.4. Respiratory endoscopy
All horses were sedated with detomidine hydrochloride (0.01
mg/kg bwt; DomosedanÒ, Pfizer Animal Health3), associated with
butorphanol tartrate (0.014 mg/kg bwt; TorbugesicÒ, Fort Dodge4)
to help suppress the intensity of the cough reflex.
Respiratory endoscopies were carried out in all horses with a
flexible video bronchoscope (10.4  3000 mm Karl Storz5, PV-G
28–300). Videos were recorded for each horse (Karl Storz AIDA
VET system) and all videos were viewed twice by two independent
clinicians in order to classify the endoscopic variables according to
Tables 1 and 2.
2.5. Broncoalveolar lavage (BAL) fluid differential nucleated cell
percentage
BAL was carried out in all horses with a BAL silicone catheter
(240 cm long and 2.5 mm internal diameter; COOK6), under the
sedation administered for respiratory endoscopy and lidocaine
(20 ml of 20 mg/ml AnestesinÒ, Medinfar7) were instilled in order
to ease the intensity of the cough reflex. Once the tube was wedged,
the cuff was inflated with air (approximately 5 ml) and held securely
at the nares. Warm (37 °C) sterile physiologic saline (between 400
and 500 ml) were instilled into the BAL tube (Hoffman, 2008) by bol-
2
MSD AgVet, Kirkland, PQ, Canada.
3
4
Pfizer, 235 East 42nd Street, New York10017, NY, USA; Tel.: +1 212 733 2323.
Fort Dodge Animal Health, Fort Dodge, Iowa 50501, USA.
5
Karl Storz GmbH&Co. KG, Postfach 230, D-78503 Tuttlingen, Germany.
6
Cook Veterinary Products, P.O. Box 489, Bloomington IN 47402-0489, USA.
Medinfar, Rua Henrique Paiva Couceiro no. 27, Venda Nova, 2700-451 Amadora,
Portugal.
1008 P. Tilley et al. / Research in Veterinary Science 93 (2012) 1006–1014

Table 1
Clinical staging of RAO sheet (only selected variables with correlation coefficients higher than 0.60 are considereda).

GRADE 0 1 2 3 4 5
CS
Cough Score None Coughs on specific times Frequent cough with periods of no coughing Very frequent cough
of day (feeding/
exercising/making beds)
Nostril flare None Flares during inspiration Flares in inspiration and exhalation (slight Flares in inspiration and expiration (no
(returns to normal at end movement can still be seen) movement can be seen)
scoreb
inspiration)
Abdominal lift None Slight flattening of Obvious abdominal flattening and ‘‘heave Obvious abdominal lift and ’’heave line’’
ventral flank line’’ extending no more than half way extending beyond halfway between
scoreb
between cubital joint and tuber coxae cubital joint and tuber coxae
CS final score = Cough score + Nostril flare score + Abdominal lift score =
Score: 0 (CS final score < 2)/ 1 (2 6 CS final score 6 4)/ 2 (5 6 CS final score 6 6)/ 3 (7 6 CS final score 6 9)

ESc
Mucus None, Little, multiple small Moderate, larger blobs Marked, confluent or stream-forming Large, pool- Extreme,
accumulation clean blobs forming profuse
amounts

Mucus colour None, Colourless White Thick White Yellow Thick

clean Yellow

Mucus None, 1/2 Ventral 2/3 Lateral 3/4 Dorsal Threading Threading
localization Clean
and
stickiness

Mucus apparent None, Very fluid Fluid Intermediate Viscous Very

Clean viscous
viscosity
ES final score = Mucus accumulation + Mucus colour + Mucus localization and stickiness + Muccus apparent viscosity =
Score: 0 (ES final score < 8.5)/ 1 (8.5 6 ES final score 6 12)/ 2 (12 < ES final score 6 16)/ 3 (ES final score > 16)

BALFS
Neutrophil% 620 21–40 41–60 P61
(400
nucleated
cell count)

BALFS final score = Neutrophil% =

Score: 0 (BALFS final score 6 20)/ 1 (21 6 BALFS final score 6 40)/ 2 (41 6 BALFS final score 6 60)/ 3 (BALFS final score P 61)

XRS
Interstitial No Mild increase of Moderate increase of parenchial opacity Increased parenchial opacity with loss of
parenchial parenchial opacity vizualization of vascular structures
pattern
opacity
Bronchial Normal Mild increase Moderate increase Severe increase
radiopacity

Tracheal No Mild increase in Moderate increase in thickening Severe increase in thickening

thickening increase in thickening
thickening

Bronchial No Mild increase in Moderate increase in thickening Severe increase in thickening

thickening increase in thickening
thickening

XRS final score = Interstitial pattern + Bronchial radiopacity + tracheal thickening + bronchial thickening =
Score: 0 (XRS final score < 5)/ 1 (5 6 XRS final score 6 6)/ 2 (6 < XRS final score 6 8)/ 3 (XRS final score > 8)
Total Score = CS final score + ES final score + BALFS final score + XRS final score =
RAO Stage = Stage 0 – Stage 1 – Latent RAO Stage 2 – Mild RAO Stage 3 – Moderate RAO Stage 4 –
No RAO Severe RAO
(Total (5 6 Total Score 6 7) (8 6 Total Score 6 9) (10 6 Total Score 6 11) (Total
Score < 5) Score P 12)

a
All grading systems suggested comprise four grades, except for endoscopic grading which was adapted from previously published data where six grades were considered.
b
Adapted from Gerber et al. (2000).
c
Adapted from Gerber et al. (2004).

using with 4–5 prefilled numbered 100 ml syringes, through a stop- from each numbered syringe was sampled into two 10 ml EDTA
cock. Then fluid was aspirated with the same syringes. BAL fluid tubes equally numbered. These were immediately refrigerated
 P. Tilley et al. / Research in Veterinary Science 93 (2012) 1006–1014 1009

Table 2
CS, ES and BALFS – proposed classification of the variables not used for RAO staging.a

GRADE 0 1 2 3 4 5
CS
Exercise intolerance None Mild or very mild limitation Moderately or very Severely or completely
limited limited
Nasal discharge None Serous (scarse or abundant) Mucous (scarse) Mucous (abundant)
Respiratory rate 8–16 17–20 (up to 20% increase) 21–24 (up to 50% 25–36 (up to 125%
increase) increase)
Pulmonary Normal Audible at inspiration and exhalation With crepitation With crepitation and
auscultation (harsh sound) whisles
ES
Airway mucosal No hiperemia Mild hiperemia Moderate hiperemia Marked hiperemia Severe Extreme
hyperaemia hiperemia hiperemia
Thickness of the No thickness Less More
carinab
BALFc
Eosinophil % 0 0<x<4 46x<8 P8
Protein g/dl 0 0 < x < 0.4 0.4 6 x < 0.8 P0.8
a
All grading systems suggested comprise four grades, except for endoscopic grading which was adapted from previously published data where six grades were considered.
b
Adapted from Koch et al. (2007).
c
Values are expressed as a percentage (%) of a 400 nucleated cell count.

(4 °C) and taken to the laboratory within two hours of collection.
One microscope slide was made from each tube in order to have
two slides from each 100 ml syringe. To make the smear, 10 ml ali-
quots were centrifuged (300 g for 10 min) and the supernatant was
decanted. The pellet was then gently re-suspended at the bottom
of the tube with the remaining fluid and a drop was transferred to
a clean slide, using a plastic transfer pipette, to make the smear.
The slides were then quickly air-dried by fanning, fixated by metha-
nol (6 min) and stained with Giemsa’s azur eosin methylene blue
solution prepared with pH 7.2 buffer tablets (Merck8) (6 min) and
400 nucleated cells (1000 magnification) were counted by an expe-
rienced clinical pathologist who was blinded to the medical history,
physical exam and endoscopic exam results.
Three average differential nucleated cell percentages were
calculated, the initial (average of the differential nucleated cell per-
centages observed in the first half of syringes recovered), the final
(average of the differential nucleated cell percentages observed in
the second half of syringes recovered) and the total (average of the
initial and the final differential nucleated cell percentages)
differential nucleated cell percentages. As there was no statistical
difference between initial, final and total differential nucleated cell
percentages, the later (total differential nucleated cell percentages)
were used to classify BAL variables (Tables 1 and 2) and to calcu-
late Pearson correlation coefficients. This is in agreement with
Hoffman (2008) where all aliquots recovered are pooled for
evaluation.
2.6. Thoracic X-ray
An X-ray generator (850 mA s, 125KV; Philips9) Optimus with a
Vertical Bucky (Philips Bucky Diagnost) equipped with a grid and a
focus-film distance of 1.50 m were used. X-rays were obtained from
all horses (35  43 cm Kodak10 X-Omat cassettes with Lanex regular
screens Retina11 XOE green sensitive radiographic films). X-rays
were developed with a developing device (Curvix 60; Agfa12).
Exposure factors were in the range of 102–117 kV and
25–32 mA s for the four radiographic fields described by Mair
and Gibbs (1990).
8
Merck S.A., Rua Alfredo da Silva, 3C, 1300-040 Lisboa, Portugal.
9
Philips Healthcare, P.O. Box 10.000, 5680 DA Best, The Netherlands.
10
Eastman Kodak Company, Rochester, New York 14650.
11
Fotochemische Werke BMBH, Berlim, Germany.
12
Agfa Gevaert, N.V., Germany.
In the sedated horses X-rays were taken at end inspiration and
classified by two independent clinicians according to Table 3.
2.7. Statistical analysis/RAO staging
Classification of all variables for clinical, endoscopic, X-ray and
BALF grading is proposed in Tables 1 and 2. CS, XRS, ES and BALFS
were attributed based on these classifications. Pearson correlation
coefficients were determined between CS (cough, nostril flare,
abdominal lift, exercise intolerance, nasal discharge, respiratory
rate and pulmonary auscultation – adapted from Gerber et al.
(2000)) and ES (airway mucus accumulation, localization, apparent
viscosity and colour, airway mucosal hyperaemia, and thickness of
the carina – adapted from Gerber et al. (2004) and Koch et al.
(2007)), BALFS (neutrophils and eosinophils percentages and total
protein) and XRS (interstitial pattern, bronchial radiopacity,
bronchial thickening, tracheal thickening, loss of diaphragmatic
convexity, air bronchograms and peribronchial infiltrate).
Only correlation coefficients higher than 0.60 were considered
significant for a RAO staging method.
Description of the criteria for establishing the grades of RAO is
shown in Table 3. One way variance analyses were used to
compare the two groups (RAO and Control) and a discriminant
analysis model was adjusted on the RAO staging method sug-
gested. The discriminant analysis was done using the stepwise
method. Initially all clinical, X-ray, endoscopic and BALF variables
were considered. Stepwise analysis showed that with two discrim-
inant axis and only some of the variables (cough, nostril flare,
abdominal lift; airway mucus accumulation, localization, apparent
viscosity and colour; neutrophil percentage; interstitial pattern,
bronchial radiopacity, and bronchial and tracheal thickening) it is
possible to discriminate between the RAO stages. In the first axis,
cough, nostril flare, abdominal lift; airway mucus accumulation,
localization, apparent viscosity and colour; interstitial pattern,
bronchial radiopacity, and bronchial and tracheal thickening
showed a greater contribution, while neutrophil percentage played
a major role in the second axis.
3. Results
There was a high (>0.60) correlation coefficient between some
CS (cough, nostril flare and abdominal lift but not exercise
intolerance, nasal discharge, respiratory rate and pulmonary
1010 P. Tilley et al. / Research in Veterinary Science 93 (2012) 1006–1014

Table 3
XRS-ray grading-proposed classification of all variables (only selected variables with correlation coefficients higher than 0.60 are illustrated).
Interstitial pattern
0 1 2 3
No parenchymal opacity Mild increase of parenchymal opacity Moderate increase of parenchymal Increased parenchymal opacity with
loss of vizualization of vascular
opacity structures

Bronchial radiopacity
0 1 2 3
Normal Mild increase Moderate increase Severe increase

Tracheal thickening
0 1 2 3
No increase in thickening Mild increase in thickening Moderate oncrease in thickening Severe increase in thickening

Bronchial thickening
0 1 2 3
No increase in thickening Mild increase in thickening Moderate increase in thickening Severe increase in thickening

Loss of diafragmatic convexity

0 1 2 3
No loss of convexity Mild loss of convexity Clear loss of convexity Marked loss of convexity
Air bronchograms
0 1 2 3
No images 1 to 2 images 3 to 4 images 5 or more images
Peribronchial infiltrate
0 1 2 3
No images 1 to 2 images 3 to 4 images 5 or more images

auscultation) and all the mucus ES (airway mucus accumulation,
localization, apparent viscosity and colour, but not mucosal hiper-
emia and thickening of the carina) (0.61–0.84), some XRS (intersti-
tial pattern, bronchial radiopacity, bronchial thickening and
tracheal thickening, but not loss of diafragmatic convexity, air
bronchograms and peribronchial infiltrate) (0.67–0.78) and BALFS
neutrophil percentages (0.63–0.84) but not eosinophil percentages,
nor total protein (Tables 4 and 5).
Selected CS, XRS, ES and BALFS, using only the variables with
correlation coefficient higher than 0.60 are shown in Table 1. These
variables were considered relevant and chosen for a score model to
characterize different stages of RAO: stages 0, 1, 2, 3 and 4 (Table 1).
Table 4
Correlation amongst selected variables (correlation coefficients higher than 0.60) and between these variables and the RAO stages. Pearson’s correlation coefficient (r) and statistical significance (P) are shown.

VARIABLES RAO
stage
Endoscopic Score Bronchoalveolar lavage Clinical score Thoracic X-ray score
fluid score
Mucus Neutrophil% Cough Nostril Abdominal Interstitial Bronchial Tracheal Bronchial
flare lift pattern radiopacity thickening thickening
Accumulation Localization Apparent Colour
viscosity

P. Tilley et al. / Research in Veterinary Science 93 (2012) 1006–1014

Endoscopic score
Mucus
Accumulation NA 0.93 0.88 0.84 0.75 0.82 0.76 0.8 0.7 0.64 0.72 0.69 0.88
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Localization 0.93 NA 0.89 0.84 0.78 0.84 0.73 0.8 0.66 0.65 0.69 0.71 0.88
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Apparent viscosity 0.88 0.89 NA 0.93 0.74 0.75 0.68 0.77 0.62 0.7 0.61 0.64 0.81
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Colour 0.84 0.84 0.93 NA 0.67 0.68 0.61 0.66 0.66 0.66 0.61 0.62 0.77
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)

Bronchoalveolar lavage fluid score

Neutrophil % (400 0.75 0.78 0.74 0.67 NA 0.84 0.63 0.76 0.73 0.76 0.73 0.71 0.8
nucleated cell count)
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Clinical score
Cough 0.82 0.84 0.75 0.68 0.84 NA 0.83 0.84 0.76 0.73 0.69 0.78 0.89
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Nostril flare 0.76 0.73 0.68 0.61 0.63 0.83 NA 0.92 0.67 0.71 0.75 0.78 0.81
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) -0.001 (0.00) (0.00) (0.00) (0.00)
Abdominal lift 0.8 0.8 0.77 0.66 0.76 0.84 0.92 NA 0.67 0.67 0.68 0.76 0.88
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Thoracic X-ray score
Interstitial pattern 0.7 0.66 0.62 0.66 0.73 0.76 0.67 0.67 NA 0.75 0.67 0.71 0.74
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) -0.001 (0.00) (0.00) (0.00) (0.00) (0.00)
Bronchial radiopacity 0.64 0.65 0.7 0.66 0.76 0.73 0.71 0.67 0.75 NA 0.7 0.88 0.78
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Tracheal thickening 0.72 0.69 0.61 0.61 0.73 0.69 0.75 0.68 0.67 0.7 NA 0.76 0.76
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Bronchial thickening 0.69 0.71 0.64 0.62 0.71 0.78 0.78 0.76 0.71 0.88 0.76 NA 0.78
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)

NA = Not applicable; values in parenthesis are P values.

1011
1012 P. Tilley et al. / Research in Veterinary Science 93 (2012) 1006–1014
Variables with correlation coefficient lower than 0.60 were not ta-
ken into account for RAO staging.
The 10 healthy controls were attributed stage 0 and the 30 RAO
horses were attributed stages according to their scores: stage 1
(4 horses), stage 2 (7 horses), stage 3 (10 horses) and stage 4
(9 horses). There was a high correlation coefficient between all
the relevant variables and the RAO stage (0.74–0.89).
Discriminant analysis of the RAO staging method was validated
originating a classification matrix with 92.5% of original grouped
cases correctly classified and 85.0% of cross-validated grouped
cases correctly classified, having confirmed a major contribution
of the same variables that had significant correlation coefficients
(Fig. 1).
Furthermore, the RAO stages attributed showed Pearson corre-
lation coefficients above 0.30 with some other investigations car-
ried out on these horses and which were not part of the criteria
used to score RAO severity and which were not described in detail
in this paper. These were namely the SPT wheal diameter (0.65),
the ECG heart rate (0.44) and QRS duration (0.43) and the EcoCG
fractional shortening (0.34), tricuspid valve diameter (0.40) and
pulse wave flow velocity (0.33).
The respiratory symptoms in the RAO affected horses were re-
duced after effective allergen eviction measures were imple-
mented (eviction of the allergens identified by SPT for each
horse) as described by Tilley et al. (2010).
4. Discussion
In our study, we examined the relationship between CS and ES,
XRS and BALFS and selected the variables in each score which
showed higher correlation coefficients.
The severity of the clinical variables – cough, nostril flare and
abdominal lift –, corresponded to the respiratory endoscopy mucus
variables, the thoracic X-ray variables – interstitial pattern,
bronchial radiopacity, bronchial thickening and tracheal thicken-
ing –, and the BAL fluid cytology variable – neutrophil percentage
–, as evidenced by the high correlation coefficients. In general, the
use of Pearson correlation coefficients confirmed most of what has
been written in the phenotypic description of RAO (Robinson,
2001), the important criteria being cough, nostril flare, abdominal
lift, airway mucus accumulation, indices of bronchial changes on
radiography and BALF neutrophil percentage.
Robinson et al. (2000) found nasal flare and abdominal effort to
be significantly related to changes in respiratory function. In fact,
nasal flaring is a means of decreasing the resistance provided by
the upper airways, to counter the increasing resistance of the
tracheobronchial tree in RAO. Both nasal flaring and abdominal
effort reflect use of accessory muscles of breathing. In a study by
Benedice et al. (2008), BAL fluid neutrophilia was a common find-
ing in horses with RAO and the onset of cough coincided with an
increased number of neutrophils above the threshold of > 5%.
Exercise intolerance didn’t relate to cough, nor did airway mechan-
ical dysfunction. Robinson et al. (2003) reported coughing to be
strongly associated with airway inflammation and Burrell et al.
(1996) reported that cough is more prevalent when airway disease
has been present for two successive months and that nasal
discharge in the absence of coughing is not a particularly useful
predictor of the presence of lower airway disease. Additionally,
Robinson et al. (2000) found respiratory rate correlation coeffi-
cients to be biologically irrelevant.
These studies are generally in agreement with our CS results
that showed significant correlation coefficients for cough, nostril
flare and abdominal lift, but no significant correlation coefficients
for respiratory rate, nasal discharge, exercise intolerance and
pulmonary auscultation.
Fig. 1. Discriminant analysis model adjusted on the RAO staging method suggested
(canonical discriminant functions). Black square – Mean discriminant score (group
centroid).
Furthermore, Costa et al. (2000) found a high correlation of clin-
ical score with mucus accumulation in the airways in horses af-
fected with summer-pasture-associated obstructive pulmonary
disease. Peribronchial inflammatory infiltrate also correlated with
percentage of neutrophils in BALF of affected horses. Robinson et
al. (2003) found a strong correlation between cough frequency
and mucus score. Neutrophil percentage in BAL fluid was also sig-
nificantly correlated with mucus score in RAO-affected horses.
Koch et al. (2007) showed thickness of the carina to be irrelevant
as a single measure when attempting to differentiate individuals
with RAO from clinically normal animals through endoscopy.
In our study there were high correlation coefficients for all the
mucus ES and for neutrophil percentage in BALFS. Monteseirín et
al. (2007) presented a novel view of the neutrophil as a cell partici-
pating in allergic processes, namely by being able to produce and re-
lease several proteins, including eosinophil cationic protein (ECP),
by an allergen-specific, IgE-dependent, mechanism. Additionally,
these authors speculate that the neutrophil might be an eosinophil-
regulatory cell in IgE-dependent hypersensitivity reactions.
Although in our study RAO horses showed a higher percentage
of eosinophils in BAL (mean 2.6 ± 2.3%, ranging from 0 to 10%) in
comparison to control horses (0.15 ± 0.2%, ranging from 0 to
0.5%), these data did not show a high correlation coefficient with
the other parameters analysed. Moore et al. (1995) and Hare and
Viel (1998) described BAL fluid eosinophilia in inflammatory air-
way disease (IAD) and hypothesised it to be allergic in nature
and potentially an early manifestation in the natural progression
of RAO. Furthermore, they considered that the characterization of
RAO as solely a neutrophil-mediated disorder was likely to be erro-
neous. Riihimäki (2008) observed a transient pulmonary eosino-
philia during the summer months in young horses. Nevertheless,
scant attention has been paid to pulmonary eosinophilia in RAO
(Riihimäki, 2008). In the International Workshop on Equine
Chronic Airway Disease held in Michigan State University, USA,
in June 2000 (Robinson, 2001) it was concluded that there was
accumulating evidence to support an allergic component in the
aetiology of RAO. The ‘‘allergy’’ theory was based on the fact that
a subset of predisposed horses develop RAO when exposed to
moldy hay. RAO likely results from complex interaction between
innate and acquired immune responses, environment and genetic
susceptibility, which may represent divergent pathways possibly
leading to a common phenotype (Lavoie, 2009).
 P. Tilley et al. / Research in Veterinary Science 93 (2012) 1006–1014 1013

Table 5
Relationship between the variables and the RAO stages and mean and standard deviation of all variables for both RAO-affected and control horses.

RAO stages Control RAO-affected

horses horses
Stage 0 (n = 10) Stage 1 (n = 4) Stage 2 (n = 7) Stage 3 (n = 10) Stage 4 (n = 9) Mean ± SD Mean ± SD
VARIABLES
Endoscopic scores
Mucus
Accumulation 1.0 ± 0.58 2.5 ± 0.71 3.7 ± 1.15 4.4 ± 0.53 4.5 ± 0.53 1.0 ± 0.58 3.8 ± 1.06
Localization 0.8 ± 0.42 2.5 ± 0.58 3.3 ± 1.38 4.3 ± 0.52 4.4 ± 0.53 0.8 ± 0.42 3.8 ± 1.03

Apparent viscosity 1.8 ± 1.14 3.0 ± 0.00 4.0 ± 0.82 4.6 ± 0.53 4.7 ± 0.48 1.8 ± 1.14 4.3 ± 0.78
Colour 1.1 ± 0.74 2.1 ± 0.00 2.1 ± 0.90 2.8 ± 0.42 2.9 ± 0.33 1.1 ± 0.74 2.6 ± 0.63
Bronchoalveolar lavage fluid
Neutrophil % (400 nucleated cell count) 3.2 ± 1.44 67.0 ± 5.18 70.2 ± 7.46 70.5 ± 17.68 71.2 ± 7.26 3.2 ± 1.44 66.2 ± 10.59
Clinical scores
Cough 0.0 ± 0.00 1.5 ± 0.58 1.9 ± 0.38 2.3 ± 0.48 2.6 ± 0.53 0.0 ± 0.00 2.2 ± 0.59
Nostril flare 0.0 ± 0.00 0.8 ± 0.96 1.3 ± 0.95 1.9 ± 0.57 2.2 ± 0.67 0.0 ± 0.00 1.7 ± 0.88
Abdominal lift 0.0 ± 0.00 1.3 ± 0.50 1.7 ± 0.49 2.0 ± 0.47 2.6 ± 0.73 0.0 ± 0.00 2.0 ± 0.72
Thoracic X-ray scores
Interstitial pattern 0.6 ± 0.52 1.5 ± 0.58 2.2 ± 0.63 2.3 ± 0.49 2.4 ± 0.73 0.6 ± 0.52 2.2 ± 0.66
Bronchial radiopacity 0.6 ± 0.52 1.5 ± 0.58 1.9 ± 0.69 2.0 ± 0.47 2.4 ± 0.53 0.6 ± 0.52 2.0 ± 0.61
Tracheal thickening 0.4 ± 0.52 1.6 ± 0.53 2.3 ± 0.50 2.4 ± 0.70 2.7 ± 0.50 0.4 ± 0.52 2.2 ± 0.73
Bronchial thickening 0.3 ± 0.48 1.7 ± 0.49 1.9 ± 0.57 2.0 ± 0.00 2.7 ± 0.50 0.3 ± 0.48 2.1 ± 0.61

n = Number of measurements; SD = standard deviation.

This study shows that in this selected population, all horses
with RAO had several positive SPT results to common aeroaller-
gens, strongly suggesting IgE-mediated hypersensitivity reactions
and the role these reactions may play in RAO (Tahon et al., 2009;
Tilley et al., 2010). Furthermore, we had clinical confirmation of
a reduction of the respiratory symptoms after aeroallergen evic-
tion, the symptoms with more marked improvement being cough,
nasal flare and abdominal lift. Therefore, the SPT were of value in
identifying possible triggers of allergy, in an attempt to minimize
environmental causes of RAO (Tilley et al., 2010).
Although the most important role of thoracic X-ray in the inves-
tigation of RAO is to rule out other types of lung disease (Robinson,
1997), there can be a correlation between radiographic alterations
(namely unstructured interstitial and structured bronchiolar inter-
stitial patterns) and the severity of the clinical signs in RAO (Bakos,
2008). Even though radiographic appearance of the lungs can vary
greatly depending on factors like body fat of the horse and the type
of radiographic equipment being used, the authors chose to include
it in the scoring system proposed here. In fact radiographic equip-
ment stays with each clinician for rather a long time, allowing for a
solid knowledge of its technical characteristics and image interpre-
tation. Therefore, provided that each clinician always uses the
same equipment and carries out and interprets the radiological
exam in the same way, variability of X-ray scores should present
a minor concern.
Koch et al. (2007) suggested that thickening of smaller bron-
chial bifurcations rather than tracheal thickening may provide a
better index of RAO as inflammation may be more severe in the
smaller airways, which would justify the inclusion of the variables
bronchial radiopacity and bronchial thickening in the staging
method proposed here. We found significant correlation coeffi-
cients for the XRS interstitial pattern, bronchial radiopacity, bron-
chial thickening and tracheal thickening.
Although some studies have been published over the years,
there is still a need to put relevant information together to estab-
lish a staging method for RAO in horses in order to avoid underdi-
agnosing and to be able to correctly stage the disease.
For this purpose, we attempted to establish the relevance of
each variable as a contribution for the characterization of different
stages of RAO and we suggested a score model for staging RAO
horses.
The staging method suggested here showed a very high per-
centage of correctly classified cases (discriminant analysis), having
confirmed a major impact of the same variables that had high cor-
relation coefficients, having the potential to contribute to the diag-
nosis of RAO in horses.
The validity of this staging method was evaluated so far in a
specific group of horses and has yet to be further tested in a
wider population. Still, it correlated well with the results of data
which were not used to score RAO severity (SPT, ECG and EcoCG
data).
This newly developed RAO staging system should now be tested
based on a gold standard diagnosis, namely objective lung func-
tional testing of each RAO stage (in remission and in exacerbation)
and of the control group.
5. Conclusions
In our population, RAO was best characterized by clinical
evaluation of cough, nostril flare and abdominal lift; endoscopic
evaluation of airway mucus accumulation, apparent viscosity,
localization and colour; X-ray evaluation of interstitial pattern,
bronchial radiopacity, bronchial and tracheal thickening; and BALF
cytology neutrophil percentage. The significant correlation coeffi-
cient between all these variables and the RAO stage, as well as
the results of the discriminant analysis of the RAO staging, support
the suggestion of a score model for the different stages of RAO.
Further study is now needed to determine the clinical value of
this RAO staging method in a larger population in order to intro-
duce it as a diagnostic tool, similar to the bronchitis index tested
by Thompson et al. (1993) in human medicine.
Horses frequently change owners and often little is known
about their clinical history. The possibility of attributing a stage
to RAO-affected horses gives us invaluable information, which is
a major aid in choosing the treatment scheme to use, managing
to increase its efficiency. Furthermore, it gives us thorough infor-
mation, allowing us to establish the prognosis of the horse, which
is always important for owners. Being able to diagnose a RAO stage
1 may imply the possibility of preventing the disease from evolving
and, on the other hand, diagnosing a RAO stage 4 may mean that
poor response to treatment is expected.
1014 P. Tilley et al. / Research in Veterinary Science 93 (2012) 1006–1014
Conflict of interest
There is no conflict of interest.
Acknowledgements
This work was supported by FMV (Faculty of Veterinary Medi-
cine of Lisbon) and CIISA (‘Centro Interdisciplinar de Investigação
em Sanidade Animal’ – Interdisciplinary Centre of Research in Ani-
mal Health), FMV’s branch of FCT (‘Fundação para a Ciência e Tecn-
ologia’). The authors are grateful to the horses and their owners, to
Dr. Bruno and Dr. Sérgio from the Pathology Laboratory, Dr. Joana
Simões from the equine hospital, and Eng. Marta Vacas Carvalho,
specialist in statistics, for making this project possible.
References
Bakos, Z., 2008. Digital luminescence thoracic radiography in horses with recurrent
airway obstruction. Vet. Rec. 162 (4), 122–124.
Benedice, D., Mazan, M.R., Hoffman, A.M., 2008. Association between cough and
cytology of bronchoalveolar lavage fluid and pulmonary function in horses
diagnosed with inflammatory airway disease. J. Vet. Int. Med. 22 (4), 1022–1028.
Bracher, V., von Fellenberg, R., Winder, C.N., Gruening, G., Hermann, M.,
Kraehenmann, A., 1991. An investigation of the incidence of chronic
obstructive pulmonary disease (COPD) in random populations of Swiss
horses. Equine Vet. J. 23 (2), 136–141.
Burrell, M.H., Wood, J.L.N., Whitwell, K.E., Chanter, N., Mackintosh, M.E., Mumford,
J.A., 1996. Respiratory disease in thoroughbred horses in training: the
relationship between disease and viruses, bacteria and environment. Vet. Rec.
139 (13), 308–313.
Costa, L.R.R., Seahorn, T.L., Moore, R.M., Taylor, H.W., Gaunt, S.D., Beadle, R.E., 2000.
Correlation of clinical score, intrapleural pressure, cytologic findings of
bronchoalveolar fluid, and histopathologic lesions of pulmonary tissue in
horses with summer pasture-associated obstructive pulmonary disease. Am. J.
Vet. Res. 61 (2), 167–173.
Couëtil, L.L., Rosenthal, F.S., DeNicola, D.B., Chilcoat, C.D., 2001. Clinical signs,
evaluation of bronchoalveolar lavage fluid, and assessment of pulmonary
function in horses with inflammatory respiratory disease. Am. J. Vet. Res. 62
(4), 538–546.
Dixon, P.M., Railton, D.I., McGorum, B.C., 1995. Equine pulmonary disease: a case
control study of 300 referred cases. Part 1: Examination techniques, diagnostic
criteria and diagnosis. Equine Vet. J. 27 (6), 416–421.
Gerber, V., King, M., Schneider, D.A., Robinson, N.E., 2000. Tracheobronchial mucus
viscoelasticity during environmental challenge in horses with recurrent airway
obstruction. Equine Vet. J. 32 (5), 411–417.
Gerber, V., StrauB, R., Marti, E., Hauptman, J., Herholz, C., Kings, M., Imhof, A., Tahon,
L., Robinson, N.E., 2004. Endoscopic scoring of mucus quantity and quality:
observer and horse variance and relationship to inflammation, mucus
viscoelasticity and volume. Equine Vet. J. 36 (7), 576–582.
Hare, J.E., Viel, L., 1998. Pulmonary eosinophilia associated with increased
airway responsiveness in young racing horses. J. Vet. Int. Med. 12 (3),
163–170.
Hoffman, A.M., 2008. Bronchoalveolar lavage: sampling technique and guidelines
for cytologic preparation and interpretation. Vet. Clin. Equine 24, 423–435.
Koch, C., Straub, R., Ramseyer, A., Widmer, A., Robinson, N.E., Gerber, V., 2007.
Endoscopic scoring of the tracheal septum in horses and its clinical relevance
for the evaluation of lower airway health in horses. Equine Vet. J. 39 (2), 107–
112.
Künzle, F., Gerber, V., Van der Haegen, A., Wampfler, B., Straub, R., Marti, E., 2007.
IgE-bearing cells in bronchoalveolar lavage fluid and allergen-specific IgE levels
in sera from RAO-affected horses. J. Vet. Med. A 54 (1), 40–47.
Laumen, E., Doherr, M.G., Gerber, V., 2010. Relationship of horse owner assessed
respiratory signs index to characteristics of recurrent airway obstruction in two
Warmblood families. Equine Vet. J. 42 (2), 142–148.
Lavoie, J.-P., 2009. Heaves in horses: allergy or non-specific response to
environmental antigens? In: Proceedings of 11th Geneva Congress on Equine
Medicine and Surgery, Switzerland.
Mair, T., Gibbs, C., 1990. Thoracic radiography in the horse. In Pract. 12 (1), 8–10.
Monteseirín, J., Vega, A., Chacón, P., Camacho, M.J., El Bekay, R., Astúrias, J.A.,
Martínez, A., Guardia, P., Pérez-Cano, R., Conde, J., 2007. Neutrophils as a novel
source of eosinophil cationic protein in IgE-mediated processes. J. Immunol. 179
(4), 2634–2641.
Moore, B.R., Krakowka, S., Robertson, J.T., Cummins, J.M., 1995. Cytologic evaluation
of bronchoalveolar lavage fluid obtained from standardbred racehorses with
inflammatory airway disease. Am. J. Vet. Res. 56 (5), 562–567.
Ramseyer, A., Gaillard, C., Burger, D., Straub, R., Jost, U., Boog, C., Marti, E., Gerber, V.,
2007. Effects of genetic and environmental factors on chronic lower airway
disease in horses. J. Vet. Int. Med. 21, 149–156.
Riihimäki, M., 2008. Inflammatory response in equine airways – Cytokines in
bronchial epithelium and bronchoalveolar lavage cells. Doctoral thesis no.
2008:12. Faculty of veterinary medicine and animal science, Uppsala, Sweden.
Robinson, N.E., 1997. Pathogenesis and management of airway disease. In:
Proceedings of the annual convention of the AAEP, vol. 43, pp. 106–115.
Robinson, N.E., 2001. International Workshop on Chronic Airway Disease, Michigan
State University, U.S.A., June 2000. Equine Vet. J. 33 (1), 5–19.
Robinson, N.E., Olszewski, M.A., Boehler, D., Berney, C., Hakala, J., Matson, C.,
Derksen, F.J., 2000. Relationship between clinical signs and lung function in
horses with recurrent airway obstruction (heaves) during a bronchodilator trial.
Equine Vet. J. 32 (5), 393–400.
Robinson, N.E., Berney, C., Eberhart, S., deFeijter-Rupp, H.L., Jefcoat, A.M., Cornelisse,
C.J., Gerber, V.M., Derksen, F.J., 2003. Coughing, mucus accumulation, airway
obstruction, and airway inflammation in control horses and horses affected
with recurrent airway obstruction. Am. J. Vet. Res. 64 (5), 550–557.
Rossdale, P.D., Hopes, R., Wingfield Digby, N.J., et al., 1985. Epidemiological study of
wastage among racehorses 1982 and 1983. Vet. Rec. 116, 66–69.
Tahon, L., Baselgia, S., Gerber, V., Doherr, M.G., Straub, R., Robinson, N.E., et al., 2009.
In vitro allergy tests compared to intradermal testing in horses with recurrent
airway obstruction. Vet. Immunol. Immunopathol. 127, 85–93.
Thompson, A.B., Huerta, G., Robbins, R.A., Sisson, J.H., Spurzem, J.R., von Essen, S.,
Rickard, K.A., Romberger, D.J., Rubinstein, I., Ghafouri, M., Daughton, D.M.,
Rennard, S.I., 1993. The bronchitis index. Chest 103 (5), 1482–1488.
Tilley, P., Sales Luis, J.P., Branco Ferreira, M., 2010. Skin Prick Tests (SPT) use in
equine Recurrent Airway Obstruction (RAO). R.P.I.A. 18 (6), 561–584.