Pathology, infectious agents and horse- and management-level risk factors associated with signs of respiratory disease in Ethiopian working horses.

BACKGROUND
Respiratory disease is a common cause for presentation of working horses to clinics in Ethiopia and a priority concern for owners.


OBJECTIVES
Identify risk factors for and association of pathogens with respiratory signs in working horses. Study design Unmatched case-control study.


METHODS
Cases were those animals recently coughing (last 7 days) or observed with coughing, nasal discharge or altered respiration at the time of examination. A physical exam and respiratory endoscopy were performed including a tracheal wash sample to detect the presence of pathogens and serology performed on blood. An owner questionnaire was administered. Risk factors were determined using multivariable logistic regression.


RESULTS
Data on 108 cases and 93 unmatched control horses were obtained. Case horses often had underlying lower airway pathology and were significantly more likely to have S zooepidemicus detected (OR 12.4, 95% CI 3.6-42.4). There was no evidence of a major role for viral respiratory pathogens. Risk factors included completion of strenuous work (OR 2.7, 95% CI 1.2-6.3), drinking from stagnant water sources (OR 2.3, 95% CI 1.0-5.2) or being housed on a cobbled floor (OR 2.0, 95% CI 1.1-3.8). There were increased odds of respiratory disease in young and old horses in this population. Main limitations Samples for pathogen detection and cytology were only taken from the trachea.


CONCLUSION
S zooepidemicus, a common commensal, may play a role in clinical respiratory disease in this population.


| INTRODUC TI ON
Respiratory disease is a major cause of time off work and poor performance for equids. In Ethiopian working horses, coughing and nasal discharge are frequently identified at clinics (SPANA clinic data, unpublished). They are a priority health concern for owners 1 and are often 'poorly defined'. 2 Limited diagnostic options mean a presumptive diagnosis is typically based on clinical history and examination alone.
Serological surveys from this study area suggested a role for Streptococcus equi subspecies equi (S. equi) 3 and alphaherpesviruses (equine herpes virus (EHV) 1/4)). 4,5 Endemicity of these pathogens occurs in other equine populations and causality in respiratory disease is well established. 6,7 There was no serological evidence of equine influenza virus (EIV) in a previous study, 3 but given its global prevalence and potential impact in a naïve population, its presence was further examined here. Additionally, the picornaviruses, equine rhinitis virus A (ERAV) and B (ERBV), were recognised as frequent contributors to respiratory disease, 8,9 and the role of lesser characterised pathogens of the equine airways was considered in this study given the limited data on respiratory disease in sub-Saharan Africa equines. The gammaherpesviruses EHV-2 and -5 are ubiquitous in equine respiratory secretions, but evidence for their causal role in disease is conflicting. 10,11 Streptococcus zooepidemicus, commonly considered a commensal, has also been frequently linked with clinical signs and outbreaks of respiratory disease 12,13 and was therefore investigated here.
Noninfectious respiratory disease includes a spectrum of asthma often presenting with coughing, exercise intolerance and nasal discharge. Severe equine asthma (previously recurrent airway obstruction, RAO) is widely recognised and is likely a hypersensitivity response to inhaled organic dusts where, in contrast to milder forms of equine asthma, an increased respiratory effort is apparent at rest.
Mild forms of equine asthma (previously inflammatory airway disease, IAD) tend to affect younger horses. Presentation may be subclinical with a variable inflammatory profile in airways. 14 This case-control study aimed to investigate the underlying pathology in airways of horses with overt clinical signs; identify any horse-or management-level factors that may increase the odds of disease; and determine associations between infectious agents and respiratory signs using molecular detection of organisms in lower airways and serological evidence for exposure to known pathogens of the equine respiratory tract. This work will aid the therapeutic decisions taken by veterinarians in Ethiopia, and the advice provided for prevention of disease in working equids.

| MATERIAL S AND ME THODS
Case and control horses were identified at four SPANA clinic locations (two static and two mobile) in Oromia region, Ethiopia, between September and December 2014. Horses reported with recent coughing (last 7 days) or observed coughing, with nasal discharge or altered respiration (rate and/or character) at examination were defined as cases. Controls were unmatched and systematically selected as the next horse presenting without a history or signs of respiratory disease that did not require immediate attention on welfare grounds. Horses with suspected African Horse Sickness, Epizootic Lymphangitis, Glanders, severe pleuropneumonia and where epistaxis was the only sign were excluded from the study. A sample size of 94 cases and 94 unmatched controls was estimated 15 to detect an odds ratio of 2.5 or greater with exposures of 20% or more, with 80% power and 95% confidence intervals. Participants were fully informed and gave verbal consent for inclusion in the study and were present throughout the procedure which they could ask to be stopped at any time. Owners also completed an individual verbal questionnaire (Data S1) in either Amharic or Afan Oromo (TA).
Clinical examination and endoscopy of the respiratory tract (to the carina) were performed by a single veterinarian (G.L.). Due to the requirement for horses to work after examination, sedative agents were not administered but a humane nose twitch was applied for a maximum 20 minutes limiting examination to a single naris insertion route, single sample collection point and brief inspection of URT. Full ethical approval was granted with abandonment of examination if animals showed any signs of distress. Secretions at the 'sump' of the trachea were sampled by flushing 20 mL of sterile phosphate-buffered saline (ambient temperature), via a sterile endoscopic catheter.
The endoscope underwent a full cleaning routine between animals with biological detergent followed by 20-minute submersion in glutaraldehyde disinfectant.
Fixed slides of tracheal wash sediment were imported to the UK (TARP/2014/163), stained with May-Grunwald-Giemsa and evaluated by a single board-certified clinical pathologist (F.C.) using methods adapted from Whitwell and Greet (1984). 16 Table 1 and Table S1.

| Data analysis
Associations between clinical, endoscopic and cytological variables and case-control status were assessed using univariable logistic regression analyses. Secondly, horse-, owner-and managementlevel factors were investigated for associations with case-control status using univariable then multivariable logistic regression models. Continuous variables were assessed for linearity using generalised additive models (GAMs). All variables with P < .25 were assessed for correlation before inclusion in a multivariable model.  and Hosmer-Lemeshow statistic was used for goodness of fit. Deltabeta values and influence plots were used to identify the sensitivity of the model to individual observations. Finally, pathogen associations with case-control status were also investigated using logistic regression but adjusted for age as a potential confounder as this demonstrated a biologically plausible and statistically significant association and may also be associated with pathogen exposure. All analysis was performed using R (version 3.1.2, R Foundation for Statistical Computing).

| RE SULTS
Clinical examination and tracheal endoscopy were performed on 108 cases and 93 unmatched controls with nine of those invited to participate declining (this initial communication was done in the local language hence, exact reasons for declining were not recorded) and nine examinations were abandoned for behavioural reasons. Cases presented with one or more signs including: coughing (89% historical and 36% during examination), 54% nasal discharge (48% serous, 52% mucoid or mucopurulent) and one epistaxis episode. A fifth (19%) of cases also showed altered respiratory signs. The majority of cases (87%) were being examined for the first time and participants reported reduced work ability in 54%. The majority of controls presented for preventive treatments or administrative documentation (75%) and lameness (23%). All animals examined were male and breed type did not vary across sample sites. Nearly all cases (96%) and controls (99%) were used as cart horses and presented by the owner (98%).
Univariable analyses of clinical and endoscopic findings are presented in Tables 2 and 3 and Figures S1 and S2. Case horses were significantly more likely to have abnormal lung sounds, increased tracheal mucus and increased neutrophil count and decreased lymphocyte proportions in tracheal wash cytology (with reference to the most widely accepted limits for normal). In addition, case horses were more likely to have intracellular bacteria in respiratory secretions. Both cases (29%) and controls (39%) frequently showed increased respiratory eosinophils but there was no significant difference in cases or controls.
Univariable analysis results of horse, owner and management variables are presented in Table S2 and S3 and the final multivariable model is shown in Table 4. The association of cases with clinic location or date examined was initially assessed as a fixed effect in a univariable logistic regression model but neither proved significant. Age of horse was significant (P = .02) and a GAM for estimated age demonstrated a nonlinear relationship with case status ( Figure 1); a quadratic polynomial provided the best fit, suggesting young horses and older horses were at increased risk, with the lowest risk for horses 9-10 years. Exercise intensity was significantly associated with respiratory cases, with horses performing strenuous exercise at nearly three times greater risk. Note: Due to apparent discordance between the Ethiopian results and published reference intervals, the lack of obvious cut-offs for normal and to avoid loss of data, cardiorespiratory and haematological parameters were analysed as continuous variables.
Abbreviations: IQR, inter-quartile range; CI, confidence interval. a Packed cell volume (PCV) and total plasma protein (TPP) were measured manually. Note: Clinical, endoscopic and cytological variables are all likely to be effects of, rather than causes of, underlying respiratory disease and as such were not assessed as potential risk factors in multivariable regression. Missing observations for some aspects of examination due to recording errors or missing data due to horse temperament. Eight tracheal wash samples had insufficient cells to perform a differential cell count.
Abbreviation: CI, confidence interval. ref, reference catgeory a There are no published accepted reference intervals for tracheal wash differential cell counts, so values used in this study were those of Richard et al, 42 who considered a range of studies to establish cut-offs. b Strong neutrophilic inflammation can make it difficult to assess other cell lines accurately.
d Haemosiderophage, an indicator of previous bleeding in the lungs, was a rare finding and was noted in just one control horse.

TA B L E 3 (Continued)
However, neither were significantly associated with respiratory signs in this study.

| D ISCUSS I ON
This is the first study to examine the association between infectious agents and signs of respiratory disease in working horses of Ethiopia. There was a significant association of S. zooepidemicus with respiratory cases in our study. Although S. zooepidemicus is considered a commensal of the URT, and the method of sampling in this study could be vulnerable to contamination from the URT, the presence of intracellular bacteria on tracheal smears supports a true septic inflammation in the LRT rather than a transient colonisation.
Furthermore, any URT contamination due to sampling methodology would affect cases and controls equally.
Streptococcus zooepidemicus in the LRT has been implicated as a primary pathogen in respiratory disease 12 and linked to equine asthma (mild) and subclinical respiratory disease. 19 attributes. 13 It is unlikely that immunity to one strain type is cross-protective to infection with others, and therefore future work should quantify S. zooepidemicus in the LRT and examine strain types.
Serostatus for S. equi, the causative agent of strangles, was not significantly associated with signs of respiratory disease (P = .09); however, 11% of controls and nearly a fifth of case horses were seropositive, hence this may be due to low power to detect an as- ERBV was detected in equal numbers of cases and controls, and in cases it was only found as a co-infection with other respiratory pathogens. There is conflicting evidence for the role of ERBV in respiratory disease, 23   to increased mucus secretion that may subsequently lead to coughing. Of those animals with increased tracheal mucus, over half were seen coughing during examination and the remainder all had a recent history of coughing in the last week. These results are in agreement with other studies that coughing is a specific indicator for tracheal mucus 25 and further supports the potential usefulness of owner reporting in this population. 26 Neutrophilia in respiratory secretions is a common occurrence in equine asthma, found in both mild and severe phenotypes, but also occurs in septic inflammation due to primary or secondary bacterial colonisation. 21 Colonisation can occur with impairment of the mucociliary clearance seen in equine asthma, or following viral insult. Septic inflammation usually presents with a high percentage of neutrophils, a frequent finding in this study, and cases showed significantly more intracellular bacteria in secretions but the presence of degenerative cells, usually found with septic inflammation, was rare.
Nevertheless, 20% of cases had neither increased tracheal mucus nor neutrophilia, suggesting that respiratory disease may involve other pathologies, or that some animals were incorrectly classified as cases. Furthermore, an absence of clinically apparent signs did not rule out airway inflammation, 19 and some control horses were found to have abnormal cytology results, inclusion of which could bias results towards the null if misclassified.
Eosinophilia was a frequent finding in cases but also in controls and was not associated with presentation for respiratory disease, as found in other studies. This may suggest subclinical airway disease or severe asthma during remission. 27 Eosinophilia may also be associated with lungworm, but it was not possible to assess prevalence in this study.
In this population, there was increased likelihood of younger and older animals presenting as cases, an age pattern found in other populations. 28 where increasing immunity to pathological challenges, or a build-up of tolerance to other aetiological agents occurs as horses reach adulthood. The risk in older animals may reflect the development of hypersensitivities and chronic respiratory disease, as seen in ageing populations elsewhere. 25,26 Other risk factors included a cobbled floor ('flat rocks') in housing.
Stabling has been identified as a risk factor for disease through increased exposure to dust and allergens especially in straw or hay, but no participants reported using bedding of any sort or feeding large quantities of hay. Increased odds could be associated with difficulties keeping cobbles clean but may also be a proxy for another management practice not identified, and this warrants further investigation.
Horses that were described by owners as performing strenuous work were at increased risk of disease in this study. High intensity or strenuous exercise can lead to airway contamination and immunocompromise, 29 and has been identified as a risk factor for disease in other populations. 30 Additionally, strenuous exercise in a dusty environment can lead to mucosal damage, and dust storms have also been associated with increased invasive bacteria in the LRT of humans. 31 Dust exposure during work was not measured during this study, but horses commonly worked on dirt roads in dry environmental conditions. It is also possible that respiratory signs may be more apparent in animals exercising intensely and therefore, they may be more likely to attend clinics as a case.
Stagnant water has not been previously identified as a risk factor for equine respiratory disease and a direct causal link cannot be determined from the current study and it may be a proxy for unmeasured variables. However, stagnant water may be contaminated with potentially harmful organisms, such as S. equi, which can survive up to a month in drinking water. 32 The broad case definition criteria in this unmatched case-control study could represent a range of aetiologies, potentially making the identification of syndrome-specific risk factors more challenging. Controls were selected at the regional clinics, hence it is also possible that they were not 'healthy' and could have been suffering from other disease leading to bias. Although respiratory disease in working horse populations is recognised as a problematic syndrome, 2 there is a lack of published information to compare these results to and they may be specific to this region of sub-Saharan Africa.
In conclusion, horses presenting with coughing, nasal dis-

E TH I C A L A N I M A L R E S E A RCH
Ethical approval for this study was granted by the Ethics Committee at the University of Liverpool (VREC143) and University Addis Ababa (VM/ERC/001/06/014).

OWN ER I N FO R M ED CO N S ENT
Owner consent for participation was obtained verbally due to assumed high levels of illiteracy.

DATA ACC E S S I B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

ACK N OWLED G EM ENTS
Elizabeth Medcalf and Toni-Ann Hammond at the Animal Health Trust, and Shirley Bonner at University of Liverpool for their guidance and assistance in the laboratory. SPANA Ethiopia team for their support in the field and we thank all the participating horse owners who contributed information for this study.

CO N FLI C T O F I NTE R E S T S
No competing interests have been declared. All authors contributed to writing the manuscript.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/evj.13339.