Training related risk factors for exercise induced pulmonary haemorrhage in British National Hunt racehorses

Background: Exercise induced pulmonary haemorrhage (EIPH) is an important condi-tion of horses performing high intensity exercise, with reported prevalence among racehorses of up to 95%, based on the detection of blood on tracheobronchoscopy. Previously identified risk factors include age, sex, season, race type, years spent in racing and lower airway inflammation. Objectives: To estimate the prevalence of EIPH in British National Hunt racehorses as indicated by two outcome measures: presence of tracheal blood on tracheobronchoscopy, and presence of moderate- large (significant) proportions of haemosiderophages in tracheal wash (TW) fluid; and to identify training- related risk factors for these indicators of EIPH. Study design: Prospective longitudinal study. Methods: Data from tracheobronchoscopy and TW cytology were analysed using univariable and multivariable mixed- effects logistic regression. Results: of

The presence of blood in the airways leads to further airway inflammation. 18 The impact-induced trauma theory proposes that the force following forelimb ground strike is transmitted via the scapula and ribs to the thorax where pressure waves are reflected along the spinal and diaphragmatic walls, leading to shearing forces within the lung parenchyma. The level of force is considered sufficient to cause oedema and haemorrhage within the lung parenchyma. 17 Accumulated time in training may be a more important risk factor for EIPH than age. It has been theorised that increased pulmonary pressures during exercise leads to venous remodelling and suggested that the degree of this remodelling may depend on genetic susceptibility combined with frequency of highpressure events occurring during strenuous exercise. 19 The current study examined a population of NH racehorses with different training histories. Some entered training for the first time as 4-year-olds while others entered NH training from flat racing, where training begins at around 2 years of age. The aim was to describe EIPH in NH racehorses. The objectives were to estimate the prevalence of EIPH as indicated by two outcome measures: presence of tracheal blood and presence of moderate-large (significant) proportions of haemosiderophages in TW fluid; and to identify trainingrelated risk factors for these outcome measures. The hypothesis was that the occurrence of EIPH would increase with time in training.

| MATERIAL S AND ME THODS
A prospective longitudinal study followed 177 healthy NH racehorses in training on five training yards over two seasons (August 2003-April 2004. 20 Yards were enrolled based on willingness to participate and being within 4 hours drive of the diagnostic laboratory. Random sampling from 3 strata was used to recruit around 15 horses from each yard at the start of each season. The strata were: (1)

K E Y W O R D S
EIPH, horse, National Hunt, racehorse was considered medium-impact work and gallop and jumping as highimpact work. Stabled horses were those sampled prior to work or having a rest day. Any nasal discharge was recorded as present or absent immediately prior to tracheobronchoscopic examination. Information about current medications and those administered within the previous month were recorded from yard staff.
Examination and sampling of horses took place within 90 minutes following exercise for horses exercised on the day of sampling.
A standardised protocol was used. Tracheobronchoscopy, to the level of the carina, was performed without sedation using a flexible endoscope (1.4 m or 2 m). The presence and amount of visible tracheal mucus was recorded and scored from 0 to 3, where 0 = none, 1 = small amounts (isolated flecks), 2 = moderate amounts (multiple larger blobs) and 3 = large amounts, at least partly confluent. 22 The presence of blood was recorded and scored in the same way. These scores were collapsed into binary variables indicating presence (score 1-3) or absence (score 0). A TW sample was collected transendoscopically, in a standard manner, from each horse undergoing tracheobronchoscopy. 30 mL of phosphate buffered saline was instilled into and aspirated from the distal trachea using a sterile polyethylene catheter. A portion of the retrieved sample was combined with an equal volume of 10% formal saline for cytological examination. Cytological assessments were performed using routine diagnostic methods adapted from those described by Whitwell and Greet (1984) 23 , with samples stained with haematoxylin and eosin. 20 Neutrophils were recorded as a proportion of total nucleated cells and categorised as 0%-10%, 11%-25%, 26%-50% and greater than 50%. Proportions of macrophages containing haemosiderin (haemosiderophages) were scored as none (0%), occasional (<10%), small proportions (10%-25%), moderate proportions (25%-50%) or large proportions (>50%). Moderate and large proportions were classified as significant haemosiderophages, likely indicating a substantial recent bleed, for analysis.
Auto-regressive variables were created from mucus, blood, neutrophil and haemosiderophage variables to indicate the status of these variables the previous month. Where observations from the preceding month were missing, including the first months' observations, auto-regressive variables were classified as missing.
The following were used as indicators of EIPH and recent EIPH, respectively: • Tracheal blood: the presence of grossly visible blood on tracheobronchoscopy.
• Significant haemosiderophages: the presence of moderate and large proportions of haemosiderophages in TW fluid.

| Data analysis
Descriptive statistics summarised horse characteristics and number of tracheobronchoscopic examinations. The sample-level and horse-level prevalence of tracheal blood and haemosiderophages in TW fluid were calculated, with 95% confidence intervals (CIs).
Distributions of continuous variables were checked for normality using histograms and Shapiro-Wilk test statistic and summarised accordingly, using median and interquartile range (IQR). Linearity of associations between continuous exposure variables and outcomes were assessed by examining plots of the log odds, in conjunction with Wald tests and likelihood ratio tests.
Univariable analysis with mixed-effects logistic regression, including individual horse identification (ID) as a random effect and yard as a fixed effect to account for unmeasured variation between horses and yards respectively, identified exposure variables associated with each outcome. A P-value <.2 was set for inclusion in further analysis.
Multivariable mixed-effects logistic regression was conducted for each outcome measure, with yard retained as a fixed effect and horse ID as a random effect, irrespective of significance. A forward stepwise manual selection process was used to identify which exposure variables should remain in the final models. The likelihood ratio test (LRT) was used to compare models and a LRT P-value of P < .05 used as the threshold for retention in the final models. Crude and adjusted odds ratios (ORs) were compared to assess for confounding.

| Univariable analysis
The variables identified by univariable analysis for inclusion in multivariable analysis (P < .2) for presence of tracheal blood were sex, age, time in training, season, work type and presence of significant haemosiderophages (Table S1). For presence of significant haemosiderophages the variables were age, time in training, season, neutrophil proportions, previous tracheal blood, current antibiotics, and previous antibiotics. (Table S2).
There was significant evidence for a linear association between time in training and odds of both tracheal blood (P = .004) and significant haemosiderophages (P < .001), and between age and odds of significant haemosiderophages (P < .001).
Likelihood ratio tests for departures from linearity were not significant, so age and time in training were analysed as continuous variables.

| Multivariable analysis
The explanatory variables that remained significantly associated with the outcomes and had a significant effect on the overall fit of the final models are presented with ORs, 95% CIs and Wald P-values in Table 1 for tracheal blood and Table 2

| D ISCUSS I ON
EIPH is known to vary by horse population and exercise performed, leading to difficulties in comparing prevalence between studies. In this study, the prevalence of EIPH diagnosed by blood on tracheobronchoscopy was lower than previous reports. EIPH has previously been shown to occur more frequently following racing compared with training. 8 This study population was examined following training only, and therefore the lower prevalence might be explained by differences in exercise intensity. 16 Cytological methods of EIPH diagnosis frequently result in higher prevalence estimates than observation of epistaxis or blood on tracheobronchoscopy. 1,6 In the current study, the 36% samplelevel prevalence of significant haemosiderophages in TW fluid was markedly higher than the 7% prevalence of tracheal blood. Similar findings have been reported in horses in flat training, with a 51% prevalence of increased haemosiderophages and 4% prevalence of tracheal blood. 6 Examination of TW samples for haemosiderophages potentially has the advantage of reducing the influence of shortterm exercise intensity on EIPH diagnosis as research examining autologous installation of blood into the airways indicates that haemosiderophages can persist in the airways for at least 28 days. 6,24 However, although the presence of haemosiderophages in TW fluid provides evidence of previous haemorrhage into the airways, the clinically significant proportions of these cells have not been determined. 1   This study population allowed further examination of the associa- The odds of tracheal blood and significant haemosiderophages were increased in both spring and winter, compared to autumn.
Others have suggested that this could be attributed to shorter term accumulated racing and training, as these months correspond to the end of the NH racing season. 6 Season is also a proxy measure for variations in ambient temperature and previous studies have demonstrated an increased risk of EIPH during colder weather. 5 The type of exercise performed prior to tracheobronchoscopy was significantly associated with tracheal blood. Compared to lowimpact activities, the odds of tracheal blood were increased for medium-and high-impact work. However, wide confidence intervals indicate that the OR estimates are imprecise, and these figures should be interpreted with caution. The observation of increasing odds of EIPH with increasing percussive impact of exercise undertaken corroborates previous studies where an increased risk of epistaxis was found following jump racing compared with flat racing, consistent with a theory of impact-induced trauma contributing to EIPH. 14 In the current study, due to low numbers of horses jumping prior to tracheobronchoscopy, it was not possible to distinguish jumping from other high-impact exercise. Anecdotally, EIPH has been associated with swimming, 13 a low-impact activity, suggesting a mechanism other than percussive impact. The association between swimming and EIPH could not be examined further in the current study due to low numbers of horses swimming and none having tracheal blood following this activity.
The presence of >20% neutrophils in TW samples indicates airway inflammation. 28 The odds of significant haemosiderophages increased with increasing proportions of neutrophils in this study, consistent with observations that blood instilled into the airways leads to inflammation. 18 However, we found no association between current inflammation or inflammation in the previous month and current tracheal blood. This suggests that inflammation associated with presence of significant haemosiderophages resulted from previous bleeding inducing pulmonary inflammation rather than current inflammation being a risk factor for current or future bleeding.
These findings are consistent with a systematic review concluding there was low-quality evidence that EIPH leads to inflammation and very low-quality evidence that inflammation causes EIPH. 11 Haemosiderophages are a recognised cytological indicator of prior pulmonary haemorrhage and their presence in TW samples has been used in previous studies to indicate EIPH. 6 The current study demonstrated fivefold increased odds of tracheal blood if horses had evidence of recent haemorrhage into the airways. As discussed by others, the high pulmonary pressures reached during intense exercise are sufficient to lead to capillary stress failure. 19 Over time this leads to remodelling of the small pulmonary veins, resulting in reduced lumen diameters and further increased pressures. This is consistent with the conclusions drawn by another study which demonstrated moderate to high-quality evidence that EIPH is progressive. 11 This can also be related to the theory that horses that experience EIPH tend to be elite athletes and are therefore likely to exert themselves to the point of experiencing EIPH. 15 Genetic variants associated with increased EIPH risk may have been maintained  32 Therefore, given a lack of standardisation for diagnosis of EIPH based on TW cytology, moderate and large proportions of haemosiderophages, representing a proportion of ≥25%, were subjectively categorised as "significant" for the purposes of this analysis in order to distinguish horses that had recent substantial haemorrhage from those that had recent smaller haemorrhages, or substantial haemorrhages occurring more than a month ago. As horses were examined monthly, bleeds experienced a month or more previously would have been evident on the previous examination. This threshold reduced the likelihood of recording the same episode of EIPH, based on small numbers of residual haemosiderophages, twice.
A further limitation of our study is that horses were examined following training, not on race days, and few after jumping. Specific information about duration or intensity of work performed prior to examination or other training-related risk factors were not available for analysis. Racing history, such as recent race dates, results or days between race starts, was also not accounted for. No variables related to medication remained in the models. This may have been influenced by inaccuracies in the data and missing data and should therefore be interpreted with caution.
In conclusion, this study demonstrates significantly increased odds of EIPH with increasing time in training. It provides further support for the association between bleeding in the previous month and current inflammation, 6 but not between current bleeding and previous inflammation, suggesting that inflammation observed may have resulted from previous haemorrhage, rather than being a risk factor for current or future haemorrhage. These findings support the capillary stress failure theory of EIPH, with increased time in training resulting in cumulative remodelling of the pulmonary vasculature, increasing susceptibility to EIPH through capillary stress failure with ongoing training. Cytological examination of TW samples was a more sensitive indicator of EIPH than tracheobronchoscopy alone and the development of standardised measures would contribute to improved detection of EIPH, particularly during training. Although EIPH in racehorses may not be avoidable, identification of horses at risk could contribute towards improved preventive measures in the future.

E TH I C A L A N I M A L R E S E A RCH
Ethics approval was obtained from the Royal Veterinary College Clinical Research Ethical Review Board (URN SR2018-1607).

ACK N OWLED G EM ENT
Richard Newton kindly reviewed the final draft of the manuscript.

CO N FLI C T O F I NTE R E S T S
None declared.

I N FO R M E D CO N S E NT
Owner consent was given for inclusion of the horses in the study.

PE E R 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.13448.

DATA ACCE SS I B I LIT Y S TATE M E NT
The data that support the findings of this study are available from the authors on reasonable request.