The present study consisted in the description of the phenotypes associated with EA in comparison to healthy controls. While the phenotypical variability increased with disease severity, the horses were grouped as mildly, moderately and severely affected (miEA, modEA, sEA). Clinical scores and BALF cytology were confirmed as accurate predictors of the EA phenotype by use of recursive partitioning trees, which did not identify sufficiently informative alternative variables.
Diagnostic value of specific examinations
In practice, clinical examination and BALF cytology are essential in the diagnosis of EA [1]. However, the variability of EA phenotypes can justify a broad range of examinations for full disease characterisation, permitting more accurate prognoses, and supporting therapeutic decisions. Moreover, the dynamic nature of EA and lack of temporal consistency in patients exposed to changing environments emphasizes the importance of owner observations and results in additional variability [22–24]. Finally, clinicians have long tried to identify parameters allowing simpler and faster identification of EA in patients.
For example, arterial blood gas analysis, as an objective tool to assess lung function and efficiency of respiratory gas exchange, has been suggested as an aid in staging EA severity [25–27]. A fast processing of the fresh sample is mandatory and usually more difficult in field settings [28, 29]. A few previous studies have reported decreased PaO2 levels at rest (< 94 ± 3 mmHg [31]) in both symptomatic and non-symptomatic asthmatic horses, with a more prominent decrease during sEA exacerbation [31–35]. While no horse was presented in exacerbation in the present study, PaO2 was in fact lower in the sEA group but stable in less severely affected horses. In racehorses with mild to moderate EA (Consensus Statement abbreviation: mEA), decreased PaO2 levels were observed after performing standardised treadmill tests [36]. However, there is limited information on blood gas alterations available in mixed populations of asthmatic horses. In addition to PaO2 and PaCO2, the calculated alveolar-arterial oxygen gradient (AaDO2) describes oxygen diffusion across lung membranes and is increased in cases of diffusion impairment, e.g., with thickened lung membranes [37, 38]. In human asthmatics, it is increased in severely diseased individuals [39]. This was also observed in asthmatic horses [35]. In our study, the diagnostic value of AaDO2 was comparable to that of PaO2. Therefore, these parameters were most useful to distinguish horses with modEA and sEA. On the other hand, the decision tree based on arterial blood gas parameters was overly complex and performed poorly to separate all horses into their respective phenotypes. Moreover, other models where this data was available did not select these parameters. In conclusion, physiological values of PaO2, PaCO2 or AaDO2 cannot rule out any EA phenotype, since physiological values were also present in one horse classified as sEA. Still, a decreased PaO2 or an elevated PaCO2 at rest were found in most cases of sEA and may be used to monitor treatment response [25].
BALF collection is considered the gold standard to diagnose EA and is often performed during airway endoscopy [1, 22, 25]. In this study, endoscopic scores were increasing with disease severity but overlapped between the phenotypes. These findings are consistent with previous studies that have shown little value of single visual abnormalities at endoscopy for diagnostics alone [6, 18, 40, 41]. Two studies found good discriminant potential for the amount of tracheobronchial mucus between control and RAO-affected horses [42, 43]. In our study, the amount of tracheobronchial mucus (secretion score) had the strongest positive correlation with diagnosis and BALF neutrophils. However, these findings did not distinguish miEA and modEA from healthy horses particularly well.
Despite its prominent role in the diagnosis of EA, the clinical relevance of slightly elevated inflammatory cells in BALF in otherwise inconspicuous horses is insufficiently supported by literature. This has led to discussions about feasible cut-off values for equine asthma [1] and resulted in various cut-offs being used in different studies [4, 22, 44, 45]. In this study, the cut-offs of 10% for neutrophils, 5% for mast cells and 2% for eosinophils [1] were chosen to avoid over-diagnosing EA in horses examined at different seasons of the year, as higher neutrophil counts were reported during the winter season [4]. Surprisingly, four out of 12 clinically healthy horses were diagnosed with miEA based on elevated mast cell counts in BALF. Although the small sample size precludes generalizing the present results to the whole population, there are likely many cases of undiagnosed, subclinical EA, as it has been observed in other horse populations [4, 46, 47]. Positive correlations of BALF cytology results with disease severity were expected, since diagnoses were partly based on BALF cytology. However, it was also found that the recovered BALF volume was negatively correlated with diagnosis. This phenomenon may be attributed to airway remodelling with increased alveolar spaces, resulting in loss of lavage fluid, as previously described in another study when comparing sEA horses to healthy controls [48].
Regarding medical history and clinical examination, the cough score was identified as potent discriminant parameter in the classification tree. Some authors suggest that a diagnosis of sEA can be made based solely on medical history [8]. However, despite the widespread understanding of its impact, it has been shown that owners have difficulty in recognition and assessment of EA [49]. Previous studies have evaluated the reliability of owner observations in describing the severity of RAO. These studies have found reliable agreement of owner reports with severe symptoms of RAO [44, 50, 51]. The present study confirms this finding, as cough was the only owner-reported symptom that reliably distinguished between healthy and diseased horses. Other studies also reported a good association of owner-reported cough and clinical signs of RAO in affected horses [52] and found an association between owner-reported cough and elevated airway neutrophils but not mast cells, which are mostly found in mild-moderate EA (mEA) [48]. The cough score used in this study included both cough at examination and history of coughing. Since this score had the highest correlation with disease severity, this further supports the importance of reported cough in diagnosing EA. However, it is important to note that reported cough as indicator is only applicable to modEA and sEA and is of no benefit in diagnosing subclinical EA (mild = miEA), if used without BALF analysis.
In conclusion, the combination of BALF cytology and the study’s cough score were the most valuable parameters for the discrimination of all phenotypes through the classification algorithm and superior to all other data subsets. Classification trees without BALF data were unable to achieve clinically realistic complexities with sufficient performance accuracies, regardless of the data subsets used. This highlights the significant importance of BALF examination and history in diagnosing EA.
Phenotypical variability in a clinical population
A major, but expected, finding of the present study is the association between EA severity and phenotype variability. Healthy and miEA horses share the similar inconspicuous clinical presentation and can only be distinguished by BALF cytology. In each horse diagnosed with modEA and sEA at least one pathological finding was noticeable during the general examination. However, the frequency and severity of symptoms did not increase linearly with diagnosis in all individuals per phenotype, as indicated by overlapping scores. The clinical scores correlated well with the underlying parameters of the clinical examination but were also superiorly correlated to diagnosis, supporting the use of defined scores as objective variables. Elevated respiratory effort at rest was widely distributed and not exclusive to the sEA phenotype, as suggested by the ACVIM consensus definition [1]. However, the absence of this finding did not preclude severe disease (e.g., ID 10, 7). The wide variety of clinical signs in mEA, including elevated breathing effort, was just previously described in other asthmatic horse populations [4, 44]. In this study, the frequency and severity of additional symptoms also varied among symptomatic horses of both phenotypes. Spontaneous cough at hospitalisation was present in some horses with modEA and sEA; and lung auscultation was also found to be physiological in horses classified as sEA. The insensitivity of lung auscultation in mild and modEA cases has been recently objectively confirmed by using a digital auscultation device [53] and other authors suggested that lung auscultation is not beneficial for EA staging [42]. Additionally, nasal discharge was rather consistent among symptomatic horses (11/14), although it is not emphasized in the definition of EA [1, 54]. Another study reported this symptom in approximately 40% of horses with mEA and sEA [44]. We found positive correlations between the occurrence of symptoms and disease probability for all general examination findings and clinical scores. However, the discriminatory potential of each parameter was poor, and the absence of individual symptoms did not necessarily indicate the absence of severe disease. The overlapping clinical presentations of the phenotypes and the increasing clinical variability of all examinations were confirmed by factor analysis of mixed data, where increasingly large clusters were observed for each group.
Management and housing before presentation may explain the variability of the phenotypes. Antigen avoidance was shown to have a high impact on the clinical presentation, possibly also influencing measurable airway neutrophilia and remodelling [22, 23, 55–61]. In some horses, clinical signs improve before resolution of BALF neutrophilia [62]. Together, these findings indicate that repeated examinations may be superior to correctly classify a horse, as short-term changes may affect clinical parameters earlier than BALF cytology. The classification algorithm suggests cut-off values of 4.75% for mast cells to differentiate healthy from miEA and 28% for neutrophils to differentiate modEA from sEA. These cut-offs are specific for the presented study population and should not be generalized to a population level. Nevertheless, they emphasize the importance of mast cell involvement in early EA development and the shift to airway neutrophilia with disease severity.
Supporting the existence of a mild EA phenotype
The ACVIM consensus statement distinguishes between two phenotypes (mEA and sEA). The mEA definition groups subclinically and mildly diseased horses as a single phenotype, while also acknowledging that EA is a spectrum [1]. However, when confronted with a clinical population, as it is presented in this study, the presence of clinically inconspicuous horses with abnormal BALF cytology is a challenge. We defined miEA as clinically indistinguishable from healthy horses with mild inflammatory cell influx to the lung. In this study, four out of 12 horses presented as healthy sport horses showed signs of airway inflammation in BALF cytology. Since BALF cytology is not performed on a regular basis by referral veterinarians in subclinical horses, this phenotype might be underdiagnosed in the whole horse population [8].
The definition of the subclinical phenotype as mild EA was suggested for racehorses with abnormal BALF findings and exercise intolerance [5, 10, 34]. However, populations of subclinical non-racehorses with abnormal BALF cytology findings were not just described in our study, which suggests the use of this phenotype description also for other horse populations than equine athletes [4, 6]. Diagnosing subclinical EA may aid to reduce the prevalence of symptomatic horses by emphasizing the need for antigen control to the owners also for clinically healthy horses. This is difficult because low-level exercise intolerance is much harder to define and detect in leisure horses, due to the variety of disciplines, expected and expectable performance [63–65]. In addition, standardised treadmill tests are not available under field conditions [66–68]. Finally, portable devices for pulmonary function testing and blood biomarkers are missing, and bronchoprovocation tests do not seem reliable in miEA [69–71]. Therefore, BALF cytology remains the only applicable diagnostic method for subclinical horses that can also be performed under field conditions. Still, the association between slightly elevated mast cells or neutrophils in BALF and clinical symptoms, except for exercise intolerance in racehorses, or progression of the disease, remains unclear [8, 22] and further diagnostic tools are desired.
Main limitations
The standardised examination protocol was advertised through an online campaign to horse owners. Therefore, the study population is likely to be biased towards affected horses and pre-informed owners. However, the aim of the study was to describe the breadth of EA phenotypes rather than prevalence of each phenotype. Moreover, the predefined protocol ensured objective examination results and consistent testing conditions.
Another issue affecting EA diagnostics in general is the difficulty of defining and assessing exercise intolerance, which was mentioned in the medical history by some owners, but could not be identified as an objective, repeatedly detectable and quantifiable symptom. As a result, low-level exercise intolerance cannot be excluded with the underlying data.
Furthermore, the use of a single examiner may be seen as a limitation, but this approach was taken to avoid examiner variance. Finally, the study population is limited in size. Therefore, the decision trees are expected to be less performant when used on a broader population. Since the main tree involves only three variables and is appealing by its simplicity, a validation study on a separate cohort is warranted.