Relationship between tracheobronchoscopic score and bronchoalveolar lavage red blood cell numbers in the diagnosis of exercise‐induced pulmonary hemorrhage in horses

Abstract Background Exercise‐induced pulmonary hemorrhage (EIPH) is diagnosed and its severity assessed by post‐exercise tracheobronchoscopy, and enumeration of bronchoalveolar lavage fluid red blood cells (BALFRBC). Minimal information is available regarding the relationship of tracheobronchoscopy score to BALFRBC number. Objective Evaluate the relationship between BALFRBC number and tracheobronchoscopy scores and determine their diagnostic sensitivities. Animals Nine sedentary horses, 21 fit Thoroughbreds, 129 Barrel Racers. Methods Normal BALFRBC number and the effect of bronchoalveolar lavage (BAL) on it were evaluated by performing 2 BALs 24 hours apart in sedentary horses. Tracheobronchoscopy followed by BAL was performed 247 times on 150 horses after treadmill, racetrack, or barrel racing exercise. Lastly, a BALFRBC diagnostic threshold number that optimized the geometric mean of the sensitivity and precision (F1‐score) was determined using Bayesian analysis. Results No increase in BALFRBC occurred after the second BAL (mean ± SD, 304 ± 173/μL). Tracheobronchoscopy scores ranged from 0 (n = 112) to 4 (n = 4) and BALFRBC ranged from 102 to 4605268/μL. Spearman correlation between tracheobronchoscopy score and BALFRBC was weak (P < .001; rs = 0.42) with large ranges of BALFRBC associated with each tracheobronchoscopy score. The highest F1‐score occurred for a BALFRBC threshold number = 992/μL. Seventy‐five tracheobronchoscopy scores equaled 0 although BALFRBC number was ≥992/μL. Sensitivity of tracheobronchoscopy for diagnosing EIPH was poor (0.59; 95% confidence intervals [CI], 0.49‐0.68), compared to BALFRBC number ≥992/μL (0.93; 95% CI, 0.88‐0.96). Conclusions and Clinical Importance False negatives are common with tracheobronchoscopy. Follow‐up determination of BALFRBC may be indicated for tracheobronchoscopy scores = 0 before EIPH can be ruled out.

Conclusions and Clinical Importance: False negatives are common with tracheobronchoscopy. Follow-up determination of BALFRBC may be indicated for tracheobronchoscopy scores = 0 before EIPH can be ruled out. Although TBE assigns 1 of 5 possible scores (0-4) as a reflection of EIPH severity, evaluation by BAL is semi-quantitative and has been used to assess the severity of EIPH in horses, cyclists, and greyhounds. [2][3][4] Distal segments of a lung are lavaged after exercise, as much lavage fluid as possible recovered, and the number of RBC in that fluid (BALFRBC) determined. The diagnostic validity of BAL with respect to EIPH has been questioned because of the possibility that the procedure itself might induce hemorrhage from the airway mucosa and iatrogenically increase the post-exercise BALFRBC number. 5 However, we could find little evidence that this concern has been investigated in horses. 6 Bronchoalveolar lavage is not commonly performed after TBE in the short-term post-exercise period for the diagnosis of EIPH, and we could find no criteria regarding the BALFRBC number that is considered positive for the condition. Some studies have utilized posttreadmill exercise BALFRBC number to assess the effects of putative prophylactic treatments for EIPH, but none of them designated horses as positive or negative for EIPH on the basis of a given number. [7][8][9][10] Tracheobronchoscopic examination is easier to perform than BAL and consequently is the diagnostic method of choice for EIPH.
Criteria for assigning a particular TBE EIPH score severity have been described in detail. 11 The TBE EIPH scores are assigned according to observations made 30 minutes to 2 hours after exercise, with a score of 0 signifying no visible evidence of EIPH. Although the specificity of TBE is apparently very high (low rate of false positives), it is also assumed to have a high sensitivity. However, this assumption has not been justified and the sensitivity of TBE may well be lower than pre-  The presence and severity of EIPH was graded according to a scale defined by the quantity of blood observed in the airways. 11

| Bronchoalveolar lavage
Bronchoalveolar lavage was performed immediately after TBE, except for the initial study using sedentary horses, which did not entail TBE.

| Statistical analysis
The correlation between EIPH TBE score and post-exercise BALFRBC number was calculated using a Spearman rank correlation with data collected after barrel racing, treadmill, and racetrack exercise events. The effect of prior BALF evaluation on BALFRBC number was assessed using a 1-way repeated-measures analysis of variance. In all instances, significance was set at P < .05.
Sensitivity and specificity of TBE in exercised horses along with the population characteristics of BALFRBC numbers in both exercised and non-exercised Thoroughbreds were estimated using a Bayesian analysis similar to that described previously. 15  used to obtain an ROC curve for BALFRBC numbers. From the ROC curve, the optimal F1-score threshold was determined for the BALFRBC number.
A Bayesian analysis was conducted to estimate log BALFRBC mean and variance for true positive and true negative horses as well as posterior distributions on sensitivity and specificity for TBE.
The sensitivity (true positive rate) and specificity (true negative rate) were reported conditional on the optimal F1-score BALFRBC threshold. All statistical analyses were performed using R (R Core

| Relationship between TBE EIPH score and BALFRBC number
The Spearman rank correlation between TBE score and BALFRBC number was significant but weak (P < .001; r s = 0.42; Figures 1 and 2).
Numbers of BALFRBC between 4167 and 41 105/μL were associated with TBE EIPH scores from 0 to 4 (Table 1, Figure 1). The color of the BALF also varied considerably, ranging from a colorless to dark red when TBE score = 0, and from pink to dark red for TBE scores of 1-4.

| Sensitivity and specificity of TBE score and BALFRBC number
Based on TBE, the prevalence of EIPH in the barrel racers was 52%, whereas that in the Thoroughbreds was 59%. The optimal F1-score However, the correlation between these 2 diagnostic methods was weak. The SD of the BALFRBC number for each TBE EIPH score was higher than the corresponding mean, and the range of RBC numbers associated with each score was very large (Figure 1). We clearly demon- It has been unclear as to whether the BAL procedure itself might be associated with airway hemorrhage. One previous report documented effects of repeated weekly BAL procedures for 10 weeks in 4 non-exercised horses and found that there was no change in BAL-FRBC number over the course of the study. 23 However, because BAL-FRBC numbers decreased rapidly in the first 3 days after instillation of 40 mL of autologous blood into lung segments, 24 we felt it was important to further assess whether the BAL procedure itself could induce hemorrhage and inadvertently increase post-exercise BALFRBC by evaluating the effects on this variable of performing 2 BALs in a period of 24 hours. The results indicated that performing the first BAL had no effect on the RBC number recovered in the subsequent procedure. Furthermore, the mean ± SD for each BAL was very similar to that previously reported in non-exercised horses lavaged weekly for 10 weeks (309 ± 73/μL). 23 These numbers were lower than those previously reported for non-exercised horses (844 ± 123/μL). 6 However, those horses had been exercised regularly until 4-6 weeks before 2 BALs were performed 1 week apart. Despite the fact that BALFRBC numbers decrease markedly soon after instillation of known volumes of blood, erythrocytes can be detected 2 and 3 weeks, respectively, after a single bout of strenuous exercise, [24][25][26] which is why we selected horses that had not been exercised for at least 3 months before this portion of the investigation. The highest BALFRBC number recorded from the 18 BALs performed on the sedentary horses was 617/μL, and the subsequently calculated EIPH positive RBC threshold number (992/μL) was >3 SDs higher than the mean BAL-FRBC numbers in these horses. Consequently, it seems very unlikely that the BAL procedure per se induces airway hemorrhage provided that an appropriate technique is utilized. It is unclear why a small number of RBCs were present in the BALF of sedentary horses, but this finding is not without precedent. 6 It is unlikely that such findings reflect previous bouts of EIPH because of the lengthy period between any previous exercise and BAL and the lack of evidence of erythrophagocytosis. Some diapedesis of blood into the smallest airways may be normal, because some leucocytes also are normally found in the airways of healthy horses. 6,27 The sensitivity of a diagnostic test is a measure of its accuracy in detecting cases that are truly positive for the condition in question, whereas the test's specificity refers to the likelihood that it will correctly identify cases that do not have the condition (true negatives).  A single unexpected instance occurred in which the TBE EIPH score was >0 but the BALFRBC number was <992/μL. This instance was regarded as a false-negative BALFRBC result. The threshold RBC was determined from the Bayesian statistical analysis rather than direct scientific evaluation and was therefore the best possible estimate based on the available data, so the occurrence of a falsenegative BAL result was not surprising. The finding also could have been an indication that all or most of the hemorrhage into the airways had cleared the smaller airways by the time the BAL was performed or that the region of lung that was lavaged was not representative of the area from which the EIPH emanated in that particular horse.
In conclusion, the diagnosis of EIPH cannot always be based on TBE. The relationship between BAL and TBE is weak to moderate at best, and it is important that veterinarians, horse owners, and trainers be aware of this poor correlation. Practitioners might be advised to perform BAL on horses that have not performed well but have a TBE