Measurements were performed in one of two barometric chambers (model PLY-351 or PLY-361, EMMS Havant, Hampshire, UK). Initially, dogs below 15 kg bodyweight were placed in the smaller chamber (PLY-351). As dogs, however, commonly touched the walls of the chamber, which interfered with the measurements, we changed the cut-off to 7–8 kg following the protocols used by the group of Liu in Cambridge.

The obtained recordings of respiratory patterns were analyzed using the eDacq software (EMMS Havant, Hampshire, UK). The set-up was calibrated with 300 mL of air according to the manufacturer’s instructions before each session. During measurements, chambers were ventilated with 20 L/min room air. If humidity increased over 60–65% or CO₂ over 0.4%, the doors were opened until humidity was below 50% and CO₂ was at 0.04%. Baseline humidity was commonly around 50–60%. To avoid further increases once the dog was in the chamber, the outflow of an air conditioning device was directed at the openings from which fresh air was pumped into the chamber (distant from the chamber itself). Before the onset of recordings, all dogs underwent a training phase in which they were positively reinforced for entering the chamber and staying in the chamber for gradually increasing amounts of time. Once the dog was familiarized with the chamber, the outer door of the chamber was closed, and breath recordings were started. The owners were actively involved in acclimatizing their dogs to the chamber and stayed next to the box during the WBP measurement (i.e., within sight for the dogs). If a dog appeared stressed, it was released from the chamber, given a 5-min break, and then re-introduced to the chamber. Measurements were discontinued if a dog showed persistent stress-associated behavior such as panting, yawning, or scratching. The goal was to record a sequence of 20 consecutive breaths in the same position. Standing, sitting, or laying positions were all accepted, as long as the head was not rested on the floor. The recording had to be restarted if the position was changed. Head movements were tolerated if (a) the head returned to the original position within a few breaths (e.g., a quick look to the side) or (b) they were small enough to not change the overall alignment of the head, neck, and body. Average respiratory rates between 10 and 40 breaths per minute were eligible for analysis.

Candidate breath sequences were selected live during the recording. Definitive sequence selection was performed once data collection was completed to ensure consistent implementation of defined selection criteria. To assess the quality of the sequence, movement was graded as “invisible,” “minimal,” or “moderate” and an interruption index, defined as the number of not recognized or not selected breaths divided by 20, was calculated. If more than one sequence was obtained, the highest-quality sequence was selected for analysis.

From the 20 selected breaths, EMMS Data ACQuisition software calculated means (av) and standard deviations (sd) for various respiratory parameters, including respiratory rate (f, measured in breaths/min), time for inspiration (Ti, measured in s), time for expiration (Te, measured in s), peak inspiratory flow rate (Pif, measured in ml/s), peak expiratory flow rate (Pef, measured in ml/s), tidal volume (TV, measured in ml), and minute volume (MV = TV × f).

The statistical model developed by Liu et al. (17, 26) predicts BOAS grades based on the means and standard deviations of Te/Ti ratio (Te/Ti), Pef/Pif ratio (Pef/Pif), minute volume per body weight (MVBW, measured in ml/kg), and f. We were kindly provided access to a web application of this model, and further analysis of WBP data focused on those parameters (Te/Ti_av, Te/Ti_sd, Te/Ti_av, Te/Ti_sd, MVBW_av, MVBW_sd, f_av, f_sd).

ETs followed the protocol performed by Riggs et al. (6) and were conducted in a corridor in the air-conditioned basement of the Small animal clinics with an ambient temperature of 20–21°C. A 5-min walking test was followed by a 3-min trotting test 30 min later. Owners were asked to walk or trot the dog on the lead at a speed of approximately 3–4 km/h in the walking test and 7–8 km/h in the trotting test. In cases of collapse, cyanosis, severe respiratory distress, or regurgitation, ETs were immediately stopped, and supportive care was initiated.

Immediately before and after both ETs, i.e., walking and trotting test, dogs were assessed for respiratory noises (including larynx auscultation), inspiratory effort, dyspnea, and cyanosis. BOAS grades 0–3 were assigned for each test separately according to Table 1. For further analysis, only the BOAS grades resulting from the respiratory functional grading before and after the trotting test were considered. Dogs with a history of collapse or cyanosis were not exercised and were directly assigned BOAS grade 3.

In addition to the parameters used for BOAS grading, rectal temperature was measured before and after each ET, and the temperature difference was calculated. Temperature measurement was omitted if a dog had shown signs of stress during previous temperature measurements (i.e., during the physical exam or before/after the walking test).

Assessment of anatomical conformation closely followed the procedures described in Liu et al. (17). Nostril conformation was assessed on a four-point scale, ranging from open to severely stenotic. Open nostrils were defined as nostrils without visible stenosis. Mildly stenotic nostrils were characterized by some degree of narrowing, which still allows for dorsolateral movement of the nostril wings upon exercise. In contrast, the nostril wings of moderately stenotic nostrils do not move after exercise, and the dorsal part of the nostril is (almost) closed (i.e., the lateral and medial walls are touching). In severely stenotic nostrils, both the dorsal and the ventral aspects of the nostril are closed, resulting in mouth breathing when walking or even at rest. An example for each grade is given in Figure 2A. For further analysis, moderately and severely stenotic nostrils were considered clinically significant stenosis (NS+), whereas open and mildly stenotic nostrils were summarized as clinically not significant stenosis (NS−).

Soft tape measurements of the head, neck, and body were taken as shown in Figure 2B to calculate the following ratios as defined in Liu et al. (17):

A total of 6 mL of venous blood was collected from the jugular vein, the vena cephalica or vena saphena. All methods used had been described previously (27). Clinical hematology included the complete blood count (CBC), reticulocyte count and maturity (low, medium, and high fluorescence), reticulocyte hemoglobin, and thrombocrit as well as total protein, albumin, bilirubin, glucose, urea, creatinine, cholesterol, triglycerides, alkaline phosphatase, alanine aminotransferase, creatine kinase, lipase, lactate dehydrogenase, phosphate, Na⁺, K⁺, Cl⁻ total Ca²⁺ and the ionized Ca²⁺. Furthermore C-reactive protein (CRP) was assessed as a clinical marker of inflammation.

Redox state markers (bulk thiols and non-protein reduced thiols) were assessed using monoromobimane (Mbbr) fluorescence and Ellman’s regagent, respectively, as described elsewhere (27). Osmoscane mode of Laser rotational red cell analyzer (Lorrca) was used to assess deformability, RBC hydration and membrane osmotic stability as described by van Cromvoirt et al. (34). Membrane surface area was estimated using eosim 5-maleimide (EMA) fluorescence (27). Avoximeter 4,000 was used to assess oxy-(O₂Hb), carboxy-(COHb) and met-hemoglobin (metHb). CO-oximetry was performed on the freshly drawn samples (within 15 min after blood collection). Precautions were taken to avoid reoxygenation.

Of the 102 recruited dogs, 84 completed the study and were included in the analysis (Figure 1). Reasons for exclusion were uncooperative behavior (11 dogs), not fulfilling inclusion criteria (5 dogs), and clotting of blood samples (2 dogs). Two-thirds of the included dogs were female (57 dogs), and roughly 40% (33 dogs) were gonadectomized. Their age ranged from 1 to 13 years (median 2 years 8 months). Further information on the dogs participating in the study may be found on in Supplementary Table 1 and on Zenodo platform (DOI:10.5281/zenodo.13358272).

The trotting test and the assessment of nostril width used in this study correspond to the Cambridge Respiratory Function Grading Scheme (RFGS). The implementation of this scheme for English Bulldogs, French Bulldogs and Pugs has meanwhile been recommended by the Fédération Cynologique Internationale (FCI) to all affiliated national breeding associations. The RFGS protocol focuses on the detection of respiratory distress and overall exercise tolerance; however, it does not explicitly assess for hyperthermia. In English Bulldogs the severity of BOAS increased the temperature change after submaximal exercise tests (21). Hence, we included rectal temperature measurements immediately before and after the exercise. Although a temperature increase of 0.5°C or more after performing the trotting test was only found in BOAS+ dogs, we did not detect significant differences in rectal temperature changes between BOAS+ and BOAS− French Bulldogs, which is similar to earlier findings in French Bulldogs (43, 44). However, in all three studies, the number of severely affected French Bulldogs was low, e.g., in our study only one of the five dogs graded BOAS 3 had temperature measurements before and after ET’s. Even if the determination of the temperature difference does not allow BOAS grading in the French Bulldog, the detection of a temperature increase of 0.5°C or more in the trotting test supports the diagnosis of BOAS + (45).

In addition to the ETs, we used the WBP as a recognized benchmark method (44) to determine the suitability of the selected blood parameters as future screening parameters. In our study, however, the agreement between BOAS grades predicted by the Cambridge web application and ET was poor (Figure 3). There are several possible explanations: first, the validation sample mainly consisted of show dogs (26), whereas our sample included primarily pet dogs not trained to sit calmly in a box. This could affect the quality of raw WBP data and thus the classification accuracy. Second, the selection of suitable breath sequences for analysis introduced a subjective component, so that highly standardized procedures could be required for reproducible findings. Third, we were not successful when using the web application developed by Cambridge at a different location (in Zurich), even when using an identical plethysmograph with the settings specified by Cambridge. Retraining the classification model using the list of parameters proposed by Liu et al. (26) but based on our WBP dataset resulted in an accuracy of 80%. While those parameters are physiologically plausible, we believe that higher classification accuracy could be achieved using different combinations of parameters for our specific study cohort and experimental setting. Due to the comparatively small sample size, we did not explore alternative parameter combinations. Even if the classification accuracy could have been improved, the question remains how much WBP could improve BOAS grading. In the meantime, trotting tests were shown to achieve a high sensitivity in severity grading of BOAS (44).

Craniofacial anatomic measurements are/were used for breeding restrictions to prevent the exaggeration of fronto-facial “baby-like” features such as relatively larger foreheads and eyes, which are related with BOAS and other health concerns.

However, in line with other studies [e.g., (17), our findings (Table 2)] do not support CFR as a robust predictor of higher BOAS. While three ratios derived from soft tape measurements were significantly associated with a higher probability of higher BOAS grade in our study, the identified ratios EWR, SI and NLR overlap only partially with those found by Liu et al. (17), NLR and NGR. Due to the considerable overlap of values between BOAS grades and the mediocre interrater reliability of soft tape measurements (17), soft tape measurements on their own lack robust predictive power sufficient for reliable BOAS grading. Stenotic nostrils on the other hand are clearly associated with a higher risk of BOAS, as shown by our data (Table 2) and in a previous study (17). Corrective surgery such as rhinoplasty may significantly improve the outcome of ETs, but not of the health status of the offspring of BOAS-prone parents. Hence, reporting of C-sections and surgeries is introduced by the British Kennel Club to exclude dogs with naturally severe BOAS from the breeding program (46). As stenotic nostrils are just one of several anatomical abnormalities contributing to BOAS, this approach may help to prevent the consequences of too narrow breeding selection. Particularly at the intermediate levels of the respective four-level grading systems, dogs may have mildly stenotic nares, but present as BOAS grade 2 in the trotting test, or vice versa may not be classified as clinically BOAS-affected despite moderately stenotic nostrils. Interestingly, neutered dogs in our study were less likely to have a higher BOAS grade (Table 2). This finding is not in line with the previous report on the lack of effect of neuter status in French Bulldogs and Pugs on BOAS severity, and a higher probability of neutered Bulldogs being affected by BOAS (17). The risk of anesthesia and the fear of neutering causing a worsening of clinical signs might explain, why dogs with more severe signs are less likely to be gonadectomized.

To date, studies linking anatomical conformation and BOAS grade in an ET are, to the best of our knowledge, cross-sectional. Longitudinal investigations might identify anatomical risk factors for clinical sign progression in dogs of BOAS grade 1 or 2 at the time point of screening. Prospective studies of Aromaa et al. elucidating the development of clinical signs of BOAS with time could also be helpful in better understanding the dynamics of the BOAS pathology (47). One more approach that could be undertaken to reduce the extreme anatomical manifestations resulting in BOAS is genetic screening for polymorphisms associated with severe phenotypes. Unfortunately, the studies suggest the involvement of multiple genes and deleterious alleles in the characteristic brachycephalic anatomy causing BOAS in different breeds making the selection of “healthy alleles” complicated (48). However, some of the polymorphisms, such as the c.2786G>A missense variant of the gene coding for the ADAM metallopeptidase with thrombospondin type 1 motif 3 (ADAMTS3) seem to be one of the promising candidates for a “genetic marker” of predisposition to BOAS development (49).

Dogs with higher BOAS grades are likely to experience more frequent episodes of hypoxia, associated by sleep apnea or increased physical activity (2). While chronic hypoxia increases red blood cell volume and hemoglobin concentration to improve tissue oxygen delivery, neither an increase of blood hemoglobin concentration nor erythrocytosis was associated with BOAS severity in our study population. This finding is similar to that in OSA patients, where only in 9 out of 527 patients had polycythemia (50). In our previous study, in which blood parameters of brachycephalic dogs were compared with those in non-brachycephalic dogs, we observed the signs of stress erythropoiesis such as an increase in reticulocytes counts and an increase in the abundance of immature reticulocytes (27). In the present study, the differences in these parameters did not reach significance when the dogs with BOAS grade 0–3 were compared with each other (Supplementary Figure S1). However, reticulocyte count and reticulocyte hemoglobin and redox state were predicted to be strongly associated with BOAS grading by varrank and caret models (Supplementary Figures S4A, B). Alterations in RBC turnover associated with the severity hypoxic condition can also be suggested from an association between the oxy-and carboxy-hemoglobin levels (Figure 5). Carboxyhemoglobin appeared to be the blood parameter showing the strongest association with BOAS severity and was clearly upregulated with an increase in severity grade from 0 to 2 (Figure 2B). Carboxyhemoglobin is accepted as a robust clinically accepted predictor of hemolytic activity (51). Hemolysis was reported to counterbalance the augmented erythropoiesis in the OSA patients (31). In human patients with OSA an increase in plasma bilirubin concentrations were detected in the morning compared to the pre-sleep levels (52) suggesting that apnea may be associated with hemolytic episodes.

One more feature of RBCs of brachycephalic dogs includes oxidative stress, manifested as GSH depletion (27). A modest reduction in the intraerythrocytic GSH levels could also be seen as BOAS severity progressed from 0 to 2 (Supplementary Figure S2E). Unfortunately, deoxygenation of hemoglobin to 50% or more may result in the release of extra GSH molecules stored within oxygenated hemoglobin resulting in an increase in free GSH in the cells, which is not associated with amelioration of oxidative stress (53). This behavior makes GSH a poor prognostic marker of intermittent hypoxia. One more interesting finding of the present study is trend to oxidation of protein thiols in reticulocytes of BOAS grade 2 (and probably grade 3) dogs (Supplementary Figure S2). Whether this trend is associated with the stress erythropoiesis condition or with the systemic oxidative stress as discussed earlier (27), remains to be explored.

We limited our research to one breed, the French Bulldog, to keep variances as small as possible. A statement is therefore only possible for this breed. It cannot be ruled out that French Bulldogs, which differ significantly from other dog breeds in the count of disorder predispositions (42) may also show differences in their blood parameters. However, since several of their “ultra-predispositions” were associated with their brachycephalic conformation, it might be assumed that the blood parameters associated with BOAS severity may be shared by other brachycephalic breeds, not only by French Bulldogs. The small number of animals in the “grade 3” group makes the data unbalanced and conclusions on the changes in blood parameters in brachycephalic dogs severely affected by BOAS questionable.

As follows from our study, none of the single-standing blood parameters qualified as a self-standing predictor of BOAS severity at the stages of 1–2 grades. Those requiring attention are CO-hemoglobin, reticulocyte counts, and their maturation, as well as RBC redox markers. They may be assessed as individual parameters or used to produce a compound variable like the principal component 1 in our PCA analysis (Supplementary Figures S4A, B). In our present study, we failed to reliably measure plasma nitric oxide derivatives NO₂⁻/NO₃⁻ as plasma samples often showed signs of hemolysis, which may interfere with the measurements. Thus, we cannot exclude that NO levels may add to the potential markers of BOAS severity, as higher levels of nitrite and nitrate in plasma was found in BOAS dogs earlier on (27, 29).

Much more research is needed to determine whether blood values are ultimately suitable as predictors of BOAS severity, which is a prerequisite to targeted selective breeding for healthier phenotype in brachycephalic breeds. According to our study results in French Bulldogs, promising candidates could be CO-hemoglobin, reticulocyte counts, and their maturation, as well as RBC redox markers. To avoid the animal welfare-related consequences of extreme brachycephalic breeding, we recommend for the time being, that all dogs of brachycephalic breeds undergo ET by trained veterinarians analogous to the Respiratory Function Grading Scheme of the Kennel Club and University of Cambridge’s, before both showing and breeding. Since only a small proportion of dogs are bred under Fédération Cynologique Internationale rules, it is crucial for the welfare of brachycephalic dogs bred in the future, that the screening test is required by law for all brachycephalic breeding dogs.