Spirometry is a common monitoring modality used in clinical practice in anesthetized dogs (1). The pressure-volume relationship of the respiratory system (P-V) is fundamental for assessing its mechanics during mechanical ventilation (2). Compliance is defined as the change in lung volume per unit change in pressure gradient; it may be measured for lung, thoracic cage, or both [i.e., respiratory system; (3)]. Compliance plays a pivotal role in understanding respiratory physiology in veterinary anesthesia, particularly when dogs are mechanically ventilated. It is commonly displayed by modern monitors, and generally calculated dividing the expired gas volume by the change in the airway pressure (plateau pressure—positive end expiratory pressure). Various types of compliance, including dynamic compliance (C_dyn), static compliance (C_stat), and quasi-static compliance, have been utilized to evaluate respiratory system function in response to interventions and different influencing factors (4–7). Both C_stat and quasi-static compliance require an inspiratory pause. Quasi-static compliance, obtained with inspiratory pauses of variable durations below 3 s, did not provide accurate C_stat values in healthy dogs (8). C_dyn offers a real-time assessment of respiratory performance during mechanical ventilation, enabling the early detection of changes in airway resistance—such as bronchospasm, secretions, and tube kinking—as well as alterations in the elastic component (9).

Despite decades of research on P-V curves, our comprehensive understanding of the P-V relationship remains incomplete (2). C_dyn can be assessed as a loop or as a numerical value. A flatter loop suggests lower compliance, potentially indicating stiffening of lung tissue or the chest wall, whereas a steeper loop indicates higher compliance (10). Veterinary anesthetists typically use the shape and steepness of the P-V loop independently of numerical values for C_dyn to guide their clinical decisions (1). The shape and steepness of P-V loops are greatly influenced by the scale of the P-V graph displayed. The axes scales can be altered either manually or automatically by devices. Therefore, interpreting C_dyn based on the steepness of a P-V curve through subjective visual observation may lead to incorrect conclusions, particularly given the existing gap in the literature concerning C_dyn usual values. Creating a quantitative reference range for C_dyn may assist clinical interpretation of changes in lung mechanics.

However, developing a reference range is likely to be challenging due to the different variables influencing compliance that have been identified in dogs. Body mass is a logical variable due to its impact on tidal volume (V_T), as described by Stahl (11), and indicated in equations reported by Asorey et al. (12) and Bradbrook et al. (13). Additionally, the influence of body position on C_dyn (13) and inspiratory time (14) have been demonstrated. These variables underscore the multifaceted nature of compliance measurements and highlight the need for establishing reference intervals specific to the canine population.

The goals of this prospective, exploratory study were to establish a reference interval for C_dyn in anesthetized dogs and to identify the variables impacting C_dyn.

A total of 515 cases were recruited. Data from 12 cases in which C_dyn varied by more than 15% across the three measurements were excluded from the analysis. Cases with missing or obviously incorrect data in selected variables were excluded when analyzing variables independently, and from the total dataset only after their inclusion in the final model was confirmed. This approach resulted in varying case numbers across analyses, as detailed in relevant sections of the Supplementary data. A total of 462 cases were included in the final model.

A variety of breeds was represented. Table 1 lists the breeds and the number of dogs of each breed included in the study. Due to insufficient numbers of each breed, the effect of breed on C_dyn was not investigated.

Based on insufficient data, the following categories were not included in the analysis: Trendelenburg (only four animals were positioned in Trendelenburg and eight in inverted Trendelenburg); breathing system type (circle breathing systems were used in most dogs; the use of a Mapleson A and D were reported in only one and seven cases, respectively); the administration of etomidate, desflurane and intravenous lidocaine (reported in only eight, one and 19 cases, respectively).

Descriptive statistics and the frequency histogram of C_dyn are reported in Table 2 and as Supplementary material (Section 1 and Supplementary Figure 1). Normal distribution was rejected.

Body mass was selected as the first predictor variable for C_dyn based on prior research indicating a significant relationship between these variables. The line of best fit of C_dyn against body mass crossing the origin and the model from Asorey et al. (12) for comparison are presented in Figure 2A. The line of best fit of C_dyn against body mass with free intercept and the model published by Bradbrook et al. (13) for comparison are presented in Figure 2B. Both linear models and their comparison are reported as Supplementary material (Section 2 and Supplementary Tables 1A, B). The model with intercept was kept for further development according to better performance metrics (Supplementary Table 1C).

The internal diameter of the orotracheal tubes was the second factor in the order of anticipated importance determined by the authors. As the correlation coefficient between body mass and internal diameter of the orotracheal tubes was 0.4, further investigation was warranted before the relationship between internal diameter of the orotracheal tubes and C_dyn was analyzed. Detailed results for the relationship between internal diameter of the orotracheal tubes and body mass are presented as Supplementary material (Section 3, Supplementary Table 2 and Supplementary Figure 2). The best fitting prediction of orotracheal tube against body mass was a sigmoid Emax model (Supplementary Section 4 and Supplementary Figure 3). Based on best models' performance metrics, internal diameter was categorized as “Small” if more than 7% smaller than its predicted value, or “Medium/large” if above. The Figure 3 represents the C_dyn against the body mass with the linear regressions. The effect of the internal diameter of the orotracheal tube on the C_dyn ~ body mass model is reported in Table 3.

The correlation coefficients between the length of the orotracheal tubes and their diameters and the length of the orotracheal tubes and body mass were 0.46 and 0.74, respectively. As the internal diameter of the orotracheal tubes impacted the C_dyn ~ body mass model, the influence of the length of the orotracheal tube was not investigated. No other multicollinearity was identified (i.e., the absolute correlation coefficient between all the other variables were <0.4).

Body condition score was also identified as factor influencing C_dyn. According to model comparisons, the BCS were best divided into the following three categories: “low BCS” (1/9, 2/9 and 3/9), “medium BCS” (4/9 and 5/9), and “high BCS” (6/9, 7/9, 8/9 and 9/9). The frequency histogram of BCS, boxplots of C_dyn for each BCS are presented in Supplementary material (Section 6, Supplementary Figures 5A–C). Linear regressions of C_dyn against body mass for the subgroup of “low BCS,” “medium BCS,” and “high BCS” were calculated (Table 3) and graphically represented (Figure 4).

The following variables had a significant impact on C_dyn and improved the C_dyn ~ body mass model when considered individually (i.e., reduction in both AIC and BIC, and p < 0.05): morphology, lateral recumbency, spirometry sensor used, the administration of FIO₂ >80% for at least 10 minutes before the measurements, time elapsed between anesthesia induction and the measurements, inspiratory time set on the ventilator, and ventilation mode. Their effect on the C_dyn ~ body mass model and associated figures is reported in Table 3. Additional information on the individual variables can be found in the Supplementary material (Section 7 and Supplementary Figures 6A, B; Section 8 and Supplementary Figures 7A, B; Section 9 and Supplementary Figures 8A, B; Section 12 and Supplementary Figures 11A, B; Section 13 and Supplementary Figure 12; Section 14 and Supplementary Figure 13; Section 15 and Supplementary Figures 14A, B).

The development of the multiple linear regression model involved using stepwise forward selection and backward elimination techniques. The procedures and criteria for model selection are detailed in Supplementary material (Section 17 and associated Supplementary Tables). Variables were added or removed iteratively, with the goal of achieving a model that balanced fit and simplicity. The final model is presented in Table 4. The variables included were selected based on their contribution to the model's performance, as reflected in the evaluation criteria described. The best-fitting model identified a linear relationship between C_dyn and body mass when the following conditions were met: a “high BCS,” “small” orotracheal tubes (<7% smaller than predicted), the use of a D-lite flow sensor, and the absence of a high FIO₂ (>80%) exposure for more than 10 minutes prior to C_dyn measurement. In cases where these conditions were not met, additional factors needed to be incorporated into the model. Factors such as a “Low (1/9, 2/9, 3/9) and Medium (4/9, 5/9) BCS,” an orotracheal tube with an internal diameter not <7% smaller than the predicted value (i.e., an orotracheal tube of the predicted size or larger) and longer inspiratory times were associated with increased C_dyn. Conversely, the use of alternative spirometry sensors, including Ped-lite, or prolonged exposure to high FIO₂ levels resulted in decreased C_dyn.

Linearity was confirmed (Residuals Plot, Figure 12). Homoscedasticity was not confirmed (p < 0.001). Independence was confirmed (Durbin-Watson test, p = 0.486). Normality of Residuals (QQ Plot, Figure 13) was not confirmed (Shapiro-Wilk test, p < 0.001). Multicollinearity was confirmed (Variance inflation factors < 1.3). To account for heteroscedasticity, robust standard errors were calculated using the more conservative Davidson and Mackinnon heteroscedasticity-consistent estimators (via the Sandwich package in R). The results, including recalculated p-values, are presented in Table 4. Additionally, to address the non-normality of the residuals, estimates were re-evaluated through bootstrapping (n = 10,000; Boot package in R) to obtain bias-corrected estimates and 95% confidence intervals (Table 4).

Age, the presence or absence of HMEs, the F_IO₂ and the PEEP levels did not have any significant effect when integrated to the multiple regression linear model. Additional information on the individual variables can be found in the Supplementary material (Section 5 and Supplementary Figures 4A, B; Section 10 and Supplemental Figures 9A–C; Section 11 and Supplemental Figures 10A, B; Section 16 and Supplemental Figures 15A, B). The administration of alfaxalone, alpha-2 adrenergic receptor agonists, anti-muscarinic, isoflurane, propofol, and sevoflurane had a significant effect on C_dyn (Yes/No for each drug; Wilcoxon rank-sum test p-value = 0.012, 0.012, 0.002, 0.007, 0.013, and 0.036, respectively). However, their individual effect was minimal to negligible on the C_dyn ~ body mass model and they were not included in the final model. The ventilation mode (Supplementary Section 15 and Supplementary Figures 14A, B), the administration of acepromazine, benzodiazepines, butorphanol, ketamine, mu-agonist opioids and neuromuscular blocking agents and the use of loco-regional blocks or epidural anesthesia did not significantly influence C_dyn (Yes/No for each drug or technique; Wilcoxon rank-sum test p-value = 0.711, 0.841, 0.222, 0.368, 0.811, 0.442, 0.235, 0.861, and 0.735, respectively).

The main aim of this study was to establish a reference interval for C_dyn in anesthetized dogs. The large variance detected in C_dyn indicated that it was too inaccurate to develop a reference range and that modeling was required to estimate C_dyn. Our findings demonstrated that C_dyn could be quantified relative to body mass. However, it was also notably affected by the internal diameter of orotracheal tubes, BCS, inspiratory time settings, the type of sensor used for C_dyn determination, and exposure to FIO₂ >80% for at least 10 minutes prior to measurement. These findings illustrate that a single reference interval would be misleading or unhelpful in clinical practice due to the varying combinations of factors that can alter the measurement in different ways. Despite this, our study provides valuable insights into the factors affecting C_dyn in anesthetized dogs and emphasizes the need for careful consideration of these variables in both clinical and research settings.

Our results demonstrated a linear relationship between C_dyn and body mass in dogs, corroborating previously documented findings (11–13). This relationship was incorporated into all subsequent modeling. However, our study offered significant insights not captured in previous investigations and highlighted that body mass alone was insufficient to reliably predict C_dyn. Our unique approach, derived from consultations with the veterinary anesthesia community (1), enabled us to further investigate prospectively the influence of numerous factors.

In our study, the impact of the inspiratory time on C_dyn was evaluated. We observed that varying inspiratory times significantly influenced the readings obtained. The characteristics of the inspiration have been shown to influence compliance values. Both C_stat and quasi-static compliance require an inspiratory pause. Varying inspiratory pauses of <3 s gave different values of quasi-static compliance and did not provide accurate C_stat values in healthy dogs (8). Here, the significant impact of the inspiratory time on C_dyn was demonstrated, confirming earlier findings (14). The first publication about D-lite flow sensors states that “some systematic error in compliance measurement is introduced if a ventilation mode without a pause and well-defined plateau pressure is used. However, even then, trends will be quite informative.” (19). In clinical anesthesia and in many studies, the inspiratory-to-expiratory ratio is set, and the respiratory rate is adjusted to maintain a target partial pressure of expired carbon dioxide. Clinicians wishing to monitor changes in C_dyn or quasi-static compliance over time should consider the importance of inspiratory time and keep it consistent in their assessments. The characteristics of the inspiratory settings (inspiratory time or the inspiratory-to-expiratory ratio and the respiratory rate set, and inspiratory pause) should also be systematically reported in research.

Our results also showed an influence of the internal diameter of orotracheal tubes on C_dyn, as documented previously (13). However, Bradbrook et al. stated that “since ETT (Endotracheal Tube) diameter was probably dependent upon body size and therefore body weight, only body weight was investigated further by general additive modeling” (13). By including more dogs in our study, we were able to further investigate and model this relationship. This allowed us to identify a diameter cut-off value of 7% below the predicted value, below which C_dyn is negatively impacted. It is worth noting that monitors using multiple linear regression of the whole respiratory cycle to calculate or estimate C_dyn might be less affected by airway resistance (20). Further investigation is necessary to validate these findings across a wider range of monitoring technologies and techniques.

Interestingly, in our study an FIO₂ above 80% for at least 10 minutes resulted in lower calculated values of C_dyn. Previous research showed that administering a low FIO₂ (40%) to ventilated dogs was associated with less atelectasis formation compared to administering a high FIO₂ (21). Mathematical modeling of absorption atelectasis kinetics indicated that the process developed fairly rapidly, based on human data (22). The 10-minutes cut-off was chosen because a complete collapse would mathematically develop in <10 minutes if a 3-minutes pre-oxygenation was followed by 100% oxygen administration (22). The time spent at a high FIO₂ was previously shown not to influence compliance (12). However, in that study, the dogs received a high FIO₂ for 34 ± 18 minutes, with the duration of high FIO₂ administration being the focus of the analysis. Our findings might reflect a clinical manifestation of absorption atelectasis, suggesting its rapid development in dogs in vivo, as supported by the mathematical model.

In our model, the ventilation mode did not influence C_dyn. However, previous research described higher compliance values when pressure-controlled ventilation was used compared to volume-controlled ventilation (23). Lower peak airway pressures tend to develop with pressure-controlled ventilation compared to volume-controlled ventilation (24), presumably because decelerating flow patterns enhance ventilation distribution in lungs with varying mechanical properties (25). In their study, Fantoni et al. used C_stat with a 4-s inspiratory hold, reflecting improved lung compliance without the resistive component of the airway and orotracheal tube. They targeted similar tidal volumes (10 mL/kg) and PEEP (5 cmH₂O) with both modes and measured airway pressure. This contrasts with how veterinary anesthetists typically set pressure-controlled ventilation, where pressure is most commonly targeted (1). The reasons for the observed differences are unclear, although variations in the settings of ventilation modes could be a contributing factor. Future studies employing more standardized ventilation strategies could further clarify this aspect. It is worth noting that the influence of inspiratory time on C_dyn has the potential to vary significantly between volume-controlled and pressure-controlled ventilation, which may contribute to explaining the discrepancies observed in our results.

Our study did not investigate thoracic shape and BCS in the same manner as Asorey et al. (12). The 20% inspiratory pause set in their study precludes direct comparison with our findings. However, they highlighted a reduction in compliance in overweight and “barrel-chested” dogs. Since our data collection began before their publication, we could not make a direct comparison of thoracic shapes. Although there is some overlap in the morphology categories (brachymorphic, dolichomorphic, and mesomorphic), direct comparison was not feasible. However, BCS remained in our model as in theirs. Our results indicated that the effect of morphology on C_dyn was not retained in the multivariable analysis, possibly because it did not provide additional relevance beyond the internal diameter of the orotracheal tubes. There was no correlation between morphology and the internal diameter of the orotracheal tubes used in our study, so this relationship was not investigated further.

This study has several limitations. Firstly, while a variety of spirometry technologies and monitors are commonly used in veterinary anesthesia (1), near-patient spirometers using Pedi-lite or D-lite flow sensors (Datex Ohmeda/GE Healthcare) were overrepresented. Additionally, several commonly available monitors were not used in this data collection. Unfortunately, no conclusions can be drawn regarding the impact of some monitors that were either underrepresented or not included in this study. Nevertheless, our model demonstrated an effect of technology used to determine C_dyn in this study and thus support the importance of considering the type of sensor when measuring compliance.

Secondly, while we acknowledge the potential for clustering due to the multicentre nature of the study, our primary aim was to assess dynamic compliance across a broad, representative sample rather than focus on center-specific effects. Incorporating a center variable would have required additional adjustment for the variability in machines used across centers, complicating the analysis without contributing significant clinical insight. Nonetheless, future studies could benefit from evaluating clustering effects, especially in contexts where center-based differences are of clinical relevance.

Thirdly, although calibration check of the monitors and correction of the data were originally planned, it proved practically impossible. It is likely that some of the monitors used in this study were not calibrated. However, this seems to represent typical veterinary anesthesia practice (1) thus the results similar expected findings in this environment.

Fourthly, although our survey among veterinary anesthetists documented the potential effect of some categories on C_dyn a few of them were not included in the analysis because they were not sufficiently represented (Trendelenburg; breathing system type; the administration of etomidate, desflurane, and intravenous lidocaine). Additionally, the limited number of cases in some categories may have caused certain variables, like PEEP levels, to appear insignificant due to under-representation, despite their expected effect on C_dyn. Further study is required to determine their possible effects on models used to calculate C_dyn.

Fifthly, the absence of preoperative imaging studies to assess for underlying respiratory diseases may limit the interpretation of the results, as undetected respiratory conditions could influence C_dyn measurements.

Finally, the diagnostic tests performed on our multiple linear regression model yielded mixed results. The assumption of linearity was confirmed, indicating that the relationship between the predictor variables and the response variable was appropriately modeled as linear. Additionally, the independence of residuals was verified, suggesting that there are no significant autocorrelation issues. Importantly, multicollinearity was not a concern, ensuring that the predictor variables are not highly correlated, and the regression coefficients remain stable. However, the tests indicated violations in both homoscedasticity and the normality of residuals. These findings suggest that the variance of the residuals is not constant, and that the residuals do not follow a normal distribution, respectively. These violations may affect the efficiency and validity of our model's estimates and hypothesis tests. To address potential bias in the linear regression coefficient estimates, bootstrapping was applied, and confidence intervals were calculated. This approach is commonly used to provide more robust estimates and mitigate the impact of violations in residual normality. Furthermore, due to the presence of heteroscedasticity, robust standard errors were calculated, and p-values accordingly adjusted. While caution is warranted when extrapolating the presented results, the application of these calculations has yielded more reliable estimates for further inference.

Many veterinary anesthetists employ P-V loops and specifically monitor respiratory system compliance (1). This study revealed that C_dyn should be evaluated relative to body mass. While a definitive reference interval was not established and may lack clinical utility, the study demonstrated that C_dyn was significantly influenced by several factors: the internal diameter of orotracheal tubes, BCS, inspiratory time settings, the specific sensor used for C_dyn determination, and exposure to FIO₂ levels exceeding 80% for at least 10 minutes before assessment. Minimal inspiratory setting characteristics (such as inspiratory time, or inspiratory-to-expiratory ratio and set respiratory rate) should be considered when clinically comparing C_dyn over time and systematically reported in research.