Using AI to expose the mechanisms of exertional dyspnoea in COPD: conflating “excessive” and “constrained” ventilation during incremental CPET

<bold>Background:</bold> There is a need of novel paradigms to interpret cardiopulmonary exercise testing (CPET) which properly considers the neurobiological underpinnings of exertional dyspnoea. <bold>Aim:</bold> We tested a conceptual framework which relates complementary metrics of demand-capacity imbalance of the respiratory system with the severity of dyspnoea as exercise intensifies. <bold>Methods:</bold> An AI-based software (<italic>DyVe-X</italic>) determined the overall burden (i.e., considering all CPET data points) of sex- and age-adjusted metrics of “excessive ventilation” (↓ dynamic ventilatory reserve (VR_dyn)), “constrained ventilation” (↓ dynamic inspiratory reserve based on tidal volume/inspiratory capacity (IR_dyn1) or inspiratory reserve volume/total lung capacity (IR_dyn2)), and ↑ dyspnoea (Borg 0-10) in 359 men and women with mild to end-stage COPD <bold>Results:</bold> Severe decrements (<25^th centile) in VR_dyn, IR_dyn1, and IR_dyn2 showed high positive predictive values for “intense” dyspnoea-work rate (>75^th centile); <italic>p</italic><0.001)). Binary logistic regression, showed that IR_dyn2 added information to the simpler IR_dyn1 and VR_dyn to predict “intense” dyspnoea-work rate (odds ratio (95% CI)= 3.26 (1.09-9.74), 21.6 (4.7-92.9), and 10.4 (3.03-35.6) respectively; 86.1 % correct). Of note, IR_dyn1 and IR_dyn2 – but not VR_dyn- were highly predictive of “intense" dyspnea-ventilation (22.8 (7.37-70.7) and 9.6 (3.1-29.7) respectively; 89.1 % correct), i.e., "constrained ventilation". <bold>Conclusions:</bold> By combining indexes of “excessive” and “constrained” ventilation throughout incremental CPET, <italic>DyVe-X</italic> may provide unique insights into the physiological underpinnings of exertional dyspnoea in individual patients.