Computer model-based injury prediction and evaluation of lung function in mice with acute and ventilator-induced lung injuries.

Abstract

Invasive mechanical ventilation is a lifesaving intervention for acute respiratory distress syndrome (ARDS) patients but also causes ventilator-induced lung injury (VILI) that can be challenging to avoid due to inter-patient and temporal heterogeneity. Thus, the aim of this study was to characterize and predict experimental VILI using readily available measures of lung function. Initially healthy (CTL) and hydrochloric acid (HCL) lung-injured mice were ventilated for four hours at PEEP = 1, 3, or 8 cmH₂O to produce graded VILI severity as measured in lung function, alveolocapillary leak, and inflammation. Optimally protective PEEP was found to be 8 and 3 cmH₂O in the HCL and CTL groups, respectively. A novel computational model was fit to the data to investigate elastance dynamics described by the "compliance factor" (C_F) which was also used to predict VILI over four subsequent hours of ventilation. The model C_F is a sensitive measure of injury-induced alterations in the pressure- and pressure-history dependence of lung elastance that are known to correlate with recruitment and derecruitment dynamics. The C_F was then combined with PEEP and plateau pressures calculated from 10 minutes at the start of prolonged ventilation and used to accurately predict VILI outcomes measured four hours later. This model outperformed other commonly utilized measures of injury such as driving pressure and mechanical power. The computer model provides a new tool for understanding lung dynamics and for predicting VILI. In future work, this approach could be used to guide identification of lung protective ventilation settings.