Development of virtual patients for use in mechanical ventilation
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
Mechanical ventilation (MV) is a core life-support therapy for patients suffering respiratory failure or acute respiratory distress syndrome (ARDS). Respiratory failure is a secondary outcome of a range of injuries and diseases, and results in almost half of all intensive care unit (ICU) patients receiving some form of MV. Funding the increasing demand for ICU is a major issue and MV, in particular, can double the cost per day due to significant patient variability, over-sedation, and the large amount of clinician time required for patient management. Reducing the cost in this area requires both a decrease in the average duration of MV by improving care, and a reduction in clinical workload.
Model-based methods offer a way of using individual patient physiology and response to MV to suggest optimal ventilator settings. In particular, models and data can be used to gain insight into specific lung mechanics, such as pulmonary elastance and resistance. Importantly, these models can be used to quantify aspects of patient lung physiology over time, capturing a patient’s time-dependent disease state as patient condition evolves. These models have the potential to enable predictive, personalised, and potentially automated, approaches to MV.
The research in this thesis explores creating a more effective method of clinical trial design for mechanical ventilation trials. The high level of patient variability and the non-normal distribution of the key clinical outcome, length of mechanical ventilation, means many MV clinical trials struggle to achieve statistical significance. As a result, very large sample sizes are required to achieve statistical power to prevent inconclusive findings that cannot be extrapolated to other care units. Equally, non-significant findings do not inform the field or allow it to improve. A Monte-Carlo simulation method was developed and used to investigate several outcome metrics of ventilation treatment. As these metrics have highly skewed distributions, it also included the impact of imposing objective clinically relevant exclusion criteria on study power to enable better design for significance. This method combined with the use of composite outcome metrics, such as ventilator free days, enables high powered studies to be developed with substantially lower sample size requirements, enabling better study design and outcomes.
This thesis primarily focusses on the development of virtual patients. Virtual patients are used to personalise and optimise care for each individual patient by predicting response to a change in treatment prior to implementing the change. This personalisation is especially critical for ICU patients, who exhibit a great deal of variation in condition, and response to treatment. In particular, these virtual patient models predict the effects of a recruitment manoeuvre (RM) on lung elastance to minimise the risk of ventilator induced lung injury (VILI) while also maximising lung recruitment, and thus oxygenation.
The model was developed using physiologically relevant basis function models describing the effects of alveolar recruitment, lung distension, and airway resistance on overall lung elastance and resistance over pressure, volume and flow. The goal was to ensure the virtual patients model predictions and information provided was clinically relevant. It was validated using clinical data from two diverse sets of data from trials in New Zealand (the CURE trial) and Germany (McREM trial). A high level of fitting accuracy (RMS error) was seen for predictions of PEEP changes of up to 10 cmH2O, indicating the selected basis functions accurately describe the behaviour of the dominant lung mechanics in an RM and could have potential as a diagnostic tool.
The model showed a high level of accuracy with predicted peak inspiratory pressure error (median [IQR]) of 6.3 [4.5 - 8.3]% in the CURE cohort and 6.2 [5.0 - 9.1] % in the McREM cohort, even for PEEP changes up to 10 cmH2O. This capability to accurately predict pressure so far ahead in an RM provides important clinical insight, as it can enable the clinician to assess early in a RM when increases should either be stopped, or when much smaller incremental changes should be made. This knowledge could significantly aid in the efficiency of RMs, reducing clinical workload and improving patient care and outcomes.
A less studied impact of increasing PEEP is the added pressure results in an increased end-expiratory (recruited) lung volume, or dynamic function residual capacity. It is essentially the residual additional lung volume additional lung volume (Vfrc), due to alveolar recruitment at this higher pressure. Determining Vfrc is invasive, typically requiring imaging that either cannot be carried out at the bedside or is not available in every care unit. A model-based method to predict additional recruited lung volume (Vfrc) gained throughout a recruitment manoeuvre was developed and validated against clinical data. Results were promising with high accuracy shown in both approximating Vfrc and using this information to predict lung behaviour at higher PEEP levels. The results offer a clinical opportunity to titrate PEEP based on the estimated lung volume recruited, a direct indication of the success of an RM. Combined with prediction of the point of minimum elastance and prediction of peak inspiratory pressure this information would allow clinicians to optimise the trade-off between the risk of VILI and lung recruitment, in real-time as patient condition evolves, improving patient care and outcomes.
The incorporation of virtual patient methods into mechanical ventilation will aid the healthcare sector in meeting increasing demand in intensive care units. In particular, a change from more generic protocols to the use of predictive, patient-specific models will improve individual patient outcomes while also reducing clinical workload. The efficacy of the physiologically relevant model in determining lung behaviour throughout an entire RM in ventilation indicates it could be used as a diagnostic clinical tool. The future use of virtual patients and cohorts will also allow new treatments and therapies to be safely and more efficiently tested, allowing for faster advancements in the field.