Breathing Easier

Mechanical ventilation (MV) is a primary therapy for intensive care unit (ICU) patients who have respiratory failure. Up to ~60% of all ICU patients require MV, and—because they are so ill—this patient group stays in the ICU 50–100% longer. They also cost almost 100% more due to this added care, and, troublingly, have a mortality rate of almost 30%.
MV provides external pressure/volume support to assist with breathing when the patient is at risk of airway or tissue closure or collapse, or when the drive to breathe is compromised. MV aims to maintain the patency of the airways and tissue, to reduce the work of breathing, and to provide ventilation at a sufficiently high rate for adequate blood oxygenation and elimination of carbon dioxide.
Of particular concern are MV patients who develop stiff and/or collapsed lungs during the course of their treatment. A stiffer (more elastic) lung requires a higher driving pressure to ventilate; however, because the stiffened or collapsed tissue is heterogeneous, the unaffected areas are susceptible to “overstretching” if the applied pressure is too high; conversely, a positive end-expiratory pressure (PEEP) that is too low can result in repeated opening and closing of airways, which is associated with additional lung injury. Patient-specific characteristics [age, gender, body mass index (BMI), and smoking history] and distribution of injury and underlying disease (if present) mean that MV patient response to therapy is highly variable, and, thus, a “one-size- fits-all” protocol to standardize care is not appropriate. In the absence of quantitative, validated tools for clinical guidance in MV, clinicians are often constrained to provide treatment based on their experience and intuition, and therapy for the individual patient changes primarily only when significant changes are noticed. So while therapy is specific to each patient, the care itself is not “patient-specific” and often doesn’t take account of the specific, transient needs of the patient, nor the patient’s respiratory status as it evolves in response to the injury and in response to treatment.
The main limitation is that it is not currently possible to titrate care to the desired clinical end points. Specifically, PEEP cannot be titrated to a desired level of measured recruitment or blood oxygenation or lung damage. What is needed is the ability to quantify and monitor the patient-specific state of the lung with regard to MV care, and the ability to do so at every breath so clinicians can be alerted to changes in patient response as soon as they occur. Mathematical modeling of human lung physiology provides an opportunity for clinicians to “see” a patient-specific and, in fact, breath-specific lung response to MV therapy in real time. The vision is that by using this unique approach, clinicians can individualize therapy based on patient-specific needs at every moment in time.

Modeling Breath in Real Time

CURE Soft is a new model-based software tool that has been developed to monitor patient-specific respiratory mechanics at every breath and the patient’s evolution of respiratory mechanics in response to care and the patient’s particular disease state, all in real time. The monitoring of respiratory mechanics provides a means of quantifying lung status and, thus, to individualize MV therapy. CURE Soft is currently part of a large randomized clinical trial to optimize MV care in the Christchurch Hospital ICU in New Zealand. Figure 1 shows the CURE Soft system connected to ventilators in the ICU as well as the basic lung model that it uses for bedside control and monitoring; Figure 2 illustrates how the system is used.

An example of CURE in action is shown for Patient 1 in Figure 3. After a recruitment maneuver, where pressure is raised and lowered to open new lung volume (i.e., to open lung tissue that was previously closed to ventilation), the PEEP that is set by the clinician is guided by the minimum model-based lung elastance to deliver maximum volume for minimum airway pressure and—in theory—minimum lung tissue stress. As shown in Figure 3, the patient’s condition improves and the required fractional inspired oxygen (FiO2) goes down even as blood oxygen saturation levels (SiO2) improve. Of course, for some patients, the opposite can occur: when monitoring the patient in Figure 4, one can see how the patient condition degrades over time. Elastance rises, indicating that the lung requires higher airway pressure—with increased risk of damaging lung tissue—to drive the desired volume of air into the lung. Detecting this degradation at a given pressure or PEEP level would be the first step in alerting clinical staff to provide a change in care, as in Figure 3.

Tomorrow’s Care today

The case study of model-based MV considered above is “tomorrow’s care coming today.” However, the question still remains as to what the “day after tomorrow” might bring and where we can go from here. Current medical practice focuses on a one-size-fits-all approach. Thus, while care may currently be consistent between patients, it is not necessarily optimal for the individual patient. There is an international push to develop “one (personalized)-method-fits-all” approaches that can cope with patient variations [1] and allow easy personalization of treatments for optimal outcomes. CURE Soft provides one approach toward this personalization. The CURE Soft model-based guidance for MV currently relies on a “singleunit” model. This works well for fitting, tracking, and predicting total lung elastance and resistance. However, it does not take account of the myriad combinations of tissue heterogeneity and airway closure distribution that could result in the fitted model values. The model currently does not include optimization of arterial blood gases or other clinical end points as objectives.
Consider two patients. Patient A is a 75-yearold male with a BMI < 20. He is a heavy smoker, has had repeated chest infections, and has pneumonia. Patient B is a 25-year-old female with moderate asthma and BMI > 30. She is postsurgical. Patient A’s age and smoking history make it likely that his tissue elastance is lower than the population average, he has some tissue abnormalities due to previous infections, he will be prone to airway closure, and he will have a marked ventilation-toperfusion (V/Q) mismatch at baseline. Patient B’s gender, asthma, and obesity mean she is likely to have relatively high baseline airway resistance and some airway closure. Her high BMI increases the likelihood of smaller-than-normal lung volume at rest, which translates to a more heterogeneous distribution of lung tissue elasticity than normal. Both patients are hypoxemic and hypercapnic, they have identical threshold values for airway opening and closing pressures, and their integrated lung elastance and resistance are the same. If their measured lung mechanics are the same, then should they both be treated with the same MV strategy? A single-unit, single-function (mechanics) model cannot provide sufficient information to support decision making in this case.
To address the complexity of device–lung interaction in the critical-care setting, the MedTech CoRE respiratory flagship (see “MedTech CoRE Seeks to Create Individualized Forms of Patient Care“) seeks to integrate biophysical models for mechanistic pathophysiological understanding with computationally simpler systems-level models (such as currently employed by CURE Soft) that are necessary for real-time execution at the bedside. This approach builds on an established body of research that has involved developing personalized physiological models of pulmonary function that are valid over the full life span [2], applying them in simulating standard pulmonary function tests [3], and supplementing in vivo clinical imaging with patient-specific functional information [4], along with modeling and identification methods to personalize bedside models with clinical data [5].
[accordion title=”MedTech CoRE Seeks to Create Individualized Forms of Patient Care”]
The development of this model-based technology is an example of how the Diagnostics and Therapeutics theme under the Medical Technologies Centre of Research Excellence (MedTech CoRE) seeks to use engineering science and technology to create new, individualized forms of patient care to improve outcomes. The MedTech CoRE and the other New Zealand Centres of Research Excellence are interinstitutional research networks funded by the New Zealand Tertiary Education Commission. The MedTech CoRE arose from the vision of the partners in the Consortium for Medical Device Technologies (CMDT). The CMDT is a national framework that couples biomedical research activities in universities with district health boards, regulatory agencies, the Health Innovation Hub, and the investment community, e.g., KiwiNet and Return on Science (see www.cmdt.org.nz). The CMDT saw an urgent requirement to bring together academic and clinical partners in the medical technologies sector to create a national platform that identifies needs, provides expert scientific research, and delivers solutions that are of economic and societal benefit, while simultaneously training skilled graduate students who understand how to translate basic research into commercial or clinical outcomes.
The MedTech CoRE represents a national collaboration of research groups that have a strong foundation of basic science research funding and strong international reputation in medical device or technology research. Several strategic “flagship” projects have been selected for initial investment by the CoRE (along with investment in seed projects, and “platform technologies” that underpin the CoRE’s scientific and translational activities). The flagship projects include respiratory and cardiac diagnostics and therapeutics, interventional technologies in gastrointestinal dysfunction and liver surgery, assistive technologies for stroke, and regenerative medicine. These areas have been selected because they are based on proven high-quality science, international collaborations, clinical engagement, and substantial links to industry.
The more than ten principal investigators and 150+ associate investigators include physical, biological, and computer scientists; mathematicians; and engineers from mechanical, electrical, and bioengineering disciplines. To ensure that the technical solutions from basic and translational research are of value to health care, the investigators also include clinicians, surgeons, radiologists, and others who work in both hospital and community health care environments. To further ensure that outcomes from the CoRE are relevant and translational, the MedTech CoRE works closely with industry partners in the CMDT.
[/accordion]
Structure-based biophysical models simulate physiological function by solving systems of equations that are based on conservation laws, with parameterization to (in this case) human physiology. The model equations are usually solved numerically, within computational domains that represent the complex tissue structure or organ anatomy. Ideally, they are data-driven and reflect mechanotransduction across spatial scales from organ to airway/ vessel and tissue, and from tissue to its constituent cells (as in the multiscale model in Figure 5). In the context of the MV lung, the models should also represent the interaction between solid and fluid mechanics (tissue deformation and blood or air transport), exchange (of respiratory gases, and, possibly, of heat and water vapor), and pertinent control mechanisms (e.g., hypoxic pulmonary vasoconstriction).

Now let us return to our two patients. A patient-specific structure- based model would allow us to predict the spatial distribution of airway opening and closure, regional tissue stress, and arterial blood gases, in response to a range of MV pressure scenarios and the patient’s FiO2. This would support the clinician in selecting the optimal ventilation strategy for the criteria that they deem most important. The sensitivity of the respiratory system to increased injury could be assessed and used as guidance for alerting the clinical team when a critical threshold for change in the patient’s lung mechanics or blood gases is reached. In silico testing could also be used to guide patient weaning from MV.
Unfortunately, the biophysical models described here are complex, time-consuming to derive, difficult to fully parameterize to individual patients, and impractical for executing at the bedside. The key to taking advantage of their much greater predictive power is to derive a model that is intermediate between the single-unit model and the full, complex model and that captures the key responses of the lung to MV. The model must be reducible to the single unit for rapid fitting of model parameters at the bedside and scalable to the complex model for offline interrogation over a longer time period.
Patient-specific modeling with seamless transition between scales of model complexity is our vision for the “day after tomorrow” in clinical decision support for MV, delivering an unobtrusive, noninvasive method to improve clinical outcomes, reduce the length of stay in ICUs, and reduce health care costs.

References

1. G. M. Shaw and J. G. Chase, “Why evidence based medicine may be bad for you and your patients,” in Care Update 2006, V. Nayyar, Ed. New Dehli: Jaypee Medical Publishers, 2007, pp. 9–20.
2. M. H. Tawhai, A. R. Clark, and K. S. Burrowes, “Computational models of the pulmonary circulation: Insights and the move towards clinically directed studies,” Pulm. Circ., vol. 1, pp. 224–38, Apr.–June 2011.
3. A. J. Swan, A. R. Clark, and M. H. Tawhai, “A computational model of the topographic distribution of ventilation in healthy human lungs,” J. Theor. Biol., vol. 300, pp. 222–231, May 2012.
4. A. R. Clark, D. Milne, M. Wilsher, K. S. Burrowes, M. Bajaj, and M. H. Tawhai, “Lack of functional information explains the poor performance of ‘clot load scores’ at predicting outcome in acute pulmonary embolism,” Respir. Physiol. Neurobiol., vol. 190, pp. 1–13, Sept. 2013.
5. P. D. Docherty, C. Schranz, J. G. Chase, Y. S. Chiew, and K. Moller, “Utility of a novel error-stepping method to improve gradient-based parameter identification by increasing the smoothness of the local objective surface: A case-study of pulmonary mechanics,” Comput. Methods Programs Biomed., vol. 114, no. 3, pp. e70–e78, Aug. 2013.
6. A. M. Lauzon, J. H. Bates, G. Donovan, M. Tawhai, J. Sneyd, and M. J. Sanderson, “A multi-scale approach to airway hyperresponsiveness: from molecule to organ,” Front. Physiol., vol. 3, pp. 191, 2012.