Digital twins suggest a mechanistic basis for differing responses to increased flow rates during high-flow nasal cannula therapy

Abstract

Background: Inconsistent responses to increased flow rates have been observed in patients with acute hypoxemic respiratory failure (AHRF) treated with high-flow nasal cannula (HFNC) therapy, with a significant minority in two recent studies exhibiting increased respiratory effort at higher flow rates. Digital twins of patients receiving HFNC could help understand the physiological basis for differing responses.

Methods: Patient data were collated from previous studies in AHRF patients who were continuously monitored with electrical impedance tomography and oesophageal manometry and received HFNC at flow rates of 30, 40 or 45 L/min. Patients, based on their responses to an increase in flow rate to 60 L/min, were categorised into two groups: five responders with reduced oesophageal pressure swings ΔP_es (- 3.1 cmH₂O on average), and five non-responders with increased ΔP_es (+ 2.0 cmH₂O on average). Two cohorts of digital twins were created based on these data using a multi-compartmental mechanistic cardiopulmonary simulator. Digital twins' responses to increased HFNC flow rates (60 L/min) were simulated with constant respiratory effort to assess changes in gas exchange and lung mechanics, and with varying respiratory effort to quantify their combined effects on lung mechanics and P-SILI indicators.

Results: The digital twins accurately replicated patient-specific responses at all flow rates. Responder digital twins showed a mean 20 mL/cmH₂O increase in lung compliance at higher flow rates, versus a 6 mL/cmH₂O decrease in compliance with non-responders. In digital twins of responders versus non-responders, increased flow rates produced a mean change in lung stress of - 1.5 versus + 1.2 cmH₂O, in dynamic lung strain of - 8.8 versus + 16.4%, in driving pressure of - 1.3 versus + 1.1 cmH₂O, and in mechanical power of - 0.8 versus + 1.2 J/min. Higher flow rate dependent positive end-expiratory pressure in digital twins of non-responders did not cause recruitment, and reduced tidal volumes due to higher functional residual capacities-to compensate for the resulting worsened gas-exchange, non-responders increased their respiratory effort, in turn increasing patient self-inflicted lung injury (P-SILI) indicators. In digital twins of responders, reductions in tidal volumes due to higher FRCs resulting from increased PEEP were outweighed by alveolar recruitment. This increased compliance and improved gas exchange, permitting reduced respiratory effort and decreases in P-SILI indicators.

Conclusions: Failure to reduce spontaneous respiratory efforts in response to increased HFNC flow rates could be due to a deterioration in lung mechanics, with an attendant risk of P-SILI.

Keywords: Acute hypoxemic respiratory failure; Digital twins; High flow nasal cannula; Non-invasive respiratory support.

Fig. 1

Patient data compared to simulator outputs

Fig. 2

Mean (standard deviation) of lung mechanics parameters and P-SILI indicators in responders (blue dashed line) and non-responders (red solid line)

Fig. 3

Digital twin analysis of non-responders and responders, average values at different flow rates and different levels of spontaneous respiratory effort