Volume–OXygenation Index to Predict High-flow Nasal Cannula Failure: How to Capture the Tidal Volume Matters

To the Editor:

We read with great interest the article by Chen and colleagues (1) in which they proposed a modified novel index volume–OXygenation (VOX), as calculated using the tidal volume (VT), as compared with the original ROX index [SpO₂ (pulse oximetry)/FiO₂ over respiratory rate (RR)] with RR. Better performance with higher sensitivity and specificity and a larger area under the receiver operating curve at an earlier phase (2 h and 6 h) after high-flow nasal cannula (HFNC) was initiated. We congratulate the authors for the nicely conducted pilot study demonstrating the important role of VT in the failure of HFNC in patients with acute hypoxemic failure. Nevertheless, as pointed out by Chen and colleagues, the VT measurement with noninvasive ventilation (NIV) and interruption of HFNC was not ideal. We would like to continue the discussion on this topic.

VT Occurring During HFNC May Not Be the Same as the VT Measured During NIV

To illustrate our concern, we have conducted a similar measurement as described in the study (1). A patient with acute hypoxemic failure was treated with HFNC (Optiflow, Fisher and Paykel). HFNC was interrupted and switched to NIV (Respironics V60, Philips) with inspiratory support of 5 cm H₂O and positive end-expiratory pressure amount of 5 cm H₂O. To compare the VT during HFNC and NIV, electrical impedance tomography (EIT; PulmoVista500, Draeger Medical) was used, and the ventilation changes were tracked (2). VT during noninvasive ventilation was 640 ml on average. When the impedance changes normalized to volume, it yielded VT 472 ± 65 ml for HFNC and 640 ± 74 ml for NIV. VOX was 30.4 instead of 41.2 if VT from NIV was used instead of HFNC. The cut-off values proposed in the study by Chen and colleagues were optimized for VT measured during NIV but not for HFNC.

For the Calculation of VOX, the Absolute VT Might Not Be Necessary

One disadvantage of HFNC interruption is that the “positive pressure” effect induced by HFNC will disappear within 10 seconds (as indicated with end-expiratory lung impedance in EIT).

On the other hand, EIT can be used to identify overdistention via monitoring VT distribution during HFNC continuously (3). This provides an objective measure for real-time VT changes and respiratory rate. In a previous study, we attempted to use EIT for early prediction (within 1 h) of HFNC failure (4). As inspired by the work (1), we combined ROX and VOX indices and retrospectively analyzed the data from the previous study (4) using EIT-based VT and RR (EIT-based min volume [MV]) at 1 hour after HFNC. This parameter ΔEMOX was calculated as follows:

$$\mathrm{\Delta }\mathit{EMOX}\hspace{0.17em}=\hspace{0.17em}[{\text{SpO}}_{2,\text{1h}}/({\text{FiO}}_{2,\text{1h}}\hspace{0.17em}\times \hspace{0.17em}{\text{MV}}_{\text{EIT},\text{1h}})\hspace{0.17em}-\hspace{0.17em}{\text{SpO}}_{2,\text{0h}}/({\text{FiO}}_{2,\text{0h}}\hspace{0.17em}\times \hspace{0.17em}{\text{MV}}_{\text{EIT},0\text{h}})]\hspace{0.17em}\times \hspace{0.17em}{\text{MV}}_{\text{EIT},0\text{h}}.$$

ΔEMOX at 1 hour after HFNC was able to distinguish HFNC failure (P < 0.05). The area under the receiver operating characteristic curve was 0.72 (same as ROX_1h). The sensitivity and specificity were 51.4 and 100, respectively, versus 77.1 and 63.6 for ROX_1h.

The VT calculation in the study (1) averaged the volume within 1 minute during stable NIV, whereas MV combines both VT and respiratory rate that reflects the respiratory drive within 1 minute. Given that respiratory efforts during spontaneous breathing could be assessed by EIT (5) and EIT data might be connected to the centralized ICU system, the combination of ROX and VOX with EIT-based MV might be more practical in clinical routine, that requires further validation.