Reply to Li et al.

From the Authors:

We read with great interest the letter by Li and colleagues and thank them for their interest in our work (1).

We fully agree that tidal volume (VT) that occurred during high-flow nasal cannula (HFNC) may not be the same as the VT measured with noninvasive ventilation (NIV) used at low degrees of support. However, this does not affect VT measured with NIV response to an early increase in respiratory drive in each patient. Although the accuracy of VT remains controversial, it should be pointed out that our study is merely a proof of concept to demonstrate that an index incorporating VT performs better than an index relying on respiratory rate.

To the best of our knowledge, there does exist only a few monitoring techniques that enable VT measurement under HFNC, such as a time-of-flight camera (2). However, it has not been widely used at the bedside. Tidal impedance change measured by electrical impedance tomography (EIT) correlates well with VT in individuals. However, such a correlation is not constant among patients. A universal cutoff value reflecting respiratory drive cannot be derived from EIT data alone. Although this method is not perfect, monitoring VT with low support level NIV in patients with HFNC is thus far a clinically acceptable and simple method. If well prepared, switching HFNC to NIV takes only a few seconds and may not result in derecruitment of the alveoli. Nevertheless, alternative ways to evaluate VT in patients under HFNC should be explored.

The second point raised by Li and colleagues is that for the calculation of Volume-OXygenation (VOX), the absolute VT might not be necessary. It is our opinion that the absolute value of VT is essential to estimate the absolute value of an early increase in the respiratory drive during HFNC. The parameter ΔEMOX proposed by Li and colleagues, that uses EIT-based VT and respiratory rate (EIT-based min volume [MV]) at 1 hour after HFNC, only reflects the change in respiratory drive before and 1 hour after HFNC but not the absolute increase in respiratory drive in each patient. Therefore, it is not surprising that ΔEMOX has only a moderate value in predicting HFNC failure (area under the receiver operating characteristic curve [AUROC], 0.72).

To verify that absolute elevation of VT rather than changes in VT after HFNC is a better predictor of HFNC failure, we retrospectively analyzed the data from our previous study, mimicking Dr. Li and colleagues’ ΔEMOX formula. The parameter ΔMOX (MV-OXygenation) at 2 hours and 6 hours after HFNC was calculated as follows (Sp_O2 = oxygen saturation as measured by pulse oximetry):

$$\mathrm{\Delta }{\text{MOX}}_{\text{2h}}\hspace{0.17em}=\hspace{0.17em}([{\text{Sp}}_{{\text{O}}_2}{}_{,\text{2h}}\hspace{0.17em}/\hspace{0.17em}({\text{Fi}}_{{\text{O}}_2,\text{2h}}\hspace{0.17em}\times \hspace{0.17em}{\text{MV}}_{\text{2h}})]\hspace{0.17em}-\hspace{0.17em}[{\text{Sp}}_{{\text{O}}_{\text{2}},0\text{h}}\hspace{0.17em}/\hspace{0.17em}({\text{Fi}}_{{\text{O}}_2,0\text{h}}\hspace{0.17em}\times \hspace{0.17em}{\text{MV}}_{0\text{h}})])\hspace{0.17em}\times \hspace{0.17em}{\text{MV}}_{0\text{h}}$$

$$\mathrm{\Delta }{\text{MOX}}_{\text{2h}}\hspace{0.17em}=\hspace{0.17em}([{\text{Sp}}_{{\text{O}}_2,\text{6h}}\hspace{0.17em}/\hspace{0.17em}({\text{Fi}}_{{\text{O}}_2,\text{6h}}\hspace{0.17em}\times \hspace{0.17em}{\text{MV}}_{\text{6h}})]\hspace{0.17em}-\hspace{0.17em}[{\text{Sp}}_{{\text{O}}_2,0\text{h}}\hspace{0.17em}/\hspace{0.17em}({\text{Fi}}_{{\text{O}}_2,0\text{h}}\hspace{0.17em}\times \hspace{0.17em}{\text{MV}}_{0\text{h}})])\hspace{0.17em}\times \hspace{0.17em}{\text{MV}}_{0\text{h}}.$$

ΔMOX, which reflects the change of respiratory drive after HFNC, was able to distinguish HFNC failure only at 6 hours but not at 2 hours after HFNC. The AUROC was 0.59 at 2 hours and 0.79 at 6 hours. The sensitivity and specificity were 58.6% and 93.9% at 6 hours after HFNC, respectively. However, the AUROC of VOX index, which reflects the absolute increase of respiratory drive, was much higher (0.93) at 6 hours after HFNC.

On the other hand, more severe patients enrolled in our study, compared with the study by Li and colleagues (3), might be the main reason for the poor predictive value of ΔMOX after a short time of HFNC (2 h). Compared with the report by Li and colleagues, a higher acute physiologic assessment and chronic health evaluation II score (21 vs. 18) and a higher intubation rate (46.7% vs. 22.9%) were observed in patients in our study.

Therefore, as the use of a change of MV does not provide any additional benefit in terms of prediction of HFNC failure, the use of the normal VOX index may be recommended.