Physiological effects of high-flow oxygen in tracheostomized patients

Background High-flow oxygen therapy via nasal cannula (HFOTNASAL) increases airway pressure, ameliorates oxygenation and reduces work of breathing. High-flow oxygen can be delivered through tracheostomy (HFOTTRACHEAL), but its physiological effects have not been systematically described. We conducted a cross-over study to elucidate the effects of increasing flow rates of HFOTTRACHEAL on gas exchange, respiratory rate and endotracheal pressure and to compare lower airway pressure produced by HFOTNASAL and HFOTTRACHEAL. Methods Twenty-six tracheostomized patients underwent standard oxygen therapy through a conventional heat and moisture exchanger, and then HFOTTRACHEAL through a heated humidifier, with gas flow set at 10, 30 and 50 L/min. Each step lasted 30 min; gas flow sequence during HFOTTRACHEAL was randomized. In five patients, measurements were repeated during HFOTTRACHEAL before tracheostomy decannulation and immediately after during HFOTNASAL. In each step, arterial blood gases, respiratory rate, and tracheal pressure were measured. Results During HFOTTRACHEAL, PaO2/FiO2 ratio and tracheal expiratory pressure slightly increased proportionally to gas flow. The mean [95% confidence interval] expiratory pressure raise induced by 10-L/min increase in flow was 0.2 [0.1–0.2] cmH2O (ρ = 0.77, p < 0.001). Compared to standard oxygen, HFOTTRACHEAL limited the negative inspiratory swing in tracheal pressure; at 50 L/min, but not with other settings, HFOTTRACHEAL increased mean tracheal expiratory pressure by (mean difference [95% CI]) 0.4 [0.3–0.6] cmH2O, peak tracheal expiratory pressure by 0.4 [0.2–0.6] cmH2O, improved PaO2/FiO2 ratio by 40 [8–71] mmHg, and reduced respiratory rate by 1.9 [0.3–3.6] breaths/min without PaCO2 changes. As compared to HFOTTRACHEAL, HFOTNASAL produced higher tracheal mean and peak expiratory pressure (at 50 L/min, mean difference [95% CI]: 3 [1–5] cmH2O and 4 [1–7] cmH2O, respectively). Conclusions As compared to standard oxygen, 50 L/min of HFOTTRACHEAL are needed to improve oxygenation, reduce respiratory rate and provide small degree of positive airway expiratory pressure, which, however, is significantly lower than the one produced by HFOTNASAL.

With HFOT NASAL , up to 60 L/min of heated and humidified air/oxygen mixture are continuously delivered to the patient through specifically designed nasal prongs [11]. Unlike standard oxygen, high flows limit dilution of inhaled gas mixture, thus enabling more accurate delivery of the set fraction of inspired oxygen (FiO 2 ) [12]. Natalini et al. Ann. Intensive Care (2019) 9:114 HFOT NASAL increases end-expiratory lung volume due to the generation of flow-dependent airway positive pressure, with highest values reached at end-expiration with closed mouth [13][14][15]. The continuous high flow washes CO 2 out from upper airways, reducing anatomical dead space and work of breathing [16]. Active heating/humidification and the comfortable interface improve comfort related to airway dryness and optimize device tolerability [16][17][18].
High-flow oxygen can be delivered also through tracheostomy (HFOT TRACHEAL ), but its mechanism of action and physiological effects appear different and have not been thoroughly elucidated [19,20]. We conducted a randomized cross-over study to assess the effects of HFOT TRACHEAL administered at different gas flow rates on gas exchange, tracheal pressure, and respiratory rate, and to establish whether the increase in airway pressure generated by high-flow oxygen is different when administered by nasal cannula or tracheostomy.

Methods
The present study was carried out in the general intensive care unit (ICU) of a tertiary-care university hospital in Rome between September 2016 and September 2017, after a preliminary study conducted on a previous cohort of patients to assess the feasibility of tracheal pressure measurement in critically ill patients [21]. The study protocol was approved by the local institutional review board; written informed consent was obtained by all patients or next of kin, according to the ethics committee recommendations.

Patients
We studied critically ill tracheostomized patients with no hemodynamic instability who had been weaned from mechanical ventilation, had been spontaneously breathing with no ventilatory support for at least 24 h and were receiving tracheal oxygen according to the prescription of the attending physician. All enrolled patients had received single-dilator percutaneous tracheostomy with PercuTwist ® technique (Rüsch, Kernen, Germany): the procedure was performed by an intensivist under bronchoscopy, which confirmed that the puncture was taking place between the first and second, or second and third, tracheal rings [22,23]. Non-inclusion criteria were age < 18 years, pregnancy, recent tracheal, esophageal, neck or thoracic surgery, presence of pneumothorax/chest drainage. For safety reasons, patients with partial pressure of arterial oxygen to nominal FiO 2 ratio (PaO 2 /FiO 2 ) below 100 mmHg and/or respiratory rate > 45 breaths per minute during standard oxygen were not enrolled.

Procedures
After study inclusion, each patient received for 30 min standard oxygen through tracheostomy with a heat and moisture exchanger (Tracheolife II HME, Mallinckrodt, United Kingdom), with oxygen flow set by the attending physician (standard oxygen step, maximal O 2 flow 8 L/ min).
Patients subsequently underwent high-flow oxygen: gas flow was provided by the dedicated module of an ICU ventilator (EvitaXL or EvitaInfinity, Drager, Lubeck, Germany), inspired gas was actively conditioned by heated humidifier set at 37 °C (HH MR850, Fisher & Paykel Healthcare, New-Zealand, absolute humidity provided 44 mgH 2 O/L) and delivered through the specifically designed interface (Optiflow ™ Tracheostomy interface OPT870, Fisher & Paykel Healthcare, New-Zealand). Three oxygen flow rates with the HFOT TRACHEAL device were tested in random order, for 30 min each: 10 L/min, 30 L/min, and 50 L/min. No wash-out period was applied between these interventions. Although 10 L/min cannot be considered as 'high-flow therapy' , this step allowed (A) to better characterize the effects of increasing flow rate with the same device on analyzed endpoints, and (B) to compare standard oxygenation device (closed system through a heat a moisture exchanger) and HFOT TRACHEAL (open system) at similar gas flow rate, highlighting the difference between these techniques. The randomization sequence was provided by S.A.S. random allocation software. FiO 2 was set to obtain a SpO 2 between 92 and 98% (88-92% in patients with PaCO 2 ≥ 45 mmHg during standard oxygen). Changes in the FiO 2 over the course of the study were discouraged and allowed only whether clinically unavoidable.

Measurements
At the end of each step, hemodynamic parameters, arterial blood gases and SpO 2 were recorded. To estimate PaO 2 / FiO 2 during standard oxygen, delivered FiO 2 was calculated using a previously described formula [24]: At study entry, a sterile, disposable 18-gauge catheter (15/25-cm length according to patient's height; 1-mm diameter; BD, CareFusion corporation, San Diego, CA, USA) connected to a differential pressure transducer was inserted in the trachea (2 cm away from carina, with the distance between tracheal stoma and carina measured on the chest X-ray) and secured to the skin with an adhesive tape. At the end of each study step, endotracheal pressure was recorded continuously for 3 min by a dedicated software at a sample rate of 200 Hz (Kleis-Tek, ICU lab, Bari, Italy). Pressure signals were offline-reviewed to assess respiratory rate and compute mean expiratory pressure (between the end of inspiration and the beginning of the following inspiration), peak expiratory and inspiratory pressure (maximal and minimal pressure achieved over the whole respiratory cycle, respectively). All these FiO 2 = oxygen flow rate in liters per minute * 0.03 parameters were measured for all breaths in the 3-min recording and values were averaged for each study step.
In a subgroup of five patients who underwent tracheostomy decannulation after study inclusion and during the ICU stay, the experimental protocol was repeated on the day of decannulation, both during HFOT TRACHEAL and during HFOT NASAL after decannulation. Briefly, when the tracheal cannula was removed, the catheter for tracheal pressure measurement was hold in situ and the stoma was covered with gauze and adherent sealing tape (percutaneous tracheostomy maintains subcutaneous tissue integrity and elasticity) [25]. After medication, absence of leaks through the stoma was assessed by hand while the patient spontaneously vocalized and coughed. This approach was clinically useful for assessing patient's tolerance to mouth/nose breathing and represented a unique opportunity to evaluate lower airway pressure during HFOT NASAL . In these 5 patients, HFOT TRACHEAL and HFOT NASAL with three flow settings (10, 30 and 50 L/min) were applied for 20-min periods in sequential order, just before and immediately after tracheostomy decannulation. No wash-out period was applied between the interventions. Heated humidifier settings were kept unchanged. Towards the end of each period, tracheal pressure tracings were recorded and were offline-analyzed to compute mean and peak expiratory pressure, as previously described.

End-points
Primary endpoint was to compare ratio of arterial oxygen partial pressure to nominal FiO 2 (PaO 2 /FiO 2 ) in the different study steps. Main secondary endpoints were to analyze the effects of the tested settings on respiratory rate, endotracheal pressure and PaCO 2 . Furthermore, we aimed at establishing whether tracheal pressure is different when high-flow oxygen is delivered through tracheostomy or nasal cannula, at similar flow rates.

Statistical analysis
Descriptive data are expressed as number and percentage and continuous data as median [interquartile range]. Because of the limited sample, adopting a conservative approach, all data were analyzed with non-parametric tests. Paired comparisons between the study steps were performed with the Wilcoxon sum of ranks test and mean differences [95% confidence interval] are displayed for most significant results. Correlation was assessed with Spearman's rank-order correlation: ρ and the p value are reported. Analysis on the mean expiratory pressure rise induced by increasing gas flow was performed with linear regression: the slope and the p value of the relationship are reported. Inter-individual variability was rated with the coefficient of variation, computed as the ratio of standard deviation to mean of the measurements [26]. Results with two-tail p ≤ 0.05 were considered significant. Statistical analysis was performed with SPSS 20.0 (IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY, USA).

Sample size
Clinical data on the effects of HFOT TRACHEAL are limited to a single exploratory study [20]: this hampered any estimation of the adequate sample needed to provide sufficient statistical power to the study. Because previous investigations with similar design demonstrate that 15-20 patients studied in a cross-over fashion represent an adequate sample to draw conclusions on similar physiological endpoints [13,15,18,20,27], adopting a conservative approach, we planned to enroll 25 patients.

Results
Twenty-six patients were enrolled and analyzed. Demographics and most relevant clinical characteristics are reported in Table 1. In the standard oxygen step, median oxygen flow was 4 [3,4] L/min and median estimated FiO 2 was 0.33 [0. 33-0.37]. No patient experienced changes in heart rate or arterial blood pressure over the course of the study. The sequence of HFOT TRACHEAL interventions did not affect PaO 2 /FiO 2 (ρ = 0.05, p = 0.69) nor respiratory rate (ρ = 0.002, p = 0.99).

Gas exchange and respiratory rate
These results are displayed in Fig. 1.

Tracheal pressure
These results are displayed in Fig. 2.
In the three HFOT TRACHEAL steps, mean and peak expiratory pressures were proportional to the delivered gas flow (p < 0.001 for all comparisons). The mean [95% CI] expiratory pressure rise induced by 10-L/min increase in flow was 0.2 [0.1-0.2] cmH 2 O (ρ = 0.77, p < 0.001). As compared to standard oxygen, 50 L/min, but not other HFOT TRACHEAL settings, led to an increase in peak and mean expiratory pressures: peak pressure 1.  (Fig. 2a, b). Mean differences [95% CI] in peak and mean expiratory pressure between HFOT TRACHEAL 50 L/min and standard oxygen were 0.4 [0.2-0.6] cmH 2 O and 0.4 [0.3-0.6] cmH 2 O, respectively. Both peak and mean expiratory pressures were lower at HFOT TRACHEAL 10 L/min than during standard oxygen (both p < 0.001).
All HFOT TRACHEAL settings yielded less negative tracheal peak inspiratory pressure, as compared to standard oxygen (p < 0.001 for all the comparisons): this effect was magnified at 50 L/min (Fig. 2c).

Discussion
In the present cross-over study, we show that, as compared to standard oxygen, HFOT TRACHEAL mitigates the negative swing in airway pressure during inspiration, and, when flow is set at 50 L/min, ameliorates oxygenation and slightly reduces respiratory rate. With similar flow rates, tracheal expiratory pressure is significantly lower with HFOT TRACHEAL than with HFOT NASAL , suggesting that the physiologic effects of HFOT TRACHEAL are milder than HFOT NASAL . A gas flow of 50 L/min should be set with the tracheal interface to slightly improve oxygenation and reduce respiratory rate. Several studies addressed the effects of HFOT NASAL in a variety of clinical scenarii [1]. Although high-flow oxygen can be delivered through tracheostomy, few data elucidate its mechanisms of action, which can be different from HFOT NASAL [20].

Oxygenation
During HFOT TRACHEAL , PaO 2 /FiO 2 ratio increases proportionally to gas flow. However, when compared to standard oxygen via heat and moisture exchangers, only 50 L/min generate improvement in PaO 2 /FiO 2 ratio. These data are partially consistent with what has been reported for HFOT NASAL [18] and may be explained by the following mechanisms: 1. Increasing flow rate up to 50 L/min can limit air dilution of inhaled gas mixture, enabling more accurate delivery of set FiO 2 . This can be demonstrated by the reduction of the inspiratory airway pressure swing during HFOT TRACHEAL . 2. Increasing flow rate yields a concomitant increase in peak and mean expiratory pressure. Although the increase in tracheal pressure generated by HFOT TRACHEAL is lower than the one reported during HFOT NASAL [11,14,15,28], this rise in expiratory pressure may still contribute to increase end-expir- Fig. 2 Peak (a), mean expiratory pressure (b) and negative peak of inspiratory pressure. Results are displayed as median, interquartile range, maximum and minimum. During HFOT TRACHEAL , tracheal expiratory pressure increases proportionally to the gas flow. All HFOT TRACHEAL settings limit the negative inspiratory pressure, especially as flow is set at 50 L/min, likely due to the capability of the high gas flow in an open system to match patient's peak inspiratory flow. As compared to standard oxygen, 50 L/min, but not 30 L/min nor 10 L/min, increase tracheal peak and mean tracheal expiratory pressure atory lung volume, reduce shunt fraction, optimize lung mechanics and improve oxygenation [11,13,18,29].
One previous report showed that, when compared to T-Piece with a Venturi generator in tracheostomized patients, airway pressure and SpO 2 /FiO 2 slightly increase during 50 L/min HFOT TRACHEAL [20]. However, because of the entrainment effect, Venturi systems can provide flows up to 30-50 L/min and cannot be considered standard oxygen devices [30]. Standard oxygen through heat and moisture exchangers represents a widely used alternative for oxygen therapy in tracheostomized patients. Fig. 3 Thirty-second recordings of tracheal pressure tracings during HFOT TRACHEAL and HFOT NASAL in 5 patients who underwent tracheostomy decannulation over the course of ICU stay. In both conditions gas flow was set at 50 L/min. Average respiratory rate for the 30-s recording is reported for all conditions. During HFOT NASAL lower airway pressure during expiration is higher and more inter-individually variable than HFOT TRACHEAL , despite a non-dissimilar respiratory rate, which was calculated on the same 30-s recording. This suggests that the HFOT NASAL -induced increase in expiratory pressure depends not only on gas flow, but also on patient's expiratory pattern and, likely, on individual respiratory system mechanical properties. Please note that, under this condition, tracheal pressure was not constant over the course of the respiratory cycle and became negative during inspiration in 4 patients, which is different from what previously reported for pharyngeal pressure [14] Natalini et al. Ann. Intensive Care (2019) 9:114 We have shown that standard oxygen through heat and moisture exchangers produces positive expiratory pressure, which is comparable to the one obtained with 30 L/min of high-flow oxygen through an open system. In fact, oxygenation between these two settings was similar. For the same gas flow (≈ 10 L/min), oxygenation and tracheal expiratory pressure were higher with the standard oxygenation (closed system) than with the HFOT TRACHEAL device (open system). This suggests that the oxygenation changes are dependent on the amount of tracheal expiratory pressure. However, mechanisms of airway pressure generation may be different between the two devices: with standard oxygen, the increase in pressure depends on the expiratory resistance produced by the heat and moisture exchanger; while, during HFOT TRACHEAL , positive expiratory pressure is produced by patient's expiration against the delivered gas flow in an open system and airway pressure is more stable over the respiratory cycle (i.e., less negative during inspiration). In this context, avoidance of excessive negative inspiratory swings in airway (and pleural) pressure is important to mitigate the risk of negative pressure pulmonary edema, whose occurrence induces lung damage and worsens oxygenation [31].

CO 2 clearance
HFOT NASAL lowers inspiratory resistance and enhances anatomical dead space clearance with CO 2 washout [32,33], finally reducing work of breathing [11,13,27,34]. Our study shows that 50 L/min HFOT TRACHEAL lowers respiratory rate without changes in PaCO 2 , as compared to standard oxygen. A reduction in respiratory rate has been reported during HFOT NASAL [5,35] and has been linked to anatomical dead space clearance, increased tidal volume, diminished resistive work of breathing and, in chronic obstructive pulmonary disease patients, increased positive expiratory pressure [13,33,36].
Work of breathing reduction by HFOT NASAL is obtained at 30 L/min and is minimally enhanced by further increases in gas flow [18]: differently, 50 L/min of HFOT TRACHEAL are needed to generate effects on respiratory rate. It is, therefore, reasonable to hypothesize that, in tracheostomized patients: 1. lower anatomical dead space and inspiratory resistance reduce the size effect of the intervention, that consequently requires higher flows to generate a significant effect; 2. inspired and expired flows are forcedly unidirectional, thus clearing anatomical dead space and improving breathing efficiency [37]: this contributes to CO 2 washout independently from the device used for oxygen therapy, thereby mitigating the effect of HFOT TRACHEAL .
Our results are consistent with recent data indicating that HFOT TRACHEAL minimally affects neuro-ventilatory coupling, work of breathing and gas exchange after weaning from mechanical ventilation [19].

Differences with HFOT NASAL
Our comparison of HFOT TRACHEAL and HFOT NASAL in the same patients represented a unique opportunity to highlight the contribution of upper airway resistance to positive-pressure generation during HFOT NASAL . In fact, to our knowledge, no other data clarify the behavior of lower airway pressure during this treatment. The average expiratory pressure reported in our study is similar to what has been reported for pharyngeal pressure [11,15,28]. However, tracheal pressure during HFOT NASAL was not constant over the respiratory cycle and became negative during inspiration in 4 of the 5 studied patients, which is different from what has been reported on upper airway pressure [14]. Our results indicate that expiratory pressure in lower airways is higher and more inter-individually variable when high flows are delivered through nasal cannula than through tracheostomy. This suggests that the mechanism of expiratory pressure generation during high-flow oxygen is dependent not only on gas flow rate, but also on the greater resistance offered by upper airways and patient's expiratory flow. In tracheostomized patients, resistance is limited, and the generated pressure is minimal. Patient's expiratory flow has wide inter-individual variability according to the resistive and elastic properties of the respiratory system and to the eventual recruitment of expiratory muscles [38]: thus, the pressure produced by HFOT NASAL is variable among subjects, also if respiratory rate with HFOT TRACHEAL is similar (Fig. 3) [39].

Clinical consequences
Our study shows that the effects of HFOT TRACHEAL are milder than HFOT NASAL , likely because the dedicated interface is completely open. HFOT TRACHEAL allows to limit the negative swing in inspiratory airway pressure, but both the dead space washout and the generation of positive expiratory pressure are limited. From a clinical perspective, our findings suggest that a minimum gas flow of 50 L/min should be set during HFOT TRACHEAL to slightly improve oxygenation and reduce respiratory rate, as compared to standard oxygen. Whether these mild physiologic effects are cost-effective and may clinically benefit the management of tracheostomized patients cannot be established from our data and should be addressed in further investigations.

Limitations
First, we did not measure effectively delivered FiO 2 , as performed elsewhere [3]. As a result, the calculation of PaO 2 /FiO 2 ratio may be subject to errors, especially if lower flows are used [40]. Nevertheless, our approach is clinically reproducible and we used a formula that has recently been shown to provide satisfactory correlation with actual FiO 2 [24].
Second, we did not measure work of breathing by esophageal manometry [41]. However, esophageal catheter insertion in awake and spontaneously breathing patients may be challenging and eventually require some sedation. Importantly, during HFOT NASAL , changes in respiratory rate have been shown to reflect variations of the work of breathing [13,33].
Third, there was no wash-out period between the applied interventions during HFOT TRACHEAL . However, our approach is consistent with previous investigations on the topic [18], and the randomized order of the interventions should have mitigated any carryover effect on the observed results. Accordingly, the main outcomes of the study were not affected by the sequence of applied flow settings.
Fourth, during HFOT NASAL , absence of major leaks through the stoma was assessed by hand. Unfortunately, we had no other way to assess if minimal leaks were present. We believe, however, that even minimal leaks, if present, should not have affected tracheal pressure measurement. In fact, the tracheal pressure values we report are similar to nasopharyngeal pressure values measured in non-tracheostomized patients by others [13][14][15].
Finally, we showed that expiratory pressure increase due to HFOT NASAL has wide inter-individual variability.
Whether and to what extent expiratory flow limitation and expiratory muscles recruitment contribute to this is unknown and remains to be established in further investigations [38,42].

Conclusions
HFOT TRACHEAL generates small flow-dependent improvement in oxygenation and increases in tracheal expiratory pressure. When compared to standard oxygen, a minimum flow of 50 L/min is needed during HFOT TRACHEAL to improve oxygenation, increase expiratory pressure, limit inspiratory airway pressure swings and reduce respiratory rate. At same gas flow, HFOT NASAL produces higher expiratory pressure than HFOT TRACHEAL .