During HFNC oxygen therapy, flow settings have been shown to have a significant impact on short-term clinical outcomes in 32 studies [6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24, 38,39,40,41,42,43,44,45,46,47,48,49,50] and long-term outcomes in one study [27].

Similar to in vitro studies, the physiological effects of HFNC are found to be flow-dependent (Table 1) [13, 15,16,17,18,19,20, 38,39,40]. As HFNC flows increase from 0 to 40 L/min, tidal volume (Vt) increases [15,16,17, 38] and respiratory rate (RR) decreases [13, 15,16,17,18,19,20]. However, Okuda et al. reported no change in minute ventilation between HFNC flows of 0 and 50 L/min [15], and Parke et al. [18] reported that RR plateaued with HFNC flows > 40 L/min.

The flow-dependent PEEP effect from the in vitro studies has also been confirmed in healthy individuals, as end-expiratory esophageal pressures or hypopharyngeal pressure gradually increase when HFNC flows are increased (Fig. 5) [13, 16,17,18, 20, 39]. However, maintaining a constant PEEP with HFNC is challenging because it can significantly decrease with open-mouth breathing [13, 17, 20, 39]. When subjects opened their mouth, hypopharyngeal pressure dropped from 5.2 (3.5, 7.0) cmH2O to 1.1 (− 0.9, 2.4) cmH2O with HFNC set at 50 L/min [13], and nasopharyngeal pressure dropped from 6.8 to 0.8 cmH2O with HFNC set at 60 L/min [17]. Caution must be taken while using very high flows, such as 100 L/min, as it can provide nasopharyngeal pressure as high as 11.9 ± 2.7 cmH2O [18], and the tolerability is concerning.

The relationship between airway pressures and HFNC flow settings. HFNC, high-flow nasal cannula

Besides the potential for significantly elevated airway pressure, an uneven distribution of delivered gas across lung regions from different HFNC flows may also pose a risk for regional overdistension. Three studies evaluating ventilation distribution across lung regions from various HFNC flows using electrical impedance tomography (EIT) report that global end-expiratory lung impedance (EELI) increases as HFNC flow increases [17,18,19]. However, increases in EELI mainly occur in non-dependent regions of the lung. Plotnikow et al. reported an increase in EELI by 35% from baseline (room air) to HFNC set at 30 L/min and by 22% from 30 to 50 L/min in the non-dependent regions [19]. In the lung-dependent regions, the EELI only increased by 18% and 7.7%, respectively [19]. Since the non-dependent lung regions are most likely open normally, these findings suggest a potential risk of over-distending the non-dependent region, resulting in lung injury.

Three studies have investigated the effects of flows on swallow function among healthy volunteers during HFNC therapy [41,42,43]. Sanuki et al. [41] reported reduced latency time of the swallow reflex with HFNC flows being increased from 15 to 45 L/min when healthy volunteers swallowed 5 mL of distilled water over 3 s. Thus, they concluded that HFNC might enhance swallowing function [41]. However, Arizono et al. reported the opposite findings, as choking was observed when HFNC flows were ≥ 40 L/min in the 30 mL swallow test [42]. Additionally, they noted that swallowing efforts were greater with HFNC flows ≥ 20 L/min than 10 L/min [42]. Allen and Galek found a flow-dependent influence on the duration of laryngeal vestibule closure (dLVC) among their healthy volunteers [43]. Since LVC is a protective reflex that helps to prevent aspiration, the authors suggest that dLVC modulation from HFNC flows might help prevent aspiration [43]. Notably, a large variation of dLVC between HFNC flows of 50 and 60 L/min was found, underscoring the reality that further research on the impact of HFNC flows on swallow function is needed [43]. Oral feeding during HFNC therapy should be closely monitored, especially in severe hypoxemic patients who might need treatment escalation and those with dysphagia or at high risk for aspiration.

During HFNC therapy for patients with AHRF, oxygenation has been assessed using SpO2/FIO2 (SF) ratio [6, 50], PaO2/FIO2 (PF) ratio [7, 8], and the ROX index (= SF/RR) [6, 9, 10]. Three studies have reported that oxygenation improves as HFNC flows increase (Table 2) [6,7,8], while Zhang et al. [9] found no significant changes in ROX index between room air and HFNC flow of 60 L/min in patients with mild hypoxemia. Likewise, Mauri et al. reported that 30% (17/57) of AHRF patients had an unchanged or decreased ROX index when HFNC flows were increased from 30 to 60 L/min. Their further analysis revealed that the 17 patients had a higher SF ratio and ROX index at 30 L/min, compared to the other 40 patients who presented an increase in ROX index with increasing flow [10]. Interestingly, the same authors implemented a study 2 years earlier on similar patient populations, and they found that 30% of patients had decreased PF ratios after increasing flows from 30 or 45 to 60 L/min [7].

Similar to the findings in healthy individuals, increasing flows also improves global EELI in patients with AHRF (Additional file 1: Table S3, appendix p5) [7,8,9]. Increasing HFNC flow generates a greater end-expiratory lung volume and PEEP [45, 46], which may cause recruitment that mainly occurs in dependent lung regions. However, it may also generate overdistension that is more pronounced in non-dependent lung regions. It appears that changes in oxygenation, that correlate with changes in EELI [9], depend on the balance of alveolar recruitment and overdistension.

The regional distribution of the aeration depends on HFNC flows and patients [7,8,9]. When flows were increased from 30 to 60 L/min, Mauri et al [7] reported that EELI increased, but not by a significant amount. Interestingly, when compared to EELI with a facemask, EELI in dependent lung regions significantly increased with HFNC at 60 L/min, while EELI in non-dependent regions remained stable. These findings suggest more recruitment in dependent lung regions than overdistension in non-dependent regions [7]. In a follow-up study that included 12 patients with AHRF [8], the same group of authors compared the effect of different flows that were set based on the patient’s predicted body weight (0.5, 1.0, and 1.5 L/Kg/min). They utilized median flows of 35, 65, and 100 L/min, respectively [8]. Compared to EELI at 0.5 L/Kg/min, EELI in non-dependent lung regions increased at 1.0 L/Kg/min and 1.5 L/Kg/min (p = 0.01), with significance reached at 1.5 L/Kg/min (p < 0.05), while EELI in dependent lung regions remained constant (p = 0.548). Both studies suggest that HFNC flows at 60–65 L/min may cause more recruitment than overdistention, while high flows (such as 100 L/min) may result in lung overdistention, especially in non-dependent lung regions [8]. The large variability between patients in these two studies should be noted, suggesting that personalized flow titration based on its physiological effects may be a pragmatic approach to be used at the bedside. For example, Mauri et al. [7] reported that 37% of patients had improvement in EELI in dependent regions with HFNC at 30 or 45 L/min, but not at 60 L/min. Similarly, Zhang et al. [9] compared EELI at baseline (room air) versus 60 L/min and used the regional recruitment (recruited pixels) to define the potential of lung recruitment, in which recruited pixels > 10% pixels at 60 L/min than at baseline was defined as the high potential of recruitment. They found that 13 in 24 patients (54%) had a high potential for recruitment. For these patients, they noted that recruitment mainly occurred in dependent lung regions when HFNC flow was increased from 0 to 60 L/min [9]. For the rest of the patients included in the study, seven had unchanged EELI and four had overdistension without lung recruitment, occurring mainly in the non-dependent lung regions [9]. The difference in regional volume distribution from various flows in the three studies might be due to the factors that cause different responses to PEEP, including disease severity, etiology, duration of pulmonary disease, and closed- vs open-mouth breathing. Regardless, close monitoring of individuals' responses in regional lung volumes to different flows might help avoid overdistension and lung injury.

Beyond the regional distribution of volume, dynamic transpulmonary pressure reflects the patient inspiratory effort and lung stress, which is associated with lung injury. Changes in esophageal pressure (\(\Delta\)Pes) are a surrogate for inspiratory effort [55]. When HFNC flows were increased, inspiratory effort (Fig. 6A), pressure–time product (Fig. 6B), and WOB (Fig. 6C) decreased [7, 24]. Mauri et al. described an exponential decay correlation between HFNC flows and patient inspiratory effort [7]. The reduction in the patient effort was caused by several factors, such as recruitment of atelectatic regions, an increase in dead space washout, a decrease in nasal resistance, an improvement in secretion clearance, and an increase in dynamic lung compliance [8, 24]. However, they also found that 43% of patients had increased \(\Delta\)Pes when HFNC flows were increased from 30 or 45 to 60 L/min [7]. The patients that demonstrated an increase in \(\Delta\)Pes might have had a compliance decrease due to alveolar overdistention, particularly in the previously relatively well-aerated regions of their lungs with 30 L/min [7]. Thus, due to the concerns that lung injury might occur in patients who have no recruitment with increasing HFNC flows, it has been suggested to titrate flow based on the inspiratory effort [9].

Effects of flow settings on \(\Delta\) Pes (A), PTP (B), and WOB (C). HFNC, high-flow nasal ca