High-flow nasal oxygen versus noninvasive ventilation in adult patients with cystic fibrosis: a randomized crossover physiological study

Cystic fibrosis (CF) is the most common autosomal recessive disorder in Caucasian populations [1]. Patients with CF experience exacerbations with hypercapnic respiratory failure associated with increased respiratory workload that may require intensive care unit (ICU) admission due to the inability of the respiratory muscles to compensate for an increased demand. Although no international recommendations currently exist [2], these exacerbations are frequently treated with noninvasive ventilation (NIV) [2, 3]. NIV has been shown to unload the respiratory muscles, increase alveolar ventilation and gas exchange [4] and reverse the rapid and shallow breathing pattern commonly adopted by CF patients with advanced lung disease [5].

NIV is a cornerstone therapy for hypercapnic acute respiratory failure [6, 7], but there is also an increasing interest in high-flow nasal therapy (HFNT) as a potential alternative treatment in this indication [8–10]. HFNT is a system delivering actively heated and fully humidified gas mixture with flow rates up to 60 L/min and adjustable FiO₂ from 21 to 100%. The high flow rates generate small amounts of positive end-expiratory pressure (PEEP) that may help counterbalance the effects of intrinsic PEEP (PEEPi) on work of breathing and might act by washing out of the physiological dead space [11]. Furthermore, it could help to facilitate secretion clearance from the humidified gas [12].

Studies have demonstrated the benefits of HFNT in acute hypoxemic respiratory failure [13], after cardiothoracic surgery [14] and in the prevention of postextubation failure among unselected cohorts of critically ill patients during weaning from invasive mechanical ventilation [15, 16]. However, the current evidence of using HFNT in patients with hypercapnic acute respiratory failure is limited [9, 10, 17, 18]. CF is a unique respiratory disease with chronic, progressive, mixed obstructive–restrictive respiratory failure. From a physiological standpoint, this device could help CF patients with exacerbations in particular by improving gas exchange, reducing respiratory workload through reducing dead space, compensating for PEEPi and facilitating mucus clearance [19]. Additionally, a non-interrupted delivery of the technique with simple nasal prongs might improve comfort and tolerability compared to NIV via total face mask.

No study has compared the physiological effects of HFNT and NIV in adult CF patients requiring ventilator support. We hypothesized that HFNT and NIV would similarly reduce the work of breathing and improving breathing pattern in patients requiring ventilator support for CF exacerbations. Our objectives were to compare HFNT- and NIV-induced changes in inspiratory work of breathing assessed noninvasively by the thickening fraction of the diaphragm (TFdi) [20], breathing pattern, CO₂ level, hemodynamics, dyspnea and comfort.

Detailed methods are available in Additional file 1.

Study design

Physiological prospective randomized crossover study (Fig. 1), using noninvasive assessments, was approved by the Ethics Committee of St. Michael’s Hospital (REB #14-338) and registered on clinicaltrial.gov (NCT02262871). Patients provided their written informed consent. The study took place in a respiratory ward where NIV is usually delivered at St. Michael’s Hospital, Toronto, from January 2015 to February 2017.

To our knowledge, this is the first physiological study that specifically investigates the role of HFNT in adult CF patients who may have unique physiological responses compared to other forms of chronic respiratory failure secondary to inspissated secretions, mixed pulmonary mechanics and potentially altered structure and function of the diaphragm. The main result of this study is that diaphragmatic activity per breath, as assessed by TFdi, is similar between HFNT and clinically set NIV session, but that HFNT is associated with reduction in respiratory rate and minute ventilation implying a further reduction in total diaphragmatic work per minute with this technique.

No difference was observed in TFdi between HFNT and NIV, supporting our hypothesis, but we were surprised that there was no measurable significant reduction from either device compared to the baseline TFdi. This finding held true after removing one patient with outlying values, further strengthening our confidence in these results.

We thought it was essential to have all assessments and measurements taken with noninvasive techniques (e.g., external device for volume, ultrasonography for diaphragm activity, transcutaneous gas values) in order to minimize a direct effect of these techniques on the observed values (Heisenberg principle) [29, 30]. Importantly, however, these techniques are subject to measurement and performance error and we took great care to standardize the technical conditions for these measurements The lack of effect of diaphragmatic work might probably be explained by the relatively normalized (approximately 30%) TFdi in these patients at baseline [31, 32] had been stabilized before protocol implementation. Despite having other clinical and biochemical indicators for ventilatory support, our subjects had improved and been stabilized prior to protocol initiation (all subjects weaned to nasal prong oxygen), making it difficult to demonstrate a difference in diaphragmatic work of breathing. This is in contrast to a report of 14 stable chronically hypercapnic COPD patients in which both HFNT and NIV significantly reduced PTPdi (diaphragmatic work) [8]. Furthermore, in a more recent study including 5 hypercapnic patients recovering from acute respiratory failure, esophageal pressure swings, PTP and work of breathing (markers of respiratory effort) were significantly reduced with HFNT at 60 L/min [33]. This difference, however, may also stem from inherent differences in the structure and function of the diaphragm between COPD and CF patients [34].

It has been previously suggested that a concomitant metabolic alkalosis contributes to hypercapnic respiratory failure in exacerbations of adult CF and that these patients less frequently exhibit acidosis than COPD patients due to the abnormal electrolyte transport in CF patients compared to COPD, even during exacerbations [35, 36]. This may have been present in our subjects as well with a median pH in the normal range NIV which has been clinically used for hypercapnic respiratory failure in CF, but little is known about its physiological effects, particularly in the adult population. A Cochrane review on NIV in 191 mixed adult and pediatric CF patients for several indications was limited by small sample sizes of adult-only trials, sparse reporting of physiological variables and heterogeneous inclusion of varying degrees disease severity [21].

Consistent with previous work, HFNT demonstrated a significant reduction in respiratory rate [37]. Despite a statistically significant, but likely clinically inconsequential small increase in Vt, there was an overall substantial decrease in MV, which has been previously reported in COPD patients treated with HFNT [17]. This finding is partly explained by HFNT washing out of anatomical dead space, allowing a larger fraction of the MV to participate in gas exchange, increasing the efficiency of the respiratory system and decreasing the work of breathing [12]. It might also be related to decreases in nasal inspiratory resistance [38], which may lead to reduced PEEPi and inspiratory load [39]. Importantly, the reduction in respiratory rate might turn into a parallel decrease in the energy expenditure of the respiratory muscles, here by approximately 15%. Despite a significant reduction in minute ventilation, we found no concurrent reduction in PtcCO₂, suggesting a reduced dead space and/or metabolic production. The significant reduction in respiratory rate and minute ventilation with HFNT suggests that the work of breathing per minute is substantially reduced, potentially between 15 and 25%, correlating with a reduction in CO₂ production by the respiratory muscles [40].

We found no significant improvement in dyspnea scores. Comfort scores were significantly lower with both devices compared to baseline oxygen. This finding might reflect relatively stabilized patients with high baseline comfort scores. Moreover, this finding may represent the response to a short-term effect and could differ with a more prolonged use, as suggested in the literature [18, 41]. Oddly, 60% of patients were prescribed NIV at home prior to hospitalization, yet had lower comfort scores compared to baseline. Patients were accustomed to this treatment, even if they did not find it comfortable. We suspect this is another sign that patients were not anymore in distress.

This short-term physiological study has limitations. Firstly, epressure support levels during NIV were lower than compared to one NIV study performed in young patients with CF (11–18 cmH₂O) [42]. These levels of pressure were not used clinically in our center, and this could have biased the results toward favoring a difference with HFNT. A recent Cochrane review, however, pointed out the paucity of studies in this situation to indicate optimal settings [21]. In addition, in a recent trial using NIV for physiotherapy, similar levels of pressure support (8 cmH₂O) and PEEP (5 cmH₂O) comparable to our study were used [43]. So, our data represent reasonable, although potentially not optimal, clinical settings. NIV was individually set according to clinical parameters and patient comfort by clinicians, allowing stabilization of the patients. This has probably resulted in normal levels of work of breathing. We are reassured that under NIV settings the TFdi was similar to values found in healthy volunteers [38]. Also, different NIV masks and NIV ventilators (based on resource availability) were used based on patient comfort. It is possible these different masks could perform differently with respect to the amount of leak, for example but the data obtained in other clinical settings do not suggest this has a great influence on work of breathing [44]. It is possible that, because 60% of our patients had used NIV at home, while none had used HFNT before, this could influence the physiological and subjective responses of our subjects.

Recent work suggests that physiological benefits of HFNT used in hypoxemic respiratory failure appear closely related to the applied flow rate [37]. Indeed, our median flow rate applied was 45 (45–55) L/min, and therefore, it is possible that further physiological improvement may have been seen if flow rates were maximized at 60 L/min. The level of flow rate, however, was dictated by patient’s preference. In addition, a recent study showed that most of the effect on inspiratory workload and CO₂ clearance is already obtained at the lowest flow rate [45]. In addition, it remains unclear whether the relationship between flow rate and physiological response holds true in chronically hypercapnic patients.

In contrast to previous physiological studies on HFNT [8, 33, 37] that used esophageal balloon catheters, we chose to utilize diaphragm ultrasonography as our primary means of estimating patient work of breathing as previously demonstrated [24]. Although our approach makes direct comparison with these previous studies more difficult, we feel that diaphragm ultrasound has become a well-studied modality, with described reproducibility, applied noninvasively and relatively easily learned [43]. Previous work has characterized the technical performance of diaphragm ultrasonography with thickness measurements being highly reproducible (mean ± SD 2.4 ± 0.8 mm, repeatability coefficient 0.2 mm, reproducibility coefficient 0.4 mm), while thickening fraction was only moderately reproducible (median 11%, IQR 3–17%, repeatability coefficient 17%, reproducibility coefficient 16%) [43].

Sputum production and tenacity is a hallmark feature of CF; however, we did not assess changes in secretion characteristics during this study. Although the added humidity of HFNT could reduce secretion thickness, this should be an area of future investigation.

An important limitation was that patients were stabilized prior to study inclusion, as already mentioned. Importantly, however, we demonstrated that HFNT significantly improves breathing pattern in patients recovering from CF exacerbations and therefore holds promise that HFNT may confer benefit even in more acute and decompensated presentations.

No difference is observed in HFNT compared to NIV with respect to the diaphragmatic work per breath in CF patients stabilized after a clinical indication for ventilatory support but significantly reduces the respiratory rate and the work per minute. These preliminary data suggest that high-flow therapy may confer physiological benefits by decreasing ventilation needs and may constitute an interesting alternative or supplement to NIV.