What's new?

The clinical presentation of COVID‑19 varies from asymptomatic infection to severe illness. A significant proportion of patients present with a rapidly progressing acute respiratory failure and require invasive mechanical ventilation. High‑flow nasal oxygen (HFNO) therapy is a technique consisting in delivering heated and humidified oxygen at high flows through a nasal cannula. The HFNO therapy in acute hypoxemic respiratory failure may help avoid intubation and mechanical ventilation. This prospective observational study showed that nearly half of patients with severe COVID‑19 pneumonia eventually required invasive mechanical ventilation and almost a third of these patients died within 30 days from hospital admission. Our results suggest that the ROX index (defined as the ratio of oxygen saturation / fraction of inspired oxygen to respiratory rate) may be a useful tool for stratification of HFNO failure risk in this population.

Introduction

The clinical presentation of COVID‑19 is highly variable and the disease affects multiple organs, with predominant involvement of the respiratory system. Approximately 15% of cases are severe, warranting hospitalization, and 5% of patients require admission to the intensive care unit (ICU).¹ The highly contagious nature of SARS‑CoV‑2 in conjunction with the lack of effective causal treatment led to an exponential increase in the number of cases and a huge burden on healthcare systems globally. Since the first case of COVID‑19 in Poland had been confirmed in March 2020, almost 1.3 million Polish citizens got infected with SARS‑CoV‑2, with a peak monthly incidence of 400 000 new active cases in November 2020. Poland is characterized by one of the lowest ICU‑to–hospital bed ratios in the European Union (2%–5%), rendering the public healthcare system particularly prone to failure in the face of an increased patient load.^2,3

A significant proportion of patients with COVID‑19 present with a rapidly progressing acute respiratory failure (ARF) and eventually require mechanical ventilation. The initially suggested approach included early intubation and mechanical ventilation with lung‑protective strategy recommended by the Acute Respiratory Distress Syndrome Network trial.⁴ Since the mortality of invasively ventilated patients remained high, it was hypothesized that some individuals with severe COVID‑19 pneumonia may benefit from other oxygenation improvement strategies allowing to avoid invasive mechanical ventilation and its adverse effects, that is, ventilator‑induced lung injury and ventilator‑associated pneumonia.^5,6 Hence, high‑flow nasal oxygen (HFNO) therapy was suggested as an optimal treatment modality in this setting.

The HFNO therapy involves delivery of oxygenated gas, heated and humidified to body conditions, via a nasal cannula at a maximum flow up to 80 l/min.^7,8 The therapy is believed to have numerous benefits, including adequate humidification, reduction of anatomical dead space and work of breathing, as well as an increase in end‑expiratory lung volume due to the provision of positive end‑expiratory pressure. During the past years, its use in the critical care setting has extended to treatment of hypoxemic ARF after extubation as well as postoperative treatment of high‑risk or obese patients.⁹ Previous studies, mainly retrospective in design and with a limited sample size, suggested potential benefits associated with the use of HFNO in the treatment of respiratory failure in COVID‑19.^10-13

Facing the growing number of patients with severe respiratory failure due to COVID‑19 along with insufficient ICU resources, our tertiary center adopted the strategy of HFNO administration by experienced respiratory physicians in noncritical care setting. The aim of this study was to assess the therapeutic success of HFNO in severe respiratory failure in the course of COVID‑19. Our secondary goal was to identify factors associated with HFNO failure.

Patients and methods

Results

Patients

The study group initially comprised 380 patients with confirmed COVID‑19, of whom 125 (32.9%) required HFNO therapy. Of those, 9 patients (7.2%) in whom endotracheal intubation was not the goal of care were excluded. Thus, a total of 116 patients were included in the final analysis (Figure 1). No patient was lost to 30‑day follow‑up.

The final study group included 91 men (78.4%) and 25 women (21.6%). The median (IQR) age was 61 (51–70) years. The most common comorbidities were hypertension (57.8%), obesity (37.1%), and diabetes (31%). On admission, the median (IQR) Modified Early Warning Score and SpO₂/FiO₂ ratio were 2 (1–3) and 101.1 (94.7–192.3), respectively. Dexamethasone was administered in 104 patients (89.7%), remdesivir in 71 (61.2%), and convalescent plasma in 17 (14.7%). Detailed demographic and clinical characteristics of the study group are presented in Table 1.

Table 1. Baseline characteristics of the study group

Parameter			Value
Data are presented as median (interquartile range) or number (percentage). a Patients receiving low‑molecular‑weight heparin SI conversion factors: to convert D‑dimer to nmol/l, multiply by 5.476; LDH to μkat/l, by 0.0167; pCO2 and pO2 to kPa, by 0.133. Abbreviations: BMI, body mass index; CRP, C‑reactive protein; HFNO, high‑flow nasal oxygen; IL‑6, interleukin 6; LDH, lactate dehydrogenase; LMWH, low‑molecular‑weight heparin; MEWS, Modified Early Warning Score; pCO2, carbon dioxide; pO2, oxygen; SpO2/FiO2, ratio of blood oxygen saturation to inspired oxygen fraction
Demographic data, measurements, and comorbidities
Age, y			61 (51–70)
Male sex			91 (78.4)
BMI, kg/m2			29.41 (26.08–32.95)
MEWS score on admission			2 (1–3)
Hypertension			67 (57.8)
Chronic heart failure			14 (12.1)
Coronary artery disease			17 (14.7)
Diabetes mellitus			36 (31)
Obesity			43 (37.1)
Chronic kidney disease			12 (10.3)
Obstructive lung disease			13 (11.2)
Laboratory results on admission
D‑dimer, mg/l			1.17 (0.78–3.29)
LDH, U/l			522 (392–620.5)
CRP, mg/l			125.5 (72.4–199.25)
IL‑6, pg/ml			69.19 (30.52–110.33)
Arterial blood gases		pO2, mm Hg	65 (56.5–77.55)
		pCO2, mm Hg	34 (31.1–38.15)
		pH	7.45 (7.42–7.48)
		Lactate, mmol/l	1.4 (1.2–1.9)
Oxygen requirement and respiratory parameters
SpO2/FiO2 on admission			101.05 (94.74–192.25)
Initial FiO2 on HFNO, %			80 (65–90)
Initial respiratory rate on HFNO			22 (20–28)
Max FiO2 on HFNO, %			92 (77.5–95)
Initial flow on HFNO, l/min			60 (50–60)
Max flow on HFNO, l/min			60 (60–60)
Duration of HFNO therapy, d			6 (2–9)
ROX index at 12 h of treatment			5.11 (4.11–6.69)
Pharmacotherapy
Systemic steroids			104 (89.7)
Remdesivir			71 (61.2)
Convalescent plasma			17 (14.7)
LMWHa	Therapeutic dose		68 (58.6)
	Intermediate dose		41 (35.3)
	Prophylactic dose		5 (4.3)

Primary outcome

The primary outcome occurred in 54 patients (46.6%). The Kaplan–Meier curves showing the probability of not developing the primary outcome stratified by the ROX index category are presented in Figure 2. Endotracheal intubation was performed in 51 patients (44%). The 30‑day mortality rates were 30.2% (35 out of 116 patients) for the entire study group and 64.7% (34 out of 51 patients) for those requiring endotracheal intubation.

Univariable analysis showed that patients in whom the primary outcome occurred were older (63 vs 58 years; P = 0.01), had higher baseline levels of lactate dehydrogenase (555 vs 504.5 U/l; P = 0.01), higher lactate levels (1.6 vs 1.3 mmol/l; P = 0.03), and a significantly lower ROX index (4.49 vs 5.47; P <⁠0.001) compared with the rest of the study group (Table 2). We did not find any differences in the duration of hospitalization (18 vs 20 days; P = 0.6). There was also no difference in the median time from initiation of HFNO therapy to endotracheal intubation between survivors and nonsurvivors (1 vs 3 days; P = 0.45). The groups did not differ in terms of administration of systemic steroids (93.5% vs 85.2%; P = 0.24), remdesivir (67.7% vs 53.7%; P = 0.18), and convalescent plasma (16.1% vs 13%; P = 0.83).

Table 2. Comparison of selected clinical characteristics between the subgroups based on the occurrence of the primary outcome

Parameter			Primary outcome not achieved (n = 62)	Primary outcome achieved (n = 54)	P value
Data are presented as median (interquartile range) or number (percentage). For conversion factors to SI units, see Table 1. Abbreviations: see Table 1
Demographic data, measurements, and comorbidities
Age, y			58 (45–66)	63 (57.25–72.75)	0.01
Male sex			11 (17.7)	14 (25.9)	0.4
BMI, kg/m2			28.73 (25.72–32.65)	30.42 (27.22–33.06)	0.23
MEWS score on admission			2 (1–2)	2 (1–3)	0.01
Hypertension			37 (59.7)	30 (55.6)	0.8
Chronic heart failure			9 (14.5)	5 (9.3)	0.56
Coronary artery disease			6 (9.7)	11 (20.4)	0.17
Diabetes mellitus			17 (27.4)	19 (35.2)	0.48
Obesity			21 (33.9)	22 (40.7)	0.57
Chronic kidney disease			4 (6.5)	8 (14.8)	0.24
Obstructive lung disease			7 (11.3)	7 (13)	0.78
Laboratory results on admission
D‑dimer, mg/l			1.11 (0.66–3.85)	1.43 (0.91–2.72)	0.16
LDH, U/l			504.5 (370.75–567.5)	555 (424–815)	0.01
CRP, mg/l			120.5 (74.28–184.25)	143.5 (67.53–211.25)	0.48
IL‑6, pg/ml			51.79 (25.12–108)	80.55 (45.81–117.75)	0.09
Arterial blood gases		pO2, mm Hg	70.2 (59.1–81.75)	63 (54–70)	0.005
		pCO2, mm Hg	34.3 (30.9–38.58)	33.6 (31.2–38)	0.7
		pH	7.46 (7.43–7.49)	7.44 (7.42–7.47)	0.16
		Lactate, mmol/l	1.3 (1.02–1.6)	1.6 (1.28–2.02)	0.03
Oxygen requirement and respiratory parameters
SpO2/FiO2 on admission			103.89 (100–214.77)	98.89 (89.47–161.67)	0.23
Initial FiO2 on HFNO, %			75 (60–85)	86 (70–93)	0.001
Initial respiratory rate on HFNO			22 (20–26)	24 (20–28)	>0.99
Max FiO2 on HFNO, %			85 (70–92)	95 (93–95)	<⁠0.001
Initial flow on HFNO, l/min			60 (50–60)	60 (60–60)	0.01
Max flow on HFNO, l/min			60 (50–60)	60 (60–60)	<⁠0.001
Duration of HFNO therapy, d			7 (5.25–10.75)	2 (2–5.75)	<⁠0.001
ROX index at 12 h of treatment			5.47 (4.64–7.13)	4.49 (3.63–5.91)	0.001
Pharmacotherapy
Systemic steroids			58 (93.5)	46 (85.2)	0.24
Remdesivir			42 (67.7)	29 (53.7)	0.18
Convalescent plasma			10 (16.1)	7 (13.0)	0.83
LMWH	Therapeutic dose		32 (52.5)	36 (67.9)	0.072
	Intermediate dose		28 (45.9)	13 (24.5)
	Prophylactic dose		1 (1.6)	4 (7.5)

Multivariable analysis revealed that after adjustment for age, sex, obesity, baseline D‑dimer levels, and disease stage, the ROX index below 3.85 was associated with increased mortality (hazard ratio [HR], 6.1; 95% CI, 3.04–12.26), whereas the ROX index between 3.85 and 4.88 was not (HR, 1.46; 95% CI, 0.73–2.93), compared with the ROX index of 4.88 or higher (set as a reference value). The results of multivariable analysis are summarized in Figure 3 and Table 3.

Table 3. Results of the multivariable analysis

Variable		HR (95% CI)
Abbreviations: HR, hazard ratio
ROX index	≥4.88	Reference
	≥3.85 and <⁠4.88	1.46 (0.73–2.93)
	<⁠3.85	6.10 (3.04–12.26)
Age		1.02 (0.99–1.04)
Male sex		1.07 (0.55–2.11)
Obesity		1.14 (0.65–2.01)
D‑dimer		1.00 (0.98–1.01)
Late disease stage		0.75 (0.42–1.35)

Discussion

This single‑center prospective observational study revealed that nearly half of patients with severe COVID‑19 pneumonia eventually required invasive mechanical ventilation and almost a third of these patients died within 30 days from admission to hospital. Moreover, our results suggest that the ROX index is a valuable tool for stratification of HFNO failure risk in this population. To our best knowledge, this is the first study of HFNO therapy in severe COVID‑19 in Europe that was conducted in a cohort including more than 100 patients.

High‑flow nasal oxygen therapy is more than just oxygen supplementation. It constitutes a very well‑tolerated and easy‑to‑apply ventilatory assist device. It is widely accepted in the treatment of hypoxemic respiratory failure and its use is recommended by the Surviving Sepsis Campaign over conventional oxygen therapy in patients with severe COVID‑19 with ARF.^15,16 A meta‑analysis of 9 randomized controlled trials of acute hypoxemic respiratory failure in non–COVID‑19 patients requiring HFNO revealed lower intubation rates without any influence on survival. Over half of our patients with baseline high risk for endotracheal intubation eventually avoided the primary outcome, which suggests that HFNO could potentially facilitate the prevention of mechanical ventilation in the COVID‑19 population. However, any final conclusions in this respect are hindered by the lack of a control group. Therefore, randomized controlled trials concerning the efficacy of HFNO in the prevention of intubation among patients with severe COVID‑19 pneumonia are warranted and highly anticipated.

Relatively high intubation and mortality rates observed in our cohort are consistent with the results of other available studies.^17-19 Another key clinical aspect concerning the treatment of ARF in the course of COVID‑19 is the optimal timing of intubation. It has been well proven that the delay of intubation in severe ARF is associated with worse outcomes.²⁰ A similar association in the COVID‑19 population remains unclear. In our study, there was no difference between survivors and nonsurvivors in time from admission to endotracheal intubation. Two multi‑center studies revealed contradictory results. A study by Hyman et al²¹ showed that each additional day between hospital admission and intubation was associated with an increased in‑hospital mortality among patients with severe ARF in the course of COVID‑19, while Dupuis and colleagues revealed that early intubation of patients admitted to ICU due to severe COVID‑19 resulted in increased mortality and number of ICU‑acquired infections.^21,22

The most important practical aim of this study was to evaluate potential factors enabling identification of patients at greater risk of HFNO failure early in the course of treatment. The ROX index is considered a useful clinical tool in the prediction of HFNO success defined as avoiding endotracheal intubation among patients with acute hypoxemic respiratory failure.²³ It was validated in a multi‑center prospective study in patients with non–COVID‑19 pneumonia and hypoxemic respiratory failure. The values of 2.85 or lower after 2 hours, 3.47 or lower after 6 hours, and 3.85 or lower after 12 hours of treatment were predictors of HFNO failure with a specificity of 98% to 99%. Our study suggests that the ROX index is a valuable predictive parameter among patients with severe COVID‑19 pneumonia. We found that the ROX index values lower than 3.85 measured in the first 12 hours after HFNO initiation, as proposed by Roca et al,¹⁴ were associated with a higher rate of intubation and death.

Relatively high intubation and mortality rates observed in our cohort are consistent with the results of other available studies. Compared with a study by Calligaro et al,¹⁰ the proportion of patients who required endotracheal intubation was somewhat lower in our study. Additionally, the 30‑day mortality rate in our cohort was 30% and all survivors were discharged from hospital at the time of data analysis, while in the latter study, the in‑hospital mortality rate was 46%. Another important observation is a very high mortality rate among patients requiring mechanical ventilation after HFNO therapy failure in both studies (64% in our study vs 76% in the study by Calligaro et al¹⁰). Better clinical outcomes observed in our study might be partially explained by a lower severity of ARF in our sample reflected by lower ROX index values in patients who developed the primary outcome (4.49 vs 2.41) as compared with those who did not (5.47 and 3.26). In another retrospective study involving a smaller cohort (n = 62), the authors observed a similar median ROX index of 5.4. Zucman et al¹² reported the need for invasive mechanical ventilation in 63% of the population; however, interestingly, the overall mortality in the ICU was markedly lower and amounted to 17%.

The strength of this study was its prospective design and a relatively large sample collected over a short period of time, with 30‑day follow‑up completed in all participants. Patients with severe COVID‑19 without contraindications received treatment including systemic steroids, low‑molecular‑weight heparin, and remdesivir according to current recommendations; thus, differences in pharmacotherapy were less likely to influence the outcomes. This is the first report on HFNO therapy provided by non‑ICU personnel in a larger cohort.