This prospective single-centre observational study was performed in the RICU of “Città della Salute e della Scienza di Torino” University Hospital, Turin, Italy, between March 24th, and June 15th, 2020.

All the consecutive patients admitted to the study center were assessed for eligibility. Inclusion criteria were: age ≥ 18 year-old, clinical diagnosis of acute respiratory failure caused by SARS-CoV-2 (confirmed by RT-PCR on rhino-pharyngeal swab), PaO2:FiO2 ratio <250 mmHg during an oxygen supplementation trial of at least one hour with a Venturi mask (FiO2>40%) or non-rebreathing mask, with respiratory distress (respiratory rate (RR) >25 breaths/minute, moderate dyspnoea as measured by the BORG [20]) and evidence of bilateral infiltrates at chest X-ray and/or computed tomography (CT).

Patients were excluded if they met one or more of the following criteria: need for intermittent positive airway pressure ventilation (NIPPV) for hypercapnic respiratory failure (PaCO2>45 mmHg), a former trial of h-CPAP or helmet NIPPV > 24 h in the emergency department or in other settings before RICU admission.

Included patients were treated in CPAP mode via a single-limb intentional leak configuration, by a turbine-driven ventilator (Philips V60 or TROLOGY EVO Respironics Ventilator) provided with a dedicated patch. Detailed description of this configuration has been previously published.18,21 In this configuration, the helmet expiratory port was capped with a connector having a 6 mm internal diameter hole to provide the intentional leak as previously described18 (Fig. 1). A high oxygen pressure inlet allowed the ventilator to guarantee a pre-set stable FiO2. The ventilators’ monitoring system ensured continuous monitoring of the overall leaks (intentional and unintentional), inspired Vt estimation, pressure and flow curves tracings. Helmets (Dimar, medical Device, Medolla (MO) Italy or CaStar StarMed, Intersurgical, Mirandola, Italy) were secured by padded armpit braces and size was chosen according to patient's neck diameter. When available, helmets equipped with an inflatable neck cushion were used to limit the unintentional leak.12 The helmet ports were provided with High-Efficiency Particulate Air (HEPA) filters to limit aerosolization.

.

Fig. 1: Helmet CPAP configuration.

The figure shows the new helmet CPAP configuration allowing tidal volume measurement. From left to right, connector (A), HEPA filter (B), expiratory port (C), helmet (D), inspiratory port (E), HEPA filter (F), single limb circuit (G).

(0.46MB).

Patients included in the study were treated with continuous h-CPAP (24 h/24 h) for at least 48–72 h. When tolerated, prone positioning was kept for at least 2 h three times per day.22 Positive pressure was chosen between 8 and 15 cmH2O and FiO2 set at the lowest possible value to achieve oxygen saturation ≥96%.

Patients on h-CPAP who did not show signs of respiratory distress (e.g. RR <25 or use of accessory muscles) and were able to maintain a SpO2>94% with a FiO2<50% and a PEEP<10 cmH2O underwent a weaning trial in HFNT or Venturi mask. Criteria to consider a patient as successfully weaned from h-CPAP was ability to maintain a PaO2:FiO2 ratio >250 on Venturi mask with a FiO2 <40% for at least 24 h.

A “leak test” (<10 l/min threshold) was periodically performed by closing with a finger (for 5–10 s) the 6 mm hole of the expiratory port. As the ventilator measured the overall leaks, unintentional leaks coming from the helmet collar were evaluated when intentional leaks (coming from the expiratory port) were excluded after expiratory port closure. At the end of each test, overall leaks were recorded. The continuous monitoring of the overall leak (unintentional plus intentional) allowed the early identification of values out of the range (greater than 30% compared to previous values). In these cases, the leak test (occlusion of the intentional leak and check of unintentional leak on the monitor) helped confirm the possible displacement or damage of the helmet. Patients received sedation in case of Richmond Agitation Sedation Score (RASS) ≥ 1.23 A detailed description of healthcare workers protection policy in our centre is provided in Supplementary material B.

Treatment effectiveness was evaluated as:

• - Reliability of Vt measurement as defined as an out of scale Vt (>50%) compared to the mean value recorded on the previous hours;

• - Proportion of patients showing lack in improvement in arterial blood gases due the interface setup malfunctioning (i.e. unintentional leak leading to a lack to maintain a correct ventilatory support)24,25;

• - Unintentional leak as measured by the leak test and dynamically by ventilator monitoring system needed helmet repositioning

Patient's treatment safety was evaluated as:

• - Rate of adverse events due the helmet (i.e. skin breakdown)26–28

• - Rate of suffocation or death due to the interface29

Healthcare workers safety was evaluated as:

• - SARS-CoV-2 infection rate among RICU staff (symptomatic or asymptomatic infection confirmed by RT-PCR test on rhino-pharyngeal swab) during the study period.

Secondary endpoints of the study included:

• - Improvement in arterial blood gases (PaO2, PaCO2, PaO2:FiO2);

• - Need for sedation due the interface intolerance (to reach a target RASS 0/-1)

• - Rate of failure (as defined as need to switch to a facial mask) due the interface intolerance

• - Respiratory rate and dyspnoea (according to Borg scale);

• - Rate of intubation not correlated to the interface setup (detailed criteria for intubation are described in Supplementary material B)

• - Death during RICU stay (not correlated to the interface setup).

• - Other adverse events not correlated to the interface setup (e.g. cardiovascular events, pulmonary embolism, acute kidney injury etc.)

Arterial blood gas analyses were recorded at admission before starting h-CPAP (T0), within 6 h (T6), and on day 1 (T24) after h-CPAP initiation. Vital parameters (systemic blood pressure - SBP, heart rate - HR, respiratory rate - RR, tidal volume – Vt) were monitored continuously and recorded together with the assessment of the dyspnoea score, as measured by the BORG20 scale, three times a day until the start of the weaning, then once a day.

Due to the nature of the study, a convenience sample of consecutive patients was included, and no a priori sample size calculation was done.

Continuous variables were expressed as median values with interquartile range (IQR), and categorical variables were expressed as proportions. Comparisons between groups were performed for continuous variables using the Wilcoxon rank-sum test. Variance analysis (ANOVA) test was used to compare results of repeated measures (T0, T6, T24). Categorical variables were examined using the Fisher's exact test. All statistical analyses were performed using STATA 13.1 (StataCorp, College Station, Tx, USA).

Seventy-one patients were admitted to the study centre during the period of study. A total of thirty-five patients (49%) were eligible and all were included in this study.

Baseline characteristics and blood gas analysis of the included patients at admission are showed in Table 1. The median age was 67 years (IQR 57–76), and 30 patients (85.7%) were men. At admission, patients showed a median PaO2/FiO2 129 (IQR 89–166), signs of respiratory distress (RR median 32, IQR 28–35) with moderate dyspnoea (BORG median 3, IQR 2–4). Details on the co-treatments received by the included patients are provided in the Supplementary material C, Table C.1.

Table 2 summarizes the results for primary and secondary outcomes. Reliability of Vt measurements was 100%. An out of scale of Vt (above 50% compared to the previous values) was never recorded in any patient at any time-point. Fig. C.1 (Supplementary material C) shows the trend of the recorded Vt of each patient during continuously hCPAP and weaning. No patients showed a lack of arterial blood gases improvement due to the configuration set up, with overall leaks reaching the safety threshold value of the turbine driven ventilator. Median value of overall leaks (intentional plus unintentional) was 68 L/min (IQR 63–75). Two patients (5.7%) reported intolerance to the helmet, and they alternated it with the mask without CPAP interruption.

No cases of suffocation or death due to the interface occurred. Six patients (17.1%) needed more than two helmet replacements every three hours, due to leak test >10 l/min. Arm oedema and skin breakdowns were reported in sixteen (45.7%) and seven (20%) patients respectively.

Among the 63 healthcare workers involved in the care of COVID-19 patients during the study period, 3 workers (4.8%) showed positive SARS-CoV-2 IgG. Of these, only one was positive at RT-PCR nasopharyngeal swab testing. All the workers remained asymptomatic during the study period. The study was conducted during the first wave of pandemic in the north of Italy, when the pandemic was out of control. Thus, it cannot be defined if the infection occurred inside or outside the hospital.

Thirteen patients (37.1%) required sedation with morphine and/or dexmedetomidine to improve the tolerability of continuous h-CPAP and reach the target of RASS 0/-1. Adverse events not related to the helmet configuration are reported in Table C.2, Supplementary material C. Table 3 shows respiratory parameters at admission in RICU (T0), after 6 h (T6), and after 24 h (T24) since the start of helmet CPAP. Respiratory parameters at the start and at the end of helmet CPAP weaning are shown in Table C.3, Supplementary material C. Table C.4 and Table C.5 (Supplementary material C) respectively show the patients who died in RICU (DNI) (n = 3) and the respiratory parameters of intubated patients (n = 9). Both patients successfully completing (S) and failing (F) h-CPAP showed an improvement of PaO2/FiO2 ratio, respiratory rate and dyspnoea within the first 6 and 24 h of treatment initiation. PaCO2 increased significantly after 6 h only in patients successfully completing h-CPAP, and in both groups after 24 h. The only difference observed between success and failure was PaO2/FiO2 ratio at 24 h (S: 257 (IQR 181–319) vs F: 171 (IQR 133–225), P = 0.02) (Table C.6, Supplementary material C). Vt, reported as ml/IB, was not significantly different at 6 and 14 h between patients who completed or failed h-CPAP (Table C.6, Supplementary material C).