Introduction

The head helmet is an alternative to the face mask for delivering both non-invasive continuous positive airway pressure (CPAP) [14] and non-invasive ventilation (NIV) [512]. This device is provided with a soft collar that ensures non-traumatic adhesion to the neck of the patient and good tightness, once connected to the ventilation system. In comparison to face masks, the helmet seems really advantageous in terms of patient comfort, tolerance and decreased skin lesions, thus improving feasibility, continuity and duration of non-invasive ventilatory assistance [3, 7, 8, 1012].

On the other hand, some concern exists about carbon dioxide (CO2) rebreathing during helmet application [9, 11, 1316]: with every breath, the CO2 expired by the patient does not completely leave the system, but partly dilutes within the internal volume of the helmet and is subsequently re-inhaled. It was recently demonstrated that, during CPAP delivered by helmet, inspired CO2 is inversely related to the flow of fresh gas “washing” the internal volume of the device [13]. Therefore, it was suggested to use the helmet only with high continuous-flow CPAP systems and to monitor the inspiratory CO2 concentration [13, 14].

Differently from CPAP, during NIV the fresh gas flow applied to the helmet is limited and typically intermittent, rather than continuous: this may worsen the CO2 storage and rebreathing. On the other hand, the helmet responds to positive-pressure ventilation with a relatively high compliance, due to its large volume, soft collar and fixation with armpit straps. In this way only part of the volume delivered by the ventilator reaches the patient respiratory system, while another fraction intermittently distends the helmet. This “helmet ventilation” may improve the washout of CO2 around the head of the patient, possibly limiting CO2 rebreathing.

In a bench study, we analysed the mechanisms of CO2 rebreathing during ventilation delivered by a low size helmet model specifically designed for NIV and provided with inflatable cushions to further decrease the internal volume. First, we looked for the most convenient method to monitor the inspiratory CO2 concentration by sampling CO2 at different sites. Then, we tested the effects on rebreathing of patient’s CO2 production and respiratory mechanics, ventilator setting, inflation of helmet cushions and application of a continuous “flow by” through the helmet. Finally, we verified the hypothesis that the inspiratory CO2 concentration depends on just two factors: the sum of all flows passing through the helmet and the patient’s CO2 production.

Materials and methods

A NIV helmet (CaStar R, Starmed, Mirandola, Italy) was connected to a physical model simulating a passive patient (Fig. 1). The model consisted of an expanded polystyrene head (about 3 l volume) connected to a plate that simulated the shoulders. The head was provided with a proximal airway that was connected, below the plate, to a passive mechanical lung simulator (Lungensimulator LS800; Dräger Medical, Lübeck, Germany) with adjustable respiratory system compliance and airway resistance. A bottle of pure CO2 with a CO2 flowmeter (RSP; Flowmeter, Wilson, OR, USA) was connected to a port in the lung model, for simulation of CO2 production.

Fig. 1
figure 1

Physical lung model and experimental setup. The patient is simulated by an expanded polystyrene head provided with an internal proximal airway, connected to a mechanical lung simulator. The polystyrene head is connected to a plate. A NIV helmet is mounted on the head and fixed to the plate by four rigid straps. a Patient airway opening (site of measurement of eiCO2, etCO2 and V'insp). b 22-mm port for connection to the Y-piece of the ventilator circuit (site of measurement of yCO2, Vt,i and Vt,e). c Additional ports, used for flow-by inflow and outflow. d 22-mm port, used for helmet CO2 (hCO2) measurement. e CO2 inflow for simulation of CO2 production

The helmet was placed around the head and secured to the plate by four straps. When the helmet was pressurized and ventilated, we could obtain a sealed connection between the helmet soft collar and the plate. As in the clinical setting, the helmet was able to slightly move up during inspiration and down during exhalation; moreover, the connection between the collar and the plate was not completely hermetic and leaks were allowed. At end-expiration, the internal volume of the helmet mounted on the head was 9 l, decreasing to 6 l once the cushions were inflated. Pressure-controlled ventilation was delivered by an Engström Carestation ventilator (GE Healthcare, Madison, WI, USA) by connecting the circuit Y-piece to the 22-mm port of the helmet.

Measurements of CO2 concentration, airway pressure and flow were obtained by the ventilator sensors (calibrated according to manufacturer's instructions) and recorded on a personal computer by means of specialized software (RDPL Tool; GE Healthcare).

Series 1

The effects of different levels of CO2 production (V'CO2) on the CO2 concentration sampled at different sites were tested. We increased V'CO2 from 100 to 550 ml/min in 50 ml/min steps. The respiratory pattern and the lung model mechanics were kept constant: PEEP 5 cmH2O, inspiratory pressure above PEEP (Pinsp) 10 cmH2O, respiratory rate (RR) 17/min, I:E ratio 1:2, airway resistance and compliance of lung model 8 cmH20/l/s and 50 ml/cmH20; no flow-by was applied and the cushions were left deflated.

Measurements and calculations

Airway pressure and flow were recorded by connection of the ventilator sensors to the airway opening of the head (Fig. 1 a). After a minimum equilibration time of 15 min and always when stable for at least 5 min, we recorded end-inspiratory (eiCO2) and end-tidal (etCO2) CO2 at the airway opening, and then end-expiratory CO2 at the Y-piece (yCO2) and mean CO2 inside the helmet (hCO2) by sequential connection of the CO2 sensor to three different sampling lines (Fig. 1 a, b, d). The hCO2 was sampled far from both the ventilator connection port and the patient airway opening.

The iCO2,mean was calculated from the instantaneous inspiratory CO2 concentration (CO2,insp) and flow rate (V'insp) sampled at 25 Hz at the airway opening:

$$ \begin{aligned} {\text{iCO}}_{{\text{2,mean}}} ({\%})=& \smallint[{\text{CO}}_{{\text{2,insp}}} (\%)\times {\text{V}}^{\prime}{\text{insp}} (ml/s)\\ &\times 0.04 (s)] /\\ & \smallint [{\text{V}}^{\prime}{\text{insp}} (ml/s)\times0.04 (s)] \end{aligned} $$

Series 2

During constant V'CO2 (300 ml/min) the effects on CO2 rebreathing of inspiratory pressure, patient passive mechanics, flow-by and inflation of helmet cushions were tested. We simulated a normal and a low compliance of 50 and 30 ml/cmH2O. The cushion inflation was performed with 3 l of air.

For the flow-by, we used the helmet ports normally dedicated to nasogastric tubes (Fig. 1 c). The flow-by inflow and outflow were adjusted to the same value by two flowmeters (Dräger Medical, Lübeck, Germany) and obtained by delivering a constant flow of air from a source independent from the ventilator and by connection to a vacuum source, respectively.

Four levels (5, 10, 15 and 20 cmH2O) of Pinsp and two values of flow-by (zero and 8 l/min) were used in any of three different settings: normal compliance, deflated cushions; low compliance, deflated cushions; and normal compliance, inflated cushions. Therefore, a total of 24 different conditions were studied. PEEP, RR and I:E ratio were kept constant as in series 1.

Measurements and calculations

We recorded inspiratory and expiratory volumes delivered by the ventilator (Vt,i and Vt,e) by connection of the ventilator sensors to the Y-piece (Fig. 1 b). End-tidal and end-inspiratory CO2 at the airway opening and helmet CO2 were recorded after a minimum equilibration time of 15 min and always when stable for at least 5 min by sequential connection of the CO2 sensor to two different sampling lines (Fig. 1 a, d). The patient tidal volume (Vt,pat) was read directly on the lung simulator, on a scale precalibrated with a super-syringe.

The minute leakage (MVleak) and the minute ventilation of the patient (MVpatient), of the helmet (MVhelmet) and of the whole system (MVtotal) were calculated as follows:

$$ {\text{MV}}_{{\text{leak}}}=({\text{V}}_{\text{t,i}}-{\text{V}}_{\text{t,e}})\times {\text{RR}} $$
$$ {\text{MV}}_{{\text{patient}}}={\text{V}}_{\text{t,pat}}\times {\text{RR}} $$
$$ {\text{MV}}_{{\text{helmet}}}=({\text{V}}_{\text{t,e}}-{\text{V}}_{\text{t,pat}}){\text{RR}} $$
$$ {\text{MV}}_{{\text{total}}}={\text{MV}}_{{\text{helmet}}}+{\text{MV}}_{{\text{patient}}}+{\text{MV}}_{{\text{leak}}}+{\text{flow-by}} $$

Paired t-test was used to compare conditions differing for just one parameter: Pinsp 5 versus 10, 15 and 20 cmH2O, normal compliance versus low compliance (by discarding the eight conditions with the cushions inflated), deflated cushions versus inflated cushions (by discarding the eight conditions with the compliance low), no flow-by versus 8 l/min flow-by.

Calculated value of mean helmet CO2 (hCO2,calc) were obtained by application of the equation hCO2,calc = V'CO2 / MVtotal and compared with measured values in series 1 and 2.

Results

Monitoring of CO2 rebreathing

In the 10 conditions of series 1, the mean inspired CO2 (iCO2,mean) was 2.2 ± 0.8% (range 1.0–3.3).

Figure 2 represents the instantaneous CO2 concentrations (10-s record) at the airway opening of the patient, at the Y-piece of the ventilator circuit and inside the helmet. The CO2 wave at the airway opening oscillated between an eiCO2 level of 0.8 ± 0.3% (range 0.4–1.2) and an etCO2 level of 3.9 ± 1.4% (range 1.7–5.9). The CO2 wave at the Y-piece oscillated between zero during inspiration and an end-expiratory value (yCO2) of 2.6 ± 1.0% (range 1.1–4.1). The CO2 concentration inside the helmet (hCO2) was extremely stable during the respiratory cycle, with a value of 2.1 ± 0.8% (range 0.9–3.3).

Fig. 2
figure 2

Continuous CO2 concentration recordings at the airway opening (black line), at the Y-piece of the ventilator circuit (dotted line) and inside the helmet (grey line) during helmet ventilation, with 16.5 l/min of total minute ventilation and 350 ml/min of CO2 production. At the airway opening the minimum value is the end-inspiratory CO2 (eiCO 2) and the maximum value is the end-expiratory CO2 (etCO 2). The CO2 concentration measured at the Y-piece ranges from zero during inspiration (due to the fresh gas flow from the ventilator) and the end-expiratory value (yCO 2), which is lower than etCO2 because the gas expired by the patient is diluted in the helmet internal volume. The CO2 concentration measured in a quiet (not crossed by flows) point inside the helmet (hCO 2) is constant during the respiratory cycle and corresponds to the mean inhaled CO2 concentration (see also Fig. 3)

Fig. 3
figure 3

Relationships between mean inspired CO2 (iCO 2 ,mean) and, respectively, end-inspiratory CO2 at airway opening (eiCO 2 , rhombs), helmet CO2 (hCO 2 , circles) and end-tidal CO2 at the Y-piece (yCO 2 , triangles). The identity line is traced as a dotted line

Figure 3 shows the relationships between iCO2,mean and, respectively, eiCO2, yCO2 and hCO2. All these three CO2 measurements were highly and linearly related with iCO2,mean (r 2 > 0.97). Differences between iCO2,mean and, respectively, hCO2, yCO2 and eiCO2 were 0.0 ± 0.1% (range −0.2 to +0.1), 0.4 ± 0.2% (0.1–0.8) and −1.3 ± 0.5% (−2.1 to −0.8).

Determinants of CO2 rebreathing

In series 1, while MVtotal was constant at 16.5 l/min, iCO2,mean was linearly and highly related to V'CO2 (iCO2,mean = V'CO2 / 169.5+0.19; r 2 = 0.98).

The results obtained in series 2 are summarized in Table 1. As in series 1, eiCO2 greatly underestimated mean helmet CO2.

Table 1 Changes of mean CO2 concentration inside the helmet (hCO 2), end-inspiratory and end-expiratory CO2 at the airway opening (eiCO 2, etCO 2) and minute ventilations (MV) due to different levels of inspiratory pressure (Pinsp) of 5, 10, 15 or 20 cmH2O, respiratory system compliance (Cpl) of 50 or 30 ml/cmH2O, deflated or inflated cushions and flow-by (FB) of 0 or 8l/min during helmet ventilation

Increasing levels of Pinsp from 5 to 20 cmH2O were associated with significant increases of MVpatient, MVhelmet and MVleak, resulting in progressively increasing values of MVtotal and decreasing values of both hCO2 and etCO2.

The selection of a low Cpl on the lung simulator led to a major decrease in MVpatient, associated with an increase in MVhelmet: this halved the patient:helmet ventilation ratio. The low Cpl setting was associated with a slight decrease in MVtotal, a slight increase in hCO2 and a large increase in etCO2.

The 33% helmet volume reduction obtained by cushion inflation was associated with a minimal decrease in MVpatient, a minimal increase in MVhelmet and no significant changes in MVtotal, hCO2 and etCO2.

The addition of a 8 l/min flow-by through the helmet went along with an equal increase of MVtotal and effectively decreased hCO2 and etCO2, despite unchanged patient's ventilation.

During helmet ventilation at constant CO2 production, we observed a very good inverse relationship between hCO2 and MVtotal (Fig. 4). The equation hCO2 = 31.8 / MVtotal–0.09 assured the best fit with the experimental data (r 2 = 0.98).

Fig. 4
figure 4

Relationship between mean CO2 inside the helmet (hCO 2) and whole flow passing through the helmet (MV total ). Data obtained during the 24 conditions of series 2 (circles) and theoretical hCO2–MVtotal curves (continuous lines) according to the equation of CO2 steady state inside the helmet at five different values of CO2 production (100, 200, 300, 400 and 500 ml/min)

Conversely, the relationship between hCO2 and MVhelmet was poor (r 2 = 0.41).

The calculated values of mean helmet CO2 were nearly identical to the measured values in series 1 and 2 (Fig. 5): mean difference was −0.1 ± 0.1%, and the equation hCO2,calc = hCO2 × 1.0–0.1 assured the best fit with the experimental data (r 2 = 0.98).

Fig. 5
figure 5

Difference between calculated (hCO 2,calc ) and measured (hCO 2) values of mean helmet CO2: Bland–Altman analysis. The calculated values were obtained by application of the equation hCO2,calc = V'CO2/MVtotal in the 10 conditions of series 1 and in the 24 conditions of series 2. Continuous line and dotted lines: mean value (−0.1%) and ± 2 standard deviations (± 0.2%) of the difference between hCO2,calc and hCO2

Discussion

The main findings of this bench study are that, during helmet ventilation: (1) CO2 rebreathing can be monitored measuring the CO2 concentration at a “quiet” point inside the helmet or alternatively at the Y-piece, and (2) CO2 rebreathing is inversely related to the total flow passing through the helmet and directly related to the patient's CO2 production.

In our experience we observed that CO2 concentration is not homogeneous inside the helmet. When the sampling was performed between patient and ventilator, at airway opening or Y-piece, CO2 oscillated between low inspiratory and high expiratory levels. In contrast, CO2 was very stable when measured at a quiet point, not affected by flows to and from the patient; moreover, this value was really equivalent to the mean inhaled CO2. Therefore, when a “quiet” point inside the helmet can be found, this seems the best site for CO2 sampling.

An interesting alternative, even if slightly less accurate, is represented by yCO2, i.e. by reading the etCO2 at the Y-piece. Of note, the value of etCO2 at the Y-piece is much lower than the actual etCO2 of the patient, because the expired CO2 dilutes within the helmet internal volume, which is much greater than the patient's expiratory volume. Therefore, yCO2 can be used to estimate CO2 rebreathing but not to estimate the patient's arterial CO2.

Usually CO2 rebreathing had been evaluated by means of the end-inspiratory value at the airway opening [1, 13, 15, 16], but in our experience eiCO2 grossly underestimated mean inspiratory CO2, particularly when it was significant. The reason was the slow decrease in inhaled CO2 concentration during the inspiratory phase of helmet ventilation.

When a steady state is reached and CO2 is stable inside the helmet, the amount of CO2 entering the helmet per minute must be equal to the amount of CO2 leaving the device in the same time. Assuming a non-significant dead space at the Y-piece of the ventilator circuit, the only source of CO2 entering the helmet is CO2 production by the patient. Concerning the CO2 leaving the system, that is a function of the mean CO2 inside the helmet and the sum of all flows directed from inside the helmet to outside: namely the expired (by patient and helmet) ventilation through the Y-piece, plus the air leaks at the collar–neck interface, plus, if present, the outflow of an additional flow-by. Accordingly, the theoretical equation of CO2 steady state inside the helmet is:

$$ \begin{aligned} {\text{V}}^{\prime}{\text{CO}}_2 &={\text{hCO}}_2\times {\text{MV}}_{\text{total}}, {\text{or }}\\ {\text{ hCO}}_2 &={\text{V}}^{\prime}{\text{CO}}_2 / {\text{MV}}_{\text{total}}. \end{aligned} $$

Therefore, the mean CO2 concentration inside the helmet should be directly related to V'CO2, inversely related to MVtotal and unaffected by any other factor. This hypothesis was fully confirmed by our experimental data (Fig. 4 and 5): of note, during phasic helmet ventilation we obtained results similar to those observed by Taccone et al. during continuous-flow CPAP [13].

As a matter of fact, all the manipulations tested in our second series affected CO2 rebreathing insofar as they were able to change the total flow passing through the helmet.

The increase of inspiratory pressure above PEEP produced a progressive increase in MVtotal; accordingly, hCO2 progressively decreased. In contrast with our findings, studies on healthy volunteers recently found no relationship between pressure support level and CO2 rebreathing [15, 16]. A possible explanation is that, in order to keep their minute ventilation constant, volunteers may react to an increase in pressure support with a decrease in spontaneous inspiratory activity, thus limiting the effect on MVtotal. Most importantly, the theoretical relationship between MVtotal and hCO2 (Fig. 4) tends to flatten for high values of MVtotal, and the flattening takes place for lower values of MVtotal when V'CO2 is lower. In the study by Costa and coworkers [15], the mean MVtotal was close to the top of our second series, while V'CO2 was much lower. Therefore, in these conditions the expected variations in helmet CO2 due to changes in MVtotal were really low. Finally, rebreathing was assessed by the end-inspiratory CO2 at the airway opening in these studies [15, 16], i.e. by a parameter we have found to have major limitations.

The relative distribution of MVtotal between patient, helmet and air leaks does not affect rebreathing. In confirmation of that, when we selected a low compliance in the lung model, this resulted in a large change in the patient:helmet ventilation ratio (from 2:1 to about 1:1) with only a small change in hCO2, the latter fully explained by a proportionate change in MVtotal.

With regard to leaks, these can decrease CO2 inside the helmet by an increase in MVtotal. In any case, a helmet provided with an “intentional” leakage port will decrease helmet pressurization and patient's inspiratory assistance, probably turning out counterproductive for CO2 removal.

Concerning the role of helmet volume, inflation of the cushions resulted in a 33% volume reduction but in our experience was associated with no change in hCO2. In clinical practice, the inflated cushions may facilitate the initial pressurization of the helmet, improve the comfort of the patient and stabilize the system, eventually limiting air leaks: the overall effect on rebreathing is difficult to predict, but probably of low significance.

The addition of a flow-by through the helmet was very effective in clearing CO2 around the head of the patient. However, the flow-by system we used in our experimental setting is not very practical for clinical application. An interesting option might be the use of the bias flow of the ventilator as a flow-by. In order to force the bias flow to pass through the helmet, the ventilator must be connected to the helmet by two independent ports, for inspiration and expiration. Moreover, for a significant effect on CO2 rebreathing the bias flow should be continuous or at least applied during the entire expiratory phase, and it should be adjustable at relatively high values, at least 10 l/min.

This study has some limitations. We performed a bench study, because the pure effects of changes of CO2 production or minute ventilation on rebreathing would have been difficult to study in patients or volunteers. The lung model was passive and the ventilation controlled, in contrast to clinical practice. The use of an active physical model would have offered the opportunity, through modifications of the spontaneous activity, to change both the patient ventilation and the patient:helmet ventilation ratio. In our experimental setting, we obtained a similar effect by changing the lung model compliance. Finally, just one type of helmet was studied. Helmet models can differ in elastic properties, tightness or inner volume: these factors are related to rebreathing through their influence on helmet ventilation and leaks or, in the case of volume, not related at all.

Conclusions

During NIV delivered by helmet, some CO2 rebreathing is necessarily present and the inspiratory CO2 must be monitored: gas sampling can be performed at a “quiet” point inside the helmet or, alternatively, at the Y-piece. The magnitude of rebreathing is directly related to the patient's CO2 production and inversely related to the total flow passing through the helmet. Therefore, significant rebreathing can be expected when the patient's metabolic requirements are elevated and/or the ventilator volume delivery is low, discouraging the use of low levels of pressure support in this setting. To further decrease rebreathing when the pressure assistance to the patient's spontaneous activity is already optimal, the addiction of a flow-by can be considered.