Introduction

Respiratory physiotherapy and suction are regularly performed in mechanically ventilated children to prevent and resolve respiratory complications. The techniques frequently used by physiotherapists for ventilated children include postural drainage, manual hyperinflation, chest wall percussion and vibration, saline instillation and tracheal suction. Routine suction does not usually include manual techniques and is usually performed by nursing staff.

Few researchers have compared physiological responses to physiotherapy or suction and none have reported changes in respiratory deadspace, although a major cause of impaired gas exchange in ventilated patients is atelectasis, causing intrapulmonary shunt. One case study reported that the alveolar deadspace (VDalv) was significantly reduced after resolution of a partially atelectatic lung [1]. Other studies indirectly attributed improvements in hypoxaemia and compliance following physiotherapy to improved matching between alveolar ventilation and pulmonary perfusion (V/Q balance) and reduced intrapulmonary shunt [2, 3]. Measurements of respiratory deadspace have shown potential clinical benefit in diagnosis, prognosis, estimation of cardiac output and efficiency of pulmonary perfusion in some adult and paediatric patients [4, 5, 6].

However, deadspace volumes are difficult to identify anatomically and to measure directly. Methods to measure it are generally based on functional rather than morphological definitions, and theoretical models for calculating deadspace volumes from expired carbon dioxide (CO2) analysis have been proposed since 1928 [7].

The principal aims of this study were to assess the effects of physiotherapy treatments and suction on deadspace volumes, carbon dioxide elimination (VCO2), end tidal CO2 (ETCO2) and arterial partial pressure of carbon dioxide (PaCO2) in fully ventilated infants and children and compare the differences between these treatments. The effects of physiotherapy and suction on respiratory mechanics, arterial blood gases and tidal volume (VT) in the same population of children have been described in the accompanying manuscript [8].

Materials and methods

Equipment

A portable respiratory monitor (“CO2SMO Plus”, Novametrix Medical Systems, USA) was used to measure respiratory volumes and expired CO2 concentration before and after physiotherapy and suction in the same ventilated children. The monitor measured pressure, flow and CO2 concentration continuously and instantaneously via a disposable, fixed-orifice, differential flow sensor with incorporated mainstream infrared absorption capnograph. The sensor was inserted at the airway opening between the tracheal tube and the ventilator circuit. Volumes were integrated from flow and the calculations were based on expired tidal volumes (VTE) to minimise any effects of tracheal tube leak [9, 10].

In vitro assessment of CO2 measurement accuracy revealed a small but systematic underestimation of gas concentration with negligible between-sensor differences (measurement error <4% and coefficient of variation <2%). In vitro and in vivo assessments of deadspace measurements by the Ventrak 1550/Capnogard 1265 (now the “CO2SMO Plus”) suggested that this was a reliable and accurate alternative to other methods of calculating deadspace [11, 12].

A neonatal sensor (deadspace: 0.8 ml), was used in children under 2 years, whereas a paediatric sensor (deadspace: <4 ml) was used in older children. Apparatus deadspace was less than 1 ml/kg in all subjects and the combined apparatus resistance was less than 20% of the patient’s intrinsic resistance at the mean flows likely to be encountered [13].

The clinical and practical validity of the “CO2SMO Plus” in measuring VTE, respiratory system compliance (Crs) and respiratory system resistance (Rrs) has previously been established in this population provided that tracheal tube leak is less than 20% [9, 10, 14, 15]. Further details on the accuracy of volume measurements is described in the electronic supplementary material (S1). Ventilator settings were held constant during the measurements and the tubing was cleared of condensed water prior to data collection.

Protocol

Data were collected in the intensive care units at Great Ormond Street Hospital for Children NHS Trust, London between April, 1998, and March, 2000. Infants and children were eligible for recruitment if they were pharmacologically paralysed, mechanically ventilated, had an arterial line in situ and were deemed by the physiotherapist to require physiotherapy. The local research ethics committee approved the study and written informed consent was obtained from parents.

This randomised crossover study involved assigning patients to receive either physiotherapy or suction in the morning and the alternative intervention in the afternoon with at least 90 min between treatments in individual patients. The order of treatment was randomised as soon as possible after recruitment by witnessed coin toss at the bedside. Arterial bloods were sampled immediately before and 30 min after each treatment using the Hewlett Packard i-Stat portable clinical analyser (i-STAT Corporation, New Jersey) [16].

The “CO2SMO Plus” respiratory monitor calculated physiological deadspace (VDphys)/kg, VDalv/kg and VDphys/VT from simultaneous PaCO2 blood gas analysis. In addition, the monitor was used to measure VTE/kg, CO2 elimination (VCO2/kg), ETCO2, mixed expired CO2 (PeCO2) and anatomical deadspace (VDana)/kg continuously for 15 min before each treatment and 30 min after each treatment. Continuous data were sampled at 100 Hz and averaged over each minute, with cumulative mean values summarised immediately before and 30 min after treatment in 15-min epochs for comparison during analysis.

Patients were muscle relaxed, sedated and had analgesia, according to standard infusion protocols, of vecuronium, midazolam and morphine, respectively. The changes in parameters were thus likely to be related to treatment rather than changes in respiratory effort. Ventilation modality, age and weight were recorded for consideration during sub-group analysis.

The staff performed the physiotherapy and suction procedures considered appropriate for the individual patients. Both procedures involved any combination of pre-oxygenation, saline instillation, manual hyperinflation and suction, with as many or few cycles as necessary. Physiotherapy procedures were distinguishable from suction by the addition of chest wall vibrations, percussion or compression and postural drainage.

Theoretical background

Physiological deadspace (VDphys) encompasses the ventilated areas of lung and upper respiratory tract which do not participate in gas exchange. It comprises the sum of alveolar (VDalv) and anatomical deadspace (VDana). The “CO2SMO Plus” included ventilator apparatus deadspace (between the patient and the flow sensor) as an extension of anatomical deadspace. Apparatus deadspace was not measured or subtracted from VDana since it remained constant within each patient.

Figure 1 illustrates the CO2 concentration in the expired breath plotted against expired volume. The “CO2SMO Plus” used a functional estimation of VDana (or “Fowler’s deadspace”) by defining it as the volume of conducting airways at the midpoint of the CO2 concentration transition between conducting airways and alveolar gas. The extrapolated phase III slope determined the point at which volumes of CO2 (represented by areas a and b in Fig. 1) were equal.

Fig. 1
figure 1

The single breath carbon dioxide (CO2) curve. Phase I is the expired volume of the proximal conducting airways containing negligible amounts of CO2. Phase II is the volume of the transitional region between alveolar gas and the conducting airways, characterised by a sharp increase in CO2 concentration as gas from the alveoli mixes with gas from the conducting airways. Phase III primarily contains gas from alveoli and provides most of the expired CO2 volume. It is usually characterised by a gently increasing slope

The “CO2SMO Plus” used the Enghoff modification of the Bohr equation to calculate VDphys/VTE:

$${\text{VD}}_{{{\text{phys}}}} /{\text{V}}_{{{\text{TE}}}} = {\left( {{\text{PaCO}}_{{\text{2}}} - {\text{PeCO}}_{{\text{2}}} } \right)}/{\text{PaCO}}_{{\text{2}}} $$

VDphys/VTE was then multiplied by VTE to obtain VDphys. PeCO2 was calculated from the concentration of CO2 in the expired breath. PaCO2 was obtained from simultaneous arterial blood gas analysis. This method compares favourably with the traditional Douglas bag method [17].

Since VDphys refers to the sum of VDalv and VDana, VDalv was calculated by subtracting VDana volume (Fowler’s method) from VDphys (Bohr equation). VDalv, or “wasted ventilation”, is the volume of gas that enters the functional gas exchange units but does not participate in gas exchange, either because the alveoli are relatively over-ventilated or under-perfused.

The volume of CO2 eliminated was the net volume of exhaled CO2 measured at the tracheal tube over each minute, divided by the weight of the infant and expressed in ml/min per kg (VCO2). Mixed expired CO2 (PeCO2), expressed in kPa, was the partial pressure of CO2 in expired gas. The peak concentration of CO2 averaged over eight expired breaths was reported as the ETCO2.

Statistical analysis

Paired t-tests with 95% confidence intervals (CI) of the difference were performed to assess the effects of respiratory physiotherapy and suction on deadspace parameters and to compare group differences [18]. In addition, the proportion of individuals in whom changes in deadspace parameters exceeded 10 and 20% following either treatment were examined and compared. There were no data available from previous studies to illustrate the extent of normal variability in VDphys and VDalv in paralysed, ventilated children in the absence of any intervention. In this study of ventilated, paralysed children, changes in excess of 10% were considered to have potentially important clinical consequences.

Results

Study population

Of the 100 fully ventilated infants and children recruited to the study and described in the accompanying manuscript [8], 87 had arterial lines in situ and were therefore eligible. Paired measurements of both treatments under similar measurement conditions were only obtained in 81 patients because of changes in ventilation or clinical circumstances between treatments and six further patients had to be excluded from data analysis because tracheal tube leak exceeded 20% during the measurement period. Forty-three of the remaining 75 children had a primary cardiac diagnosis (49% male: median age 8 weeks, range 3 days–16 years, median weight: 4 kg, range 2–37 kg) and 32 a primary respiratory diagnosis (65% male, median age: 22 months, range 3 days–16 years, median weight: 10 kg, range 3–87 kg). Pressure pre-set ventilation modes were used in 56 children, while 19 received volume pre-set modalities.

Infants and children with a primary cardiac diagnosis included those who had undergone surgery for congenital cardiac defects such as transposition of the great arteries, truncus arteriosus and septal defects, but also those with pulmonary hypertension and those who had undergone cardiac transplantation. Six infants, all in the cardiac population, were receiving nitric oxide (NO) therapy for suspected pulmonary hypertension. In these patients, physiotherapy and suction treatments were, as per normal practice, modified to ensure continued delivery of NO by manual inflation via the ventilation circuit.

Patients with primary or secondary respiratory problems included those with respiratory failure, head injury, abdominal surgery, bone marrow transplantation, asthma, gastric transposition surgery, inhalation injuries and tracheal reconstruction. Both groups included medical and surgical patients.

There was no difference in baseline PaCO2, PeCO2, VDphys or VDalv between medical and surgical groups. However, baseline PaCO2 and PeCO2 values in the younger cardiac patients were lower than in the respiratory group (4.85 vs 5.85 and 2.7 vs 3.8, respectively). As a consequence, the median weight-corrected VDphys and VDalv were higher in the cardiac patients than the respiratory group (VDphys: 3.9 ml/kg, range 0.7–7.6 versus 2.3 ml/kg, range 1.0–4.6 and VDalv: 2.0 ml/kg, range: 0.2–4.3 versus 1.3 ml/kg, range: 0.3–3.7) (Fig. 2).

Fig. 2
figure 2

Baseline values of physiological deadspace (VD phys ) marked by cardiac or respiratory diagnosis

The median baseline oxygenation index (OI) for the group was 4.4 with a range between 2.2 and 36.4. The median OI for patients with cardiac diagnoses was marginally higher than that in the older respiratory group [4.5 (range: 2.3–36.4) versus 3.7 (2.2–10.5)]. Mean tracheal tube leak, assessed on a breath by breath basis, was less than 10% in both groups throughout the measurement period with no significant change in leak as a result of intervention in either group. The median change in percent of leak following treatment was 0.8% (−10.2 to 8%).

Differences between treatments

As described in the accompanying manuscript [8], the physiotherapy treatments were significantly longer, involved more saline instillation and required more suction catheters per treatment (p<0.005).

Effects of physiotherapy and suction

There were significant increases in VDphys/kg (p<0.0005), VDalv/kg (p<0.0001) and VDphys/VT (p<0.05) following physiotherapy (Tables 1 and 2). There was also a tendency for VTE to increase (p=0.08) and for PeCO2 to decrease (p=0.06) following physiotherapy (Table 1). There were no changes in VDana, VCO2 or PaCO2 following either intervention (Tables 1 and 2).

Table 1 Effect of physiotherapy on deadspace and blood gas parameters
Table 2 Effect of suction on deadspace and blood gas parameters

Comparison of treatments

The significant changes in mean within-subject values following physiotherapy and suction and the differences between treatments are summarised in Table 3. There were significant differences between treatments for VDphys/kg and VDalv/kg (p≤0.005). In addition, there were significant differences in VTE, PeCO2 and ETCO2 (p<0.05) because of the difference in direction of change following physiotherapy and suction. There were no significant differences between treatments with respect to VDairway, VDphys/VT, VCO2, pH, PaO2 or PaCO2.

Table 3 Significant differences between physiotherapy and suction (paired data)

Within-subject changes

There was considerable individual variation in response to both physiotherapy and suction with respect to expired CO2 and deadspace parameters. The proportions of individuals demonstrating changes in VDphys in excess of 10% and 20% in response to either treatment are shown in Fig. 3. VDphys and VDalv increased by more than 10% in 36/75 (48%) patients undergoing physiotherapy compared with only 18/75 (24%) of those following suction (p>0.002).

Fig. 3
figure 3

Relative changes in physiological deadspace (VD phys ) following physiotherapy and suction

There was no relationship between OI and adverse events or response to treatment for either physiotherapy or suction. Despite the differences in baseline VDphys and VDalv, both cardiac and respiratory groups independently showed significant increases in VDphys and VDalv in response to physiotherapy while neither responded to suction alone.

Discussion

The significant increases in VDphys, VDalv and VDphys/VT following physiotherapy and significant differences between treatments with respect to VDphys, VDalv, VTE, PeCO2 and ETCO2 were likely to be related to some or all of the physical differences between physiotherapy and suction delivery including factors such as treatment duration, saline instillation, suction catheters used and the additional performance of chest wall vibrations.

Factors influencing deadspace volumes

Deadspace is proportional to body weight, height and body surface area [19] and affected by body and neck position, alveolar volume at the end of expiration, VTE, intubation status, ventilation modality and length of tracheal tube [20]. Since most of these variables remained constant, changes were likely to reflect changes in alveolar ventilation, VTE or pulmonary perfusion. Increases in VDalv reflect the degree to which alveolar ventilation and perfusion fail to match each other, regionally or globally in the lung [21]. This may be caused by any condition that abnormally elevates the V/Q ratio such as pulmonary hypotension, pulmonary embolus, hyperventilation, obstruction of pulmonary arterioles and alveolar hyperinflation. In children both gross hypo- and hyper-perfusion cause an increase in VDalv [22].

The modified Bohr equation requires PaCO2, PeCO2 and VTE and VDairway to derive VDphys. These parameters fluctuate in a ventilated child and VDphys is thus dependent on the relative magnitude of each of these variables at any specific time. The tendency for VTE to increase and PeCO2 to decrease following physiotherapy resulted in an amplification of VDphys and VDalv (since group changes in PaCO2 and VDairway were negligible).

An increase in VDphys or VDalv appears at face value to be a cause for concern. However, since VDalv is derived from four dynamically fluctuating variables, changes may reflect a transient pulmonary state, before a new equilibrium between ventilation and perfusion is reached. One case report indicated a reduction in VDalv after manual hyperventilation and tracheal suction in a child with acute lobar atelectasis [1], perhaps reflecting a rapid reduction in the PaCO2/PeCO2 gradient following resolution of atelectasis, and a relatively smaller increase in VTE. If this child had presented with a chronic lobar atelectasis, in which there was substantial hypoxic pulmonary vasoconstriction (HPV), a rapid recruitment of alveoli may not immediately have been matched by reperfusion. In this example, the increase in VT would, at least temporarily, have been reflected as “wasted ventilation” with increased VDalv.

The time taken to restore V/Q equilibrium following the recovery of atelectasis is variable and may depend on the resolution of HPV. Research suggests that there is a time continuum in which HPV occurs and in which it is relieved [23]. Resolution depends on the proportion of lung involved, age, gravity and posture, period of localised atelectasis [24], mixed venous oxygen saturation [25], vascular distending pressure [26] and presence of NO or other pharmacological factors influencing the reactivity of pulmonary vasculature [27]. Disconnection of NO during treatment might result in pulmonary vasoconstriction and increased deadspace. However, routine modification of physiotherapy and suction ensured this did not occur. Finally, pulmonary microvascular networks are extremely complex in nature and the reactivity to hypoxia is perceptibly different in different lung regions [28].

Within the clinical environment, therefore, it is impractical to predict the length of time required for pulmonary reperfusion following the resolution of atelectatic areas. It is possible that sampling arterial blood gases only 30 min after treatment would not have allowed sufficient time for V/Q equilibrium. By contrast, the tendencies for VTE to increase and PeCO2 to decrease following physiotherapy were likely to be immediate and may have indicated some recruitment of atelectatic areas. In the absence of a concurrent increase in pulmonary perfusion, VDphys would have increased as a result of the increased gradient between PeCO2 and PaCO2 [22].

In this study, there were no group changes in the “gold standards” VCO2 or PaCO2 after either treatment, indicating that pulmonary homeostasis was preserved despite the change in V/Q balance suggested by the increase in VDphys and VDalv. A compromise in CO2 elimination would be reflected by a rise in PaCO2, which was not observed. This may illustrate the relative insensitivity of these standard measures in identifying subtle changes in regional ventilation and V/Q balance. VDalv and VDphys could offer a more sensitive measure of subtle changes in respiratory physiology following interventions such as physiotherapy, but have not commonly been used as outcomes in clinical research. While measurements of VDphys and VDalv are not dependent on pharmacological paralysis, they do require minimal tracheal tube leak, adequate expiration time (tE), an arterial line in situ for blood gas analysis and PeCO2 calculation. This may make it an impractical measure for many patients in the intensive care unit.

Volume or pressure pre-set ventilation modalities may have had the potential to influence deadspace measurements. There were no group changes in VTE in the subgroup of patients receiving volume pre-set ventilation. However, even in this subgroup, VDalv increased significantly following physiotherapy alone, suggesting that changes in the distribution of alveolar ventilation occurred, rather than gross changes in VTE. A proportion of the ventilation delivered could, for example, have been re-directed from hyper-expanded regions to previously atelectatic regions, recruited during treatment. In this subgroup, there was also a significant reduction in PIP (p<0.05) after physiotherapy, indicating an improvement in respiratory compliance. The recently proposed “avalanche” model of inflation of collapsed regions of the lung suggests that rapid and variable pressure changes within the lung (such as those that may be generated during physiotherapy) may open up a successive stream of previously closed airways [29]. This theory may account for changes in regional ventilation reflected by the increase in VDphys and VDalv, despite the absence of gross changes in VT or PaCO2.

Cyanotic heart disease

Children with cyanotic congenital heart disease have significantly larger VDphys/VT than normal children or those with acyanotic heart disease (p<0.01) [30], as a consequence of intra-pulmonary shunt. This was confirmed by the increased baseline values of VDphys and VDalv in the cardiac population, though most infants had already undergone corrective surgery. The changes observed after physiotherapy, however, could not be attributed to an increase in right-left shunt, since this would very likely be accompanied by an increase in the arterial-alveolar gradient (reflected by an increase in PaCO2). This was not supported by the data. In addition, suction alone did not result in any change in VDphys/VT in the cardiac population and significant increases in VDphys were also demonstrated in the non-cardiac population after physiotherapy.

Broncho-constriction

Some physiotherapy techniques may cause broncho-constriction [31] and VDana decreases in response to bronchial constriction [32]. In this study, VDana did not change in either group, indicating that broncho-constriction was not an important consequence of treatment in this sedated and muscle-relaxed patient population. VDana may be a more important outcome in self-ventilating patients or those with underlying bronchial hyper-responsiveness.

ETCO2 vs PaCO2

End tidal carbon dioxide monitoring is easily available, non-invasive and has been investigated at length as an alternative to PaCO2. While some studies suggest that ETCO2 and PaCO2 may sometimes be used interchangeably [33, 34], others have noted that there are many clinical scenarios in which ETCO2 cannot accurately predict PaCO2 [35, 36, 37]. The current study confirmed that dynamic and sometimes rapid changes in deadspace volumes following clinical interventions make ETCO2 an extremely unreliable substitute for PaCO2 in an intensive care unit.

Conclusion

We have previously reported material differences in the physical delivery of physiotherapy and suction treatments with subtle differences in their effects on respiratory function [8]. The findings presented in this study confirm that the interventions have small but distinguishable effects on regional ventilation and V/Q balance, with physiotherapy resulting in significant increases in VDalv and VDphys.

Measures of VDphys and VDalv show promise as sensitive indicators of subtle changes in gas exchange and regional ventilation. However, changes due to clinical interventions such as physiotherapy are likely to be complex. Meaningful interpretation of these outcomes is absolutely dependent on concurrent evaluation of changes in the parameters from which they are derived (VTE, VDana, PeCO2 and PaCO2).