INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Extracorporeal membrane oxygenation (ECMO) is a technically demanding and expensive means of providing temporary support during respiratory failure. It is most commonly used in newborn infants, and has been available for neonates in the United Kingdom since 1989. Although early published studies had suggested that use of ECMO led to a reduction in mortality (1), these were not conclusive. In addition, concerns were raised that ECMO might improve survival at the cost of long-term disability. For these reasons, a national, randomized trial was undertaken of ECMO (hereafter referred to as the main ECMO trial) for mature neonates with reversible respiratory disease, in which the main outcome measures were death or severe disability at 1 yr of age. This trial showed that ECMO conferred a survival advantage over conventional management (CM), without a concomitant increase in severe disability (4, 5).

At the inception of the main ECMO trial, it was envisaged that one possible outcome would be similar rates of survival and severe disability in the ECMO and CM groups, but better respiratory status in one group or the other. Whether or not ECMO conferred a survival advantage, it was considered essential to investigate the respiratory function of survivors of both limbs of the trial. ECMO could potentially result in survival of infants with severe respiratory dysfunction who would otherwise have died, which would result in poorer respiratory status in this group. Alternatively, those infants receiving ECMO might be spared aggressive ventilation and consequent barotrauma, which has been shown to be associated with subsequent alterations in respiratory mechanics (6). In this case the infants in the CM control group would be at a disadvantage.

In other studies of respiratory function in infants receiving ECMO, measurements were made during or shortly after ECMO (9), or at age 6 mo (12). Follow-up studies of respiratory function in subjects receiving intensive care in the neonatal period have shown abnormalities extending into childhood and beyond (13). We chose to study the infants reported here at the age of 1 yr because at this age the acute effects of disease and treatment would have receded, and any effect on growth of lungs or airways should be apparent. In addition, this timing of respiratory function tests in our study population provided an opportunity for us to gain an overview of clinical status and respiratory morbidity during the first year of life.

The aim of this study was therefore to compare respiratory health and function at 1 yr of age in infants who were assigned to receive ECMO with that of similar infants who were assigned to CM.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

All infants who were recruited into the main ECMO trial and who survived to 1 yr of age were eligible for the respiratory follow-up. Entry criteria for the main ECMO trial are published elsewhere (4). In brief, neonates were eligible if they were born after at least 35 completed weeks of gestation, had a birth weight of over 2 kg, and had severe respiratory failure. Infants had to be less than 28 d old at trial entry and had to have no contraindication for ECMO. One hundred and eighty-five infants were entered into the main ECMO trial, of whom 103 were discharged alive (65 assigned to ECMO and 38 to CM). Two infants died before 1 yr of age, and another two were lost to follow-up, leaving 99 infants (62 in the ECMO group) who were eligible for the respiratory follow-up. Twenty-one infants did not participate in respiratory follow-up because of parental unwillingness (eight cases), because the underlying diagnosis or clinical condition made testing inappropriate (eight cases), or because of repeated cancellations caused by onset of upper respiratory tract infection (URTI) (two cases), inability to attend before the child was 18 mo old (two cases), or adverse social circumstances (one case). Hence, respiratory follow-up was performed in 78 infants (51 in the ECMO group). Permission for the study was granted by the ethics committees of all the participating centers, and written consent was obtained from a parent of each infant.

Protocol

As each infant approached the age of 1 yr, the infant's family was contacted by the data coordinating center for our study and chose the most convenient center to attend for the follow-up (Leicester Royal Infirmary or Institute of Child Health, London). Parents were asked to defer the appointment if their infant developed a URTI during the 3 wk before testing.

Upon arrival, parents were asked not to disclose to the staff performing the test whether their infant had been in the ECMO or CM group of the ECMO trial. Infants wore a high-necked tunic to hide any neck scars that would have indicated to which limb of the trial they belonged. The infant was examined by a clinician uninvolved in testing, and a questionnaire was completed with the parents. A baseline measurement of oxygen saturation (Sa_O2) was made prior to sedation, and for safety reasons, (Sa_O2) was then monitored throughout measurements made in the study. The infant was weighed and sedated with chloral hydrate (up to 100 mg/kg) or a similar dose of triclofos sodium (up to 125 mg/kg). Occasionally, additional sedation was required, but in no infant did the total dose of either chloral hydrate or triclofos sodium exceed 1.5 g.

When the infants were sleeping, measurements were made of maximal expiratory flow at FRC (max_FRC), using the technique of rapid thoracoabdominal compression (RTC); this was followed by plethysmographic measurements of FRC (FRC_pleth) and airways resistance (Raw), using standardized techniques (18, 39). In brief, a plastic jacket with a double-walled anterior portion was wrapped around the chest and abdomen. In one center the infant's arms were outside the jacket, which was positioned high into the axillae. In the other center the infant's arms were positioned at the sides within the jacket, which encompassed the shoulders. The infant breathed through a pneumotachograph and face mask positioned around the nose and mouth, permitting the recording of respiratory flow and volume. Inflating the jacket at end-inspiration produced a firm squeeze that caused rapid exhalation, generating a partial maximal expiratory flow-volume curve. The RTC maneuver was repeated several times with a range of applied pressures, starting at approximately 2 kPa and increasing gradually to a maximum of 8 kPa (arms in) or 10 kPa (arms out), until additional increases in pressure no longer produced any increase in flow. Maximal flow provides a measure of peripheral airway function that is relatively independent of upper airway and nasal resistance (18, 19, 39), and is determined largely by the geometry of and resistive pressure losses along the small airways under conditions of dynamic compression. This measure has been widely used to characterize the normal growth and development of the airway during infancy and the pulmonary abnormalities associated with acute and chronic lung disorders during early childhood (19). Measurements of maximal flow were followed by plethysmographic measurements of resting lung volume at FRC and of Raw, using standardized techniques (18, 39). The face mask and pneumotachograph were attached to a block containing two pneumatic valves that could be used to switch the infant from breathing room air from within the plethysmograph to breathing heated, humidified air from a rebreathing bag as Raw was measured. These valves could also be closed simultaneously at end-inspiration for two or three respiratory efforts during the measurement of FRC_pleth. Recordings were examined immediately after airway occlusion, and pressure in the plethysmograph was plotted against that at the face mask. A straight line showed that the two signals were in phase, indicating that equilibration between alveolar and face-mask pressure had occurred, thereby providing reassurance that the measurements were accurate. Whenever possible, five or six technically satisfactory measurements of both Raw and FRC_pleth were made according to a predetermined protocol (20).

Equipment Details

The equipment in the two laboratories was broadly comparable. At Leicester, the infant whole-body plethysmograph was a commercially available instrument (Baby bodyplethysmograph; Jaeger GmbH, Wuerzburg, Germany) with minor modifications, and the rebreathing apparatus had been designed and built within the department. The pneumotachograph, pressure transducers (for measurement of flow, plethysmographic pressure, pressure at the airway opening, and pressure in the jacket) were part of the Jaeger Baby Bodytest system (Jaeger GmbH). In London, the plethysmograph had been built within the department and all transducers were of the same type (MP45; Validyne, Inc., Northridge, CA). The pneumotachograph was a Fleisch No. 1 capillary type (Fleisch, Lausanne, Switzerland). In both centers the transducers were matched and all equipment had an adequate frequency response. Dead space was similar though not identical in the two centers, and appropriate corrections were made during analysis.

In both centers all signals were collected in an IBM-compatible 486 personal computer (City Business Systems, Leicester, UK) and analyzed with identical software (RASP; PhysioLogic, Newbury, Berks, UK). Sampling rates were similar at both centers (80 Hz for all measurements in Leicester and for RTC in London, and 50 Hz for FRC_pleth and 100 Hz for Raw in London).

Data Analysis

Using standardized techniques (18, 39), we calculated the following measures from the recorded plethysmographic signals: tidal volume (VT), respiratory rate (RR), FRC_pleth, Raw at initial inspiration and at end expiration (RawII and RawEE), and max_FRC, the last of which was recorded both as the highest value and as the mean of the best four values. The optimal jacket pressure (the lowest pressure used to generate one of the four highest flows) was noted. The number of individual measurements and their standard deviations (where appropriate) were also recorded. Values of airway conductance (GawII and GawEE) were calculated from resistance values, and specific airway conductance (SGaw) was computed by dividing Gaw by FRC_pleth. This provides a measure of effective airway caliber corrected for lung size and hence somatic growth, and will be reduced if airway caliber is diminished and/or lung volume is increased in relation to body size. VT and RR were mean values from 25 breaths (the first 5 tidal breaths from the first five sets of data collected during RTC). FRC_pleth was the mean of all technically acceptable measurements made at end-inspiration. During the analysis of Raw, the operator selected portions of the signal train that were technically acceptable (i.e., best phase relationship between the recorded signals of plethysmographic pressure and flow) and analyzed all breaths individually. She then selected the seven breaths that appeared to be technically the best, and calculated the mean value. If fewer than seven breaths were acceptable, the mean of a minimum of five breaths was accepted.

The computerized analysis used in the study was highly interactive. To avoid potential bias in analysis techniques, raw data from each infant were exchanged on floppy disk and analyzed by both study centers (20). Data were acceptable if the two analyses agreed to within 5% (FRC_pleth), 10% (VT, RR, max_FRC), or 15% (Raw and Gaw). In cases for which the agreement fell outside these limits, each center checked its analysis, and differences were usually resolved. To ensure blinding, the analysis used to generate final results was that from the center at which the infant had not been tested.

The two groups of infants were compared with regard to weight, length, gender, and underlying diagnosis. In cases in which data were normally distributed, results of respiratory function tests were compared through means and 95% confidence intervals (CIs) of the differences between the two groups (21). When data were not normally distributed, results were compared through medians and interquartile ranges.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

The 78 infants who participated in the respiratory follow-up did not differ from the follow-up group of 99 survivors with respect to gestational age, birth weight, gender distribution, primary diagnosis, or age and severity of disease at trial entry (20, 22). At the time of testing, the two groups of infants were comparable in terms of demographic data (Table 1). Management between trial entry and discharge was similar, apart from a greater number of days on a ventilator or with very high oxygen requirements in the CM group (Table 2). Pulse oximetry was normal throughout testing in both groups. Each test center studied a similar number of infants, and the proportion from each limb of the trial was the same in both centers. Three infants in the ECMO group had had repair of congenital diaphragmatic hernia, whereas none of the infants with this condition in the CM group survived to 1 yr (5). One infant (in the ECMO group) was still receiving domiciliary oxygen at the time of respiratory follow-up, but was well enough to manage in room air for short periods, and his measurements were made during breathing of room air.

Design and analysis of the main ECMO trial was based on intention to treat, a policy that we continued when comparing respiratory function. Because the condition of three of the infants assigned to receive ECMO improved between trial entry and arrival at one of the ECMO centers, they were not connected to the ECMO circuit. Nevertheless, data from these three infants have been included in the data for the ECMO group.

There were no statistically significant differences between the ECMO and CM groups in family history of atopy, postnatal smoke exposure, household pets with fur or feathers, or central heating in the home (20, 22). Members of the CM group were more likely to be receiving respiratory medication at the time of testing than were members of the ECMO group, and tended to have more respiratory symptoms (5, 22), but these observations were likely to have been a consequence of the two treatment modalities, and should not invalidate any comparison of respiratory function test results in the two groups.

Respiratory Function

Complete data were available for most infants in the study. Measurements of RR and VT were missing for only one infant (in the CM group), who woke before measurements were complete. Measurements of FRC_pleth were missing for six infants (three in each group), and those for max_FRC were missing for five infants (four in the ECMO group). Fewer measurements of Gaw were available: data were missing for 16 infants in the ECMO group and six in the CM group. In some cases this was because the infant woke before measurements were complete; on other occasions accurate measurements could not be made for physiologic or technical reasons.

With the exception of one infant (who was hospitalized on a long-term basis for nonrespiratory reasons), all infants attended the study from home, and none was acutely unwell at the time of study. However, among individual infants there was a broad spread of respiratory function, with some subjects in both groups having entirely normal findings whereas others had severely compromised function.

Lung volume was greater in the CM group (95% CI of the difference between the ECMO and CM groups: 67, 4 ml) (Table 3, Figure 1). This statistically significant difference was equivalent to 13% of the lung volume of the ECMO group. The statistical significance of the difference was retained if lung volume was expressed per unit body weight (FRC_pleth was 27.0 ± 6.86 ml/kg [mean ± SD] in the ECMO group and 31.2 ± 7.60 ml/kg in the CM group; 95% CI [ECMOCM]: 7.86 to 0.64 ml/kg).

View larger version (15K): [in this window] [in a new window]   Figure 1.   Individual measurements of FRCpleth plotted against body length. Open circles: ECMO infants; open triangles: CM infants; open squares: infants with diaphragmatic hernia (all received ECMO); multiplication symbols: infants assigned to ECMO but who did not receive it.

Measured values of Raw (or its reciprocal, Gaw) did not differ in the two study groups at either initial inspiration or end expiration (Table 3). However, SGaw, measured at initial inspiration, was marginally lower in the CM group (95% CI of the difference between the ECMO and CM groups: 0.03; 0.98 s¹ · kPa¹) (Table 3), showing that effective airway caliber (corrected for lung size) was reduced in these infants.

There was a trend toward reduced maximal flows in the CM group (95% CI of the difference between the ECMO and CM groups: 2, 67 ml/s, p = 0.07) (Table 3), which just failed to reach the conventionally accepted criterion for statistical significance. This was strongly suggestive of diminished peripheral airway function in these infants. There was no difference in the optimal jacket pressure between the two groups (4.01 ± 1.82 kPa [mean ± SD] for the ECMO and 3.59 ± 0.24 kPa) for the CM group, p = 0.29), although higher jacket pressures were needed to achieve a similar transmitted pressure when the infants were studied with arms outside the jacket (18, 39).

There were no differences between the two groups in absolute VT or RR. Similarly, there was no difference in VT when expressed per unit body weight (8.7 ± 1.37 ml/kg [mean ± SD] in the ECMO group and 8.8 ± 1.88 ml/kg in the CM group).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main ECMO trial provided a unique opportunity to assess the respiratory outcome of a group of infants who had experienced neonatal respiratory failure, for whom comprehensive data describing status at trial entry and throughout infancy had been collected. The present study is the first reported respiratory follow-up of survivors of a randomized controlled trial of ECMO at 1 yr of age.

Infants in both the ECMO and CM groups showed a broad spectrum of abnormalities of respiratory function, from entirely normal function to major functional disturbances. Our results therefore confirm those of others who have shown that ECMO does not prevent sequelae of severe respiratory disease in the newborn period (12). In contrast to previous studies (9, 10, 12, 23), the present study included a contemporaneous control group of infants not assigned to ECMO, thereby enabling us to directly compare the respiratory outcome in the two groups.

For such a comparison to be meaningful, it was essential that infants in both groups were as similar as possible in every respect other than their respiratory management. In the present study there were no group differences with respect to age, weight, length, gestational age, birth weight, or ethnic origin. Nor were sociodemographic factors such as antenatal and postnatal smoke exposure, type of heating in the home, or exposure to pets different. These grounds justified a direct comparison of respiratory function between the two study groups.

For results of multicenter studies to be robust, it is clearly important for all centers to study groups that are representative of the whole population under examination. In the work reported here, the two test centers studied similar numbers and almost identical proportions of ECMO and CM infants. Furthermore, results from the two centers were very similar (20).

The measured value of FRC_pleth was higher in the CM than in the ECMO group. One possible explanation for this finding would be that the ECMO group had small lungs. The reasons for a reduced lung volume are usually congenital, and include hypoplasia associated with congenital diaphragmatic hernia, oligohydramnios, or thoracic dystrophy. The ECMO group included all three infants who had had repair of diaphragmatic hernia. Although we recognize that extreme caution was needed in dividing the infants in our study into diagnostic or other subgroups because of potential loss of power, we examined the study data to see whether these three infants had markedly low values of FRC_pleth and could therefore have reduced the mean FRC_pleth value for the ECMO group. This proved not to be the case (Figure 1), possibly because infants with congenital diaphragmatic hernia have measured lung volumes within the normal range by 1 yr of age, as the lungs expand (by increase in alveolar size rather than in number of alveoli) to fill the space available (24). Measurements of resistance and max_FRC were available for two of the three infants who had repaired diaphragmatic hernias: excluding these infants from the analysis produced no substantive changes in the results, although it is noteworthy that one of the three infants showed the lowest value of inspiratory resistance (1.01 kPa · L¹ · s) of any infant in the ECMO group, a finding that has also been reported previously (24).

An alternative explanation for differences in lung volume between the two study groups would be that the infants in the CM group (who received more mechanical ventilation than those in the ECMO group) were hyperinflated, a finding usually associated with gas trapping caused by peripheral airways obstruction (25). Peripheral airways obstruction is reflected in the measurement of max_FRC, which was generally lower in the CM group (although this just failed to reach statistical significance), supporting the hypothesis that infants in this group were more prone to gas trapping and hyperinflation.

Predicted values of lung volume are generally based on body length and, for an infant of 78 cm (the mean body length of the study group), the predicted lung volume is approximately 267 ml (19, 26), indicating that the infants in the CM group were indeed slightly hyperinflated.

Measurements of lung volume made by nitrogen washout in infants receiving full ECMO support, and repeated 24 h before weaning (10), have shown an increase from 3.6 ml/kg to 7.9 ml/kg. These values reflect both the very low lung volumes obtained during ECMO treatment at lung-rest ventilator settings, and differences in measurement technique. Mean lung-volume values measured in another group of 19 ECMO survivors studied at 6 mo (12) were smaller than those reported in the present study (18.5 ml/kg, versus 26.4 ml/kg [ECMO] and 30.4/ml/kg [CM]). All of these measurements were made by plethysmography, and the discrepancies may relate to the age at which infants were studied, since the infants in the present study were 6 mo older and therefore had had more time for lung growth and repair. Alternative reasons for the contrasting findings could relate to differences in populations and their diagnostic profiles. However, the possibility that apparently minor differences in equipment and technique could have resulted in substantial differences in calculated values of lung volume cannot be discounted, since the study by Garg and colleagues reported that the "low" volumes (with respect to most published data [27]) in their ECMO group were "normal" in relation to those of their own, albeit limited, reference population. This highlights the need for developing a standardized approach to hardware, software, and protocols in comparative studies of infant lung function (20).

Measured values of airway resistance in the ECMO and CM groups were not significantly different, although they were higher than the predicted value of 1.28 kPa · L¹ · s (27). Garg and colleagues reported a mean airway resistance of 5.07 kPa · L¹ · s with a specific conductance of 1.7 s¹ · kPa¹ in their ECMO group. Thus, although resistance was lower and lung volumes greater in the present study, mean SGaw was similar to that reported by Garg and colleagues (12). The finding of reduced airway function in survivors of neonatal lung disease has been reported since the 1970s (6, 28, 29), and attempts have been made to relate such reduction either to the underlying condition (8, 30) or to therapy (7, 17, 35). Differences between neonatal intensive care units in the management of severely ill newborns, and rapid changes in technology and available therapy (e.g., availability of pulse oximetry, surfactant therapy, high-frequency ventilation, nitric oxide), limit the value of comparison of the present group of infants with those reported in the literature. However, the comparison of the ECMO group with the contemporaneously treated CM group in the present study is valid. This shows that assignment to ECMO does not confer a statistically significant improvement in Raw, but that SGaw (measured in early inspiration) is better in the ECMO group. SGaw can be diminished either if airway caliber is diminished or if lung volume is increased, so that poorer specific conductance in the CM group could be attributed to the hyperinflation in this group. However, although the difference failed to reach significance, inspiratory airway resistance was increased to a greater degree in the CM group. We speculate that this group may have had a degree of volutrauma subsequent to their mechanical ventilation in the neonatal period, since this treatment is known to be associated with poorer airway resistance in infancy (7, 37). The precise mechanisms for the airway narrowing that results in increased resistance are uncertain, but may include inflammation of the airway wall, hypertrophy of bronchial smooth muscle, or the presence of airway secretions.

The measurement of max_FRC showed considerable variation in both the ECMO and CM groups, demonstrating that some infants had entirely normal functioning of the small airways whereas others were severely compromised. As compared with data obtained from a reference population (38) in whom expiratory flows in excess of 150 ml/s were reported at 1 yr in all infants, mean values of expiratory flow were reduced in both the ECMO and CM groups in our study. The variability inherent in the measurement of expiratory flow, depending on the stability of the end-expiratory level, may have contributed to the lack of a statistically significant difference in our observations. Our approach to attempting to minimize this variability was to examine both the highest recorded max_FRC and the mean of the four highest values recorded. This practice made no substantial difference in our findings, producing little change in the 95% CIs (Table 3). There was no difference in the optimal jacket pressure between the ECMO and CM groups, but higher pressures were needed at the center in which the jackets were used with the arms outside (20). This reflects rather lower pressure transmission from the jacket to the chest under this condition than when the arms and shoulders are enclosed within the jacket (18, 39). No current recommendations exist about the number of maneuvers to be performed for any one infant, nor is there any consensus about whether a single best value or a mean value should be reported. The results reported here suggest that there is little advantage in either approach, but recent developments in methodology and approach to the analysis of infant flow-volume curves may improve the reproducibility of max_FRC and other forced expiratory measurements (18, 39).

The mean values of max_FRC for the ECMO and CM groups in our study showed a strong trend toward better airway function in the ECMO group, although this just failed to reach statistical significance (95% CI: 2, 67 ml/s, p = 0.07). The original aim of this study was to examine at least 30 infants in each group, to achieve an 85% statistical power at the 5% level to detect differences equivalent to one standardized difference in each of the three respiratory outcome measures that we investigated. However, early termination of the main trial, when the survival advantages of ECMO became apparent, meant that we fell just short of this target.

When the ECMO and CM groups were compared, RR and VT did not differ. When compared with other published data for survivors of ECMO studied after the neonatal period (12), the mean RR reported here was slightly lower (33 versus 35 breaths/min), presumably reflecting minor differences in equipment or the later age of the infants in our study. The RR showed considerable intersubject variability (range: 19 to 74 breaths/min); the infant with the highest RR had received ECMO, and was still receiving domiciliary oxygen at the time of respiratory function testing. VT had a similarly wide spread of values. When expressed in terms of body weight, the mean values of VT were 8.8 and 8.7 ml/kg in the ECMO and CM groups, respectively. These were similar to the value reported in the literature for healthy infants (19), although lower than the mean values observed by Garg and colleagues (12).

Three infants assigned to the ECMO group did not receive ECMO because their condition on arrival at the ECMO center had improved. Because the study was randomized, it is likely that a similar number of infants in the other limb also showed an equivalent improvement, and we therefore had to conduct our analysis on an intention-to-treat basis. However, even if we had excluded these three infants, there would have been no substantive changes in any of our reported results. Figure 1 shows individual measurements of lung volume.

In summary, the findings of increased lung volumes, reduced SGaw during inspiration, and a tendency toward reduced forced expiratory flows in infants who were managed conventionally in our study suggests that the larger proportion of CM infants receiving respiratory medication and reporting respiratory symptoms at about 1 yr of age may be attributed to subtle impairment of small-airway function in these infants relative to those assigned to ECMO. These findings probably reflect differences in management during the neonatal period, because initial disease severity and all other background characteristics were similar in our ECMO and CM groups.

Between study entry and discharge, infants in the ECMO group were exposed for a significantly shorter time than those in the CM group to high fractions of inspired O₂ (> 0.90). They were also ventilated for significantly fewer days, which, in accord with the agreed ECMO protocol, included ventilation at lung-rest settings (4), thus limiting barotrauma and associated sequelae (7). It would therefore appear that treatment with ECMO may provide a small benefit over CM in terms of respiratory function, as well as a reduction in symptoms and decreased need for medication, as reported elsewhere (22). Although it could be argued that these small differences might not justify treatment as complex and expensive as ECMO, the survival advantage it confers is such that its continued use is likely. Although changing clinical practice since the inception of our study suggests that newer therapies (such as use of high-frequency ventilation and nitric oxide) may reduce the need for ECMO, there are few data on the outcome of such treatments in infants at 1 yr of age. At present, suggestions about whether it is better for an infant to receive ECMO or another, newer therapy are purely speculative. Furthermore, this study provides reassurance that respiratory function following ECMO is no worse, and indeed appears slightly better, than that following conventional treatment.