Elective High-frequency Oscillatory Ventilation Versus Conventional Ventilation in Preterm Infants With Pulmonary Dysfunction: Systematic Review and Meta-analyses
Objectives. To systematically review the evidence to determine whether the routine use of high-frequency oscillatory ventilation (HFOV) as compared with conventional ventilation (CV) is beneficial or harmful in preterm infants requiring mechanical ventilation for pulmonary failure principally due to respiratory distress syndrome.
Methods. All randomized controlled trials of elective HFOV versus CV in preterm infants <36 weeks' gestation with respiratory failure mainly attributable to respiratory distress syndrome were identified from the literature through a search of MEDLINE, EMBASE, Oxford database of Perinatal trials, and previous reviews including cross-references and abstracts. Meta-analyses using event rate ratios (ERR), event rate difference, and if significant, number needed-to-treat were calculated (95% confidence limits were used for all analyses). Two prespecified subgroup analyses were performed.
Results. Four published trials918 were included. Meta-analyses revealed the following ERR (95% confidence intervals) for HFOV versus CV: mortality at 28 to 30 days, 1.02 (0.76, 1.39); chronic lung disease (CLD) at 28 days, 0.86 (0.73, 1.01); mortality or CLD, 0.9 (0.80, 1.01); air-leak syndromes, 1.13 (0.97, 1.33); mechanical ventilation at 28 days, 1.06 (0.84, 1.33); supplemental oxygen at discharge, 0.59 (0.37, 0.92); intraventricular hemorrhage (IVH) all grades, 1.11 (0.95, 1.29); IVH (grades 3 or 4), 1.32 (1.01, 1.72); and periventricular leukomalacia, 1.39 (0.91, 2.13). In the subgroup of trials in which a high volume strategy (HVS) was used18 the ERR for CLD was 0.53 (0.36, 0.78); mortality or CLD, 0.56 (0.40, 0.77); supplemental oxygen at discharge, 0.57 (0.36, 0.92); IVH (all grades), 0.90 (0.61, 1.33); and IVH (grades 3 or 4), 0.84 (0.39, 1.84). Results were similar to these for the trials using surfactant.1920 One recent trial suggests that HFOV may reduce the cost of in-hospital care.19
Conclusions. The overall meta-analysis is dominated by the HIFI study,9 which was criticized for its methodology11 and surfactant was not used. Subsequent studies, most of which used HVS and/or surfactant, have shown benefits in measures of CLD without an increase in rates of IVH. Caution is warranted in interpreting these results because: 1) the treatment is not blinded and this could affect some outcomes; 2) except for one small trial20 postneonatal survival, lung function, and neurodevelopment have not been reported from HVS trials; and 3) the benefits and disadvantages have not been reported in infants born at different gestational ages or different birth weights. Importantly, results from groups experienced in the use of HFOV may not be readily generalizable.
- high-frequency ventilation
- high-frequency oscillatory ventilation
- chronic lung disease
- neonatal ventilation
- neonatal morbidity
Although assisted ventilation has reduced mortality, morbidity with chronic pulmonary disease is a significant problem.1 Chronic lung disease (CLD) develops in 20% to 60% of preterm infants with respiratory distress syndrome (RDS) due in part to barotrauma from conventional ventilation (CV).1
Animal studies have suggested that the use of high-frequency ventilation is associated with effective gas exchange, less barotrauma, and it may be life saving in situations in which CV has failed.5,6 In premature baboon models of hyaline membrane disease, high-frequency oscillatory ventilation (HFOV) results in more uniform lung inflation pattern, improves oxygenation, and reduces the severity of lung pathology produced by assisted ventilation.7,8
Surprisingly the first multicenter randomized trial of HFOV versus CV by the HIFI Study Group9 failed to show any benefit in decreasing the frequency of CLD. On the contrary, it showed an increased incidence of intraventricular hemorrhage (IVH) and neurodevelopmental abnormalities at follow-up.10 There were criticisms of the trial methodology particularly regarding the large intercenter variability in outcomes and the failure to use measures to recruit and maintain lung volumes.11 Since the HIFI study,9 other randomized trials have been conducted in the face of changing perinatal practice such as introduction of surfactant replacement therapy and increasing use of antenatal corticosteroids. In view of the small number of patients in the recent trials, it was felt appropriate to do a systematic review and a meta-analysis of the results of all these trials.
All randomized studies were sought with the use of the MEDLINE bibliographic retrieval system (National Library of Medicine) by means of the MeSH (medical subject heading thesaurus) terms “high frequency ventilation,” and “high frequency oscillatory ventilation” from the years 1980 to 1995. The initial search was performed in June 1995 and updated in October 1996. The EMBASE database was also searched under the same terms from the years 1989 to 1996. Further studies identified in reference lists of publications noted above and in a review article were also included. The Oxford database of perinatal trials was also searched and trials identified by the Neonatal Review Group of Cochrane Collaboration were available. Information was also obtained from experts in the field. Expert informant searches were carried out in the Japanese language by Professor Ogawa.
Because it has been shown in the laboratory that there is difficulty in achieving alveolar expansion after exposure to CV12 it was decided to include only studies that randomized patients to HFOV early and electively rather than as rescue therapy. To be included in the review trials had to meet each of the following additional six criteria: 1) published randomized controlled study; 2) study infants had to be <36 weeks' gestational age, or with a birth weight <2 kg; 3) ventilated for pulmonary dysfunction principally due to RDS; 4) electively randomized in the first 24 hours of life; 5) the intervention was HFOV; and 6) no mandatory crossover.
The outcomes examined included: mortality at 28 to 30 days, CLD, which was defined as oxygen dependency at 28 to 30 days with chest radiography changes, supplemental oxygen at discharge, oxygen at 36 or 37 weeks' postconceptual age, mechanical ventilation (MV) at 28 to 30 days, air-leak syndrome (ALS) (pneumothorax, pulmonary interstitial emphysema, and so forth), all grades of IVH, IVH grades 3 or 4,13 and periventricular leukomalacia (PVL). Also reviewed were long-term pulmonary and neurodevelopmental follow-up outcomes.
This review used the guidelines of the Cochrane Collaboration as outlined in the Cochrane Library14 and in the textEffective Care of the Newborn Infant.15 An earlier version has been published as a Cochrane review.16The data were extracted separately by each author and then compared. Meta-analyses using event rate ratios (ERR), event rate difference (ERD), and if significant, number needed-to-treat (NNT) were calculated (95% confidence limits were used for all analyses). The data were synthesized using Meta-Analyzer, version 1.2 (Update Software Ltd, Oxford, England).
For these analyses the authors chose to base validity assessment on four methodologic criteria that can be associated with significant bias in trials assessing treatment effect17 and these were evaluated separately by each author. These included concealment at randomization, blinding of treatment, blinding of outcome assessment, and completeness of follow-up. No scoring system is incorporated in this particular method of evaluation. The final assessment of the validity of the studies included is therefore left to the individual reader.
Two subgroup analyses were determined a priori and independently confirmed as relevant by a group of four neonatologists. The first subgroup analysis was based on whether a high volume strategy (HVS) was used in the HFOV ventilated group. To be included in this subgroup, trials needed to fulfill two of the following three criteria with which patients in the HFOV group were ventilated after randomization. The three criteria were: ventilation with mean airway pressures more than those of CV after randomization, adequate alveolar recruitment maneuvers (sighs, bagging, or increasing mean arterial pressure briefly), and weaning of fractional inspired oxygen before mean airway pressures. The second subgroup analysis was based on whether surfactant replacement therapy was used in all patients with RDS.
Additional information was obtained from Clark et al18regarding the outcome of infants excluded after randomization and this allowed intention-to-treat analyses. Gerstmann et al19provided additional information regarding methodology in their trial.
Eight randomized controlled trials were found in the search.9,18 Four met the inclusion criteria and were published in full.9,18 There were two recently presented abstracts23,24 that were not included in the meta-analyses because it was not possible to assess the methodology and the outcomes according to the a priori protocol. The HIFO study21 was excluded from the analysis because HFOV was used for rescue rather than electively. The study by Ramanathan et al22 published as an abstract was also excluded because there was a mandatory crossover from HFOV to CV at 96 hours of age. Three trials were deemed as having used HVS based on predetermined conditions.18 The HIFI study9 did not meet the criteria to be included in the subgroup analysis of HVS. No subgroup analysis by gestational age or birth weight, except for some pulmonary outcomes, was possible because only one trial published stratified outcomes.19 Characteristics of the trials included in the review are shown in Table1. The important methodologies used in each trial are summarized in Table 2.
There was no significant difference in neonatal mortality in any individual trial or in the overall analysis (Table3). Subgroup analysis also failed to show any difference in mortality (see Tables 9 and 10).
There were no significant differences in incidence of ALS in individual trials or the overall analysis (Table4). In the subgroup analysis of trials using a HVS the incidence of ALS was similar in the two treatment groups (see Table 9). In the subgroup analysis of trials in which surfactant was used there was a trend towards a reduced incidence of ALS in patients randomized to HFOV (see Table 10).
In all but one trial9 and in the overall analysis, the trend is towards a reduced incidence of CLD at 28 to 30 days of age that was not significant (Table5). In the subgroup in which HVS was used there was a significant reduction in risk of CLD at 28 to 30 days [ERR, 0.53 (0.36, 0.78), NNT, 7 (4, 24) (see Table 9)]. This result suggests that, on average, for every seven infants treated, one case of CLD at 28 to 30 days would be prevented. In the subgroup with surfactant replacement therapy there is a similar significant reduction in CLD at 28 to 30 days in the HFOV group [ERR, 0.60 (0.37, 0.96); see Table 10].
In the overall analysis there is a trend towards a reduced risk of death or CLD in the HFOV group (Table 6). In the subgroup analysis of trials using HVS for HFOV death or CLD is significantly reduced in the HFOV-treated infants [ERR, 0.56 (0.40, 0.77), NNT, 6 (4, 15); see Table 9] and there is a similar effect in the subgroup in which surfactant was used [ERR, 0.54 (0.33, 0.87); see Table 10].
Oxygen use at 36-weeks' postconceptual age in survivors to discharge was a documented outcome in only one of the studies and was reduced in the HFOV group [3/24 vs 10/22; ERR, 0.38; (0.16, 0.86)].18 Oxygen therapy at discharge is documented in two studies.18,19 In the study by Clark et al18none of the 23 in the HFOV group and 2 of 21 in the CV group required supplemental oxygen at discharge [ERR, 0.23 (0.01, 5.06)]. In the trial by Gerstmann et al19 17 of 63 in the HFOV group and 27 of 59 in the CV group required supplemental oxygen at discharge. The ERR of the combined data18,19 is 0.58 (0.36, 0.92); NNT, 8 (4, 77).
Use of MV
The use of MV at or beyond 28 days was reported in three studies.9,19,20 In the HIFI study9 there is no difference in the rate of MV at 28 days [HFOV 87/327 vs CV 85/346 all infants; ERR, 1.08 (0.8, 1.46)]. Ogawa et al20reported that 13 out of 46 in the HFOV group and 9 out of 45 in the CV group who survived to 28 days, were still on MV (ERR, 1.41; 0.6, 3.31). Gerstmann et al19 reported this outcome and found that 9 out of 64 in the HFOV group and 12 out of 59 survivors in the CV group were on MV beyond 28 days (ERR, 0.69; 0.29, 1.64). The pooled ERR was 1.06 (0.85, 1.34). In the trial by Gerstmann et al19 the median (95% confidence interval) days on MV in those with a birth weight less than 1 kg is 24.7 (3.7, 61.4) in the HFOV group and 53.7 (28.4, 103) in the CV group, which is not significantly different. In this trial there are also similar median durations of MV in infants with birth weights more than 1 kg [(HFOV group, 4.1 (1.7, 6) vs CV group, 4.5 (3, 6.1)]. Clark et al18 reported median and wide ranges for the days on MV for all infants entered in the study that was not significantly different between the HFOV group (16; 1.8, 67) and the CV group (30.3; 0.5, 222).
Long-term Pulmonary Outcomes
Follow-up assessments (in 82% of survivors), including pulmonary function tests (in 43% of survivors) were carried out at 9 months corrected age on infants who were in the HIFI trial.25There were no significant differences in the rate of growth, incidence of respiratory tract infections, hospital readmission, retractions and episodes of wheezing, or in respiratory function tests. Twelve month follow-up of patients in the trial by Ogawa et al20 showed persistence of abnormal fibrous or emphysematous shadows on chest radiography in 2 of the infants in the HFOV group and 4 in the CV group.
In the overall analysis there is a trend towards increased risk of IVH of all grades in those treated with HFOV, which was not statistically significant (Table 7). This trend was not evident in the subgroups in which HVS or surfactant were used (Tables 9 and 10). The rates of more severe IVH (grades 3 or 4) are increased in the HIFI study9 and there is a similar effect in the overall analysis, which is statistically significant [ERR, 1.32 (1.01, 1.72), NNT, 35, (14, 84) Table8]. In the subgroup analysis of infants treated with surfactant and those in which a HVS was used for HFOV, there is a nonsignificant trend towards lower rates of grades 3 or 4 IVH in the HFOV-treated infants (Table 9and Table 10).
PVL was reported in four of the five studies and there is a nonsignificant trend towards an increased rate in the HIFI study9 and in the overall analysis [43/429 vs 32/447, ERR, 1.39 (0.91, 2.13), ERD, 2.32 (−1.41, 5.87)]. There is no such trend in the subgroup analyses of patients in which the HVS strategy was used or in those with surfactant treatment.
Neurodevelopmental Outcomes at Follow-Up
Neurodevelopmental status was assessed at 16 to 24 months corrected age in 77% of survivors of the HIFI study trial9 (HFOV, 185 and CV, 201) using Bayley's psychometric evaluations and central nervous system examinations.10 Cerebral palsy was diagnosed in 10% of HFOV-treated infants and 11% in CV infants. A significantly higher incidence of hydrocephalus (12% vs 6%) was present in the HFOV group. The overall proportion of children with abnormal neurodevelopmental status was significantly higher in the HFOV group (65% vs 54%). The authors concluded that there were more neurologic deficits related to higher proportion of survivors with major IVH in the HFOV group.
One year follow-up in the trial by Ogawa et al20 found 4 infants in each group (8.7% vs 9%) had delays in motor and/or mental development, although the method of neurologic assessment was not given. No follow-up results have been reported for the other trials.
Total Hospital Costs
The total hospital costs from a subgroup of patients from one center in the trial by Gerstmann et al19 suggests that the median hospital costs were less in the group of patients randomized to HFOV. However, there was no reduction in the length of hospital stay.
The methods used in this review were those recommended by the Cochrane Collaboration.14 The main strength of this approach is that it endeavors to minimize bias by the use of rigorous methodology. This includes a priori setting up a protocol for the review that explicitly states the objectives, the criteria for considering studies for inclusion and exclusion, the search strategy, which is as comprehensive as possible, and the standards for assessing trial quality and for data extraction and analysis. If subgroup analysis is planned, the criteria are established before the search or data analysis, in keeping with the initial objectives of the review. Reviewer bias is minimized by independent assessment of trials for quality and independent data extraction by at least two authors.
In this review, the search revealed eight possible trials, four of which were published in full and the others were in various stages of publication. Additional information was obtained from Clark et al18 and Gerstmann et al19 to complete missing information about methodology or results. Wherever possible raw data was reanalyzed on an intention-to-treat basis.18 It is possible that there are other trials that had not been published or published in a language not covered by this systematic review.
Four of the trials were included for analysis because they met the prespecified criteria. The range of gestational ages and birth weights of infants enrolled in the trials was large. Although some authors stratified by weight or gestation at randomization, little data has been published by these strata. The four trials excluded were the HIFO study,21 which was mainly designed to detect the difference in the rate of ALS rather than mortality and morbidity. Furthermore, this study randomized infants with more severe RDS and at a later age (mean 21 hours). This was considered to be primarily a rescue rather than an elective use of HFOV. The other three studies were by Ramanathan et al,22 Rettwitz-Volk et al,23 and Lambert et al.24 The study by Ramanathan was excluded from the main analysis because there was a mandatory crossover from HFOV to CV at 96 hours of age. This was felt to impair the ability to evaluate respiratory outcomes in the same way as assessed by the other studies in which crossover was allowed but not mandatory. Nevertheless, because IVH is primarily a disorder occurring in the first few days of life, this outcome is still of interest and is discussed below. The results of trials by Rettwitz-Volk23 and by Lambert et al24 were excluded from the meta-analysis because they had not been peer reviewed and it was difficult to assess the methodology and outcomes from the abstracts.
In the overall analysis of studies in which HFOV was used electively, there is no evidence to suggest that this form of therapy affects the incidence of air-leak syndrome. In the overall analysis, there is also no evidence for a reduction in short-term measures for CLD (oxygen dependency and an abnormal radiograph at 28 to 30 days); however, this result is dominated by the HIFI study,9which is the largest. After the completion of this study, it was criticized because the methodology used to apply HFOV did not include methods to recruit lung volume.11 This criticism is supported by the results of subsequent trials that used a HVS in which there is a significant reduction in this measure of CLD. Furthermore, two of the trials found reductions in longer term oxygen requirements at 36 weeks18 or at discharge.18,19 This was the main reason for introducing HFOV for the treatment of RDS and the finding is supported by animal studies reporting the reduction in lung injury with this form of treatment. The number of patients receiving more than one dose of surfactant was also lower in the HFOV group in the study by Gerstmann et al.19 This trial was also the only trial to show a reduction in the total costs in the HFOV group, the reasons were thought to be attributable to patients needing lower degrees of support, exhibiting stability earlier in the hospital course, and having fewer critical setbacks. This data however was only from one of the centers of this multicenter trial. The total number of hospital days was not significantly different in the two groups. In the excluded abstract by Lambert et al24 there was no significant difference in the incidence of CLD.
A major concern, which first arose with the HIFI study,9 was the increased rates of acute and chronic neurologic injury that seemed to be associated with HFOV. This has been measured in the neonatal period by assessing rates of IVH and PVL and at neurodevelopment follow-up within the first 3 years of life.
Adverse neurologic outcomes have not appeared in all trials. The increased rates of all IVH in the large HIFI study9contributes to the nonsignificant trends in these outcomes in the overall meta-analysis. In the excluded trial by the HIFO study group,21 81 infants randomized to initial HFOV treatment and 84 infants randomized to CV had preentry and postentry ultrasounds. At study entry, 10 patients in each group had grade 1 or 2 IVH. After study entry, the incidence of all grades of IVH was greater in the HFOV group (29/81, 36%) than in the CV group (17/84, 20%;P = .037). The number of infants for each group with grade 3 or 4 IVH was also higher in the HFOV group (HFOV, 6 vs CV, 2;P = .041). This latter result was no longer significant when confounding factors such as birth weight were controlled for in the logistic regression. In the abstracts reporting the trials by Ramanathan et al,22 Rettwitz-Volk,23 and Lambert et al24, no difference in the rates of grade 3 or 4 IVH was found.
The authors of the HIFI study9 suggested that the nearly constant mean airway pressure during HFOV might restrict venous return, increase intracranial venous pressure, and decrease cerebral blood flow. However, animal studies,26 and a recent human study,27 failed to show these changes. Failure to use lung volume recruitment and the consequent cardiorespiratory instability were thought to be other mechanisms implicated.11 A recent publication of another high-frequency mode, namely jet ventilation, found an increased incidence of cystic PVL with the ventilatory strategy used in that study.28 Another recent meta-analysis of IVH and high-frequency ventilation in which two different modes of high-frequency ventilation (HFOV and high-frequency jet ventilation) were combined showed a nonsignificant trend towards increase in IVH but a significant increase in PVL if the HIFI study9 was included.29
Whether it was one of the mentioned mechanisms or just lack of experience with a new technology at the time, the HIFI study9 demonstrated how narrow the risk benefit can be with this technique. This adverse outcome is not apparent in the subgroup analysis of trials in which HVS was used with or without surfactant. Other factors such as improved overall neonatal care, increased use of antenatal corticosteroids (19% CV vs 30% HFOV in the trial by Gerstmann et al19 and 64% CV vs 78% HFOV in the trial by Rettwitz-Volk et al23), and increased experience with HFOV may also have contributed. Thus, it would be very important to have long-term outcome data from the trials that have used HVS to establish beyond doubt that it is safe to use HFOV electively in preterm infants with RDS.
Implications for Clinical Practice
The results of this review and the meta-analyses of elective HFOV in preterm infants with RDS suggest that there is some evidence of benefit in terms of decreased incidence of CLD at 28 to 30 days (a short-term pulmonary outcome), supplemental oxygen at discharge and CLD or mortality in a subgroup of infants who were ventilated with the HVS with or without surfactant. However, caution is warranted interpreting this result because: 1) the treatment is not blinded and this could affect assessment of some outcomes; 2) except for one small trial20 postneonatal survival, lung function, and neurodevelopment have not been reported from HVS trials; 3) the benefits and disadvantages have not been reported in infants born at different gestational ages or of different birth weights. Importantly, results from groups experienced in the use of HFOV may not be generalizable.
Implications for Research
Future studies need to target infants most at risk of developing CLD. Stratification by gestational age, birth weight, and severity of disease would be important. There is a need for more data on the optimum strategy for safely ventilating neonates with HFOV. The economic implications along with important long-term pulmonary and neurodevelopmental outcomes, are also important issues that need to be addressed.
- Received January 24, 1997.
- Accepted April 11, 1997.
No reprints available.
- CLD =
- chronic lung disease •
- RDS =
- respiratory distress syndrome •
- CV =
- conventional ventilation •
- HFOV =
- high-frequency oscillatory ventilation •
- IVH =
- intraventricular hemorrhage •
- MV =
- mechanical ventilation •
- ALS =
- air-leak syndrome •
- ERR =
- event rate ratio •
- ERD =
- event rate difference •
- NNT =
- number needed-to-treat •
- HVS =
- high volume strategy •
- PVL =
- periventricular leukomalacia
- Sauve RS,
- Siaghal N
- Truong WE,
- Standaert TA,
- Murphy JH,
- Woodrum DE,
- Hodson WA
- Meredith KS,
- deLemos RA,
- Coalson JJ,
- et al.
- deLemos RA,
- Coalson JJ,
- deLemos JA,
- et al.
- The HIFI,
- Study Group
- Froese AB,
- Bryan AC
- Papazoglou K,
- Suzuki H,
- Bohn DJ,
- Bryan AC
- ↵The Cochrane Library [database on disk and CD Rom]. The Cochrane Collaboration; Oxford Update Software; 1996 updated quarterly; available from BMJ Publisher
- ↵Sinclair JC, Bracken MB. Effective Care of the Newborn Infant. New York, NY: Oxford University Press; 1992
- ↵Bhuta T, Henderson-Smart DJ. Elective high frequency oscillatory ventilation vs conventional ventilation in preterm infants with acute pulmonary dysfunction. In: Sinclair J, Bracken M, Soll RF, Horbar JD, eds. Neonatal Module of The Cochrane Database of Systematic Reviews. Available in: The Cochrane Library [database on disk and CD Rom]. The Cochrane Collaboration; Issue 4. Oxford: Update Software; 1996
- Clark RH,
- Gerstmannn DR,
- Null DM Jr.,
- deLemos RA
- Gerstmann DR,
- Minton SD,
- Stoddard RA,
- Meredith KS,
- Bertrand JM
- ↵Ramanathan R, Ruiz I, Tantivit P, Cavabvab R, deLemos R. High frequency oscillatory ventilation compared to conventional mechanical ventilation in preterm infants with respiratory distress syndrome. Abstract. San Diego, CA: Society for Pediatric Research; 1995
- Rettwitz-Volk W,
- Veldman A,
- Roth B,
- et al.
- ↵Lambert J, Clarke O, Debnuche C, et al. High Frequency Oscillation Versus Conventional Mechanical Ventilation for Respiratory Distress Syndrome. Abstract. Lyon, France: European Society for Paediatric Research; 1996
- The HIFI,
- Study Group
- Laubscher B,
- van Melle G,
- Fawer C-L,
- Sekarski N,
- Calarne A
- Wiswell TE,
- Graziani LJ,
- Kornhauser MS,
- et al.
- Clark RH,
- Dykes FD,
- Bachman TE,
- Ashurst JT
- Copyright © 1997 American Academy of Pediatrics