Acute respiratory distress syndrome (ARDS) was characterised by Ashbaugh and colleagues1 nearly half a century ago, and yet, to the casual observer, it might appear that little progress has been made since. Clinicians are still struggling with how best to define and treat ARDS.2 ,3 A major advance was the recognition that mechanical ventilation may induce further lung injury and the application of so-called lung protective ventilation (6 ml/kg ideal body weight tidal volumes and plateau pressures <30 cm H₂O) reduces mortality.4 High-frequency oscillatory ventilation (HFOV) is an attractive mode of ventilation because it combines improved lung recruitment with small tidal volumes, often below anatomical dead space, applied at very high frequency. Thus, it theoretically provides ultra-lung protective ventilation, limiting atelectrauma (caused by cyclical alveolar opening and closing) and volutrauma (alveolar over-distension). For those not familiar with this mode of ventilation, it should be noted that, unlike conventional ventilation, increasing the frequency of HFOV reduces CO₂ elimination. HFOV has often been used as a rescue therapy for severe ARDS and refractory hypoxaemia; however, uncertainty remains over its role in early ARDS. Although there is substantial experimental data, clinical studies are limited to case series and two small trials5 ,6 where controls did not receive what is now accepted as standard lung protective ventilation following the ARDSNet study.4

Two recently randomised clinical trials sought to establish if HFOV could reduce mortality in early ARDS.7 ,8 The UK-based OSCAR study7 randomised 795 patients to receive either HFOV or conventional ventilation. No difference was observed in 30 day mortality between groups (41.7% vs 41.1%, respectively). The larger OSCILLATE study8 planned to recruit 1200 patients but was stopped after 548 patients on recommendation from the data monitoring committee, as in-hospital mortality was significantly higher in those assigned to HFOV (47% vs 35%). Recruitment was broadly similar in both trials, which aimed to enrol patients with early ARDS and a PaO₂/FiO₂ ≤200 mm Hg rather than using HFOV as a rescue therapy. From these trials, it would appear that HFOV does not improve mortality in early ARDS, and at worse, might be harmful. However, although ostensibly both trials applied a similar intervention to a similar group of patients, there were significant differences between studies and also in how this technology was applied in UK Intensive Care Units prior to publication of these studies. The major differences between studies are summarised in tables 1 and 2.

Ventilator selection

For the past two decades, adult HFOV in Europe and North America has largely been undertaken using the SensorMedics 3100B oscillator (CareFusion). The OSCAR trial chose to use the Novalung R100 ventilator (Metran) device which, despite having been used in Japan, was relatively untried in the UK and had only been awarded a CE mark shortly before the start of the study. All 25 critical care units in the UK providing HFOV prior to the trial used the SensorMedics 3100B oscillator7; no explanation was given for choosing this unfamiliar equipment.

Centre selection

Centre characteristics can be an important determinant of outcome in mechanical ventilation as intensive care units dealing with higher caseloads have better outcomes.9 HFOV can be a complex intervention, with centres building up experience in its use over several years.10 OSCILLATE used 39 centres, predominantly North American hospitals with previous experience of HFOV employing the 3100B oscillator. In OSCAR, only 3 centres had significant experience of HFOV, 6 had limited experience and the remaining 20 had none. No centre had used the Novalung R100 previously and a substantial effort was put into training over 2300 clinical staff. Lack of equipoise is likely to have existed in the UK centres with substantial prior HFOV experience. The inability to use HFOV as a rescue therapy in the OSCAR trial is likely to have contributed to this. In OSCILLATE, 34 patients in the control group received HFOV (31 in accordance with the trial protocol); however, neither trial specifically recorded the use of HFOV outside of the trial in the severe ARDS subgroup.

Participant selection

Each trial broadly included similar patients; it is, however, notable that APACHE II scores were significantly higher in the OSCILLATE study (29 vs 21.8). In OSCAR, inclusion was dependent on the onset of mechanical ventilation within the previous 7 days, but up to 14 days of respiratory failure were permitted in OSCILLATE. Patients were included if the PaO₂/FiO₂ was ≤200 mm Hg on a PEEP of ≥5 cm H₂O⁵ or FIO₂≥0.5 irrespective of PEEP.8 The OSCILLATE trial used an additional step of standardised ventilator settings for 30 min (FIO₂≥0.6, PEEP of ≥10 cm H₂O and tidal volume 6 ml/kg predicted weight) to reassess if PaO₂/FiO₂ was still ≤200 mm Hg, but this excluded only 19 individuals.

Ventilation protocol

Ventilation strategies in these two studies were very different for HFOV and conventional ventilation arms (table 2). OSCILLATE adopted an ‘open lung’ approach from the outset by using recruitment manoeuvres in both arms. Following this, the smallest tidal volumes for either HFOV or conventional ventilation were applied, aiming for a pH>7.25. The oscillation strategy was based on a round-table discussion of experts in HFOV.11 Target oxygen saturations were 88–93% and were achieved using mean airway pressure-inspired oxygen charts, which in the conventional arm, were based on a previous trial.12 Recruitment manoeuvres were mandated frequently with increases in mean airway pressure (HFOV arm) or PEEP (conventional arm) within defined safety parameters. Importantly, cross-over to the other arm of the study could occur as a rescue intervention in refractory hypoxaemia. In contrast, OSCAR did not mandate recruitment manoeuvres in either arm of the trial and allowed ventilation in the conventional limb according to local practice but encouraged lung protective ventilation and setting of PEEP and inspired oxygen according to ARDSNet.

This lack of protocolised ventilation in the control arm of the OSCAR study may explain the large disparity in mortality in conventional ventilation groups between trials (table 2), especially as patients in OSCILLATE had higher APACHE II scores. Tidal volumes were also much lower in OSCILLATE; a recent observational study of ARDS found that non-adherence to tidal volumes of <6.5 ml/kg ideal body weight increased absolute mortality by 7.8% at 2 years.13 Failure to adhere to lung protective ventilation (30%) was also a feature of a recent UK trial comparing extracorporeal membrane oxygenation with conventional ventilation.14 It is conceivable that had the OSCAR trial achieved tidal volumes approaching 6 ml/kg ideal body weight, the two arms would not have demonstrated equivalently. In OSCILLATE, more patients in the intervention group required vasoactive drugs and required them for a longer duration, possibly due to the high inflation pressures and frequent recruitment manoeuvres associated with the protocol and perhaps a less aggressive recruitment strategy would have mitigated this.