Skip to main content

Foundational Trials

OSCILLATE

Image: Shutterstock / ChaNaWiT

Publication

  • Title: High-frequency oscillation in early acute respiratory distress syndrome
  • Acronym: OSCILLATE
  • Year: 2013
  • Journal published in: New England Journal of Medicine
  • Citation: Ferguson ND, Cook DJ, Guyatt GH, et al. High-frequency oscillation in early acute respiratory distress syndrome. N Engl J Med. 2013;368(9):795-805.

Context & Rationale

  • Background
    • ARDS has high mortality and morbidity despite lung-protective ventilation, and historically many “physiologically plausible” strategies have failed when tested against patient-centred outcomes.
    • High-frequency oscillatory ventilation (HFOV) delivers very small oscillatory tidal volumes around an elevated mean airway pressure, aiming to maintain recruitment while minimising volutrauma and atelectrauma.
    • Adult HFOV trials before OSCILLATE were relatively small, often compared against less standardised conventional ventilation, and largely showed improved oxygenation with uncertain mortality effects.
    • By the late 2000s, conventional ventilation had evolved (low tidal volume, plateau pressure limits, and higher-PEEP strategies), creating genuine equipoise about whether early HFOV adds benefit—or causes harm—when compared against contemporary best practice.
  • Research Question/Hypothesis
    • In invasively ventilated adults with early, moderate-to-severe ARDS, does early HFOV using an open-lung strategy reduce in-hospital mortality compared with protocolised conventional lung-protective ventilation?
  • Why This Matters
    • HFOV required specialised ventilators and typically deeper sedation/neuromuscular blockade, with potential haemodynamic consequences and major resource implications.
    • Clarifying benefit versus harm was essential to determine whether HFOV should be used early, reserved for rescue, or abandoned in adult ARDS.

Design & Methods

  • Research Question: Whether early HFOV with an open-lung strategy reduces in-hospital mortality versus protocolised conventional lung-protective ventilation in early moderate-to-severe ARDS.
  • Study Type: Multicentre, randomised, controlled, parallel-group, investigator-initiated trial conducted in ICUs (pilot phase: 12 centres; main trial: 39 ICUs in 5 countries); allocation via central web-based randomisation (variable block sizes of 2 and 4), stratified by centre; open-label.
  • Population:
    • Inclusion: Age 16–85 years; invasive mechanical ventilation; acute onset of pulmonary symptoms within 2 weeks; bilateral airspace opacities on chest radiograph; PaO2/FiO2 ≤200 mm Hg with FiO2 ≥0.5.
    • Eligibility confirmation: After enrolment, all patients placed on standardised settings (pressure-control ventilation; tidal volume 6 mL/kg predicted body weight; FiO2 0.60; PEEP 10 cm H2O or higher if needed); after 30 minutes, randomised if PaO2/FiO2 remained ≤200 mm Hg.
    • Timing: Not enrolled if eligibility criteria met for >72 hours.
    • Key exclusions: Hypoxaemia primarily related to left atrial hypertension; suspected vasculitic pulmonary haemorrhage; neuromuscular disorder known to prolong mechanical ventilation; severe chronic respiratory disease; pre-existing condition with expected 6‑month mortality >50%; risk for intracranial hypertension; lack of commitment to life support; expected duration of mechanical ventilation <48 hours; weight <35 kg or >1 kg per cm of height; already receiving HFOV.
  • Intervention:
    • Modality: HFOV using SensorMedics 3100B with an “open-lung” strategy.
    • Initiation: Recruitment manoeuvre 40 cm H2O for 40 seconds; initial mean airway pressure 30 cm H2O.
    • Targets and adjustments: Oxygenation adjusted using a prespecified algorithm (PaO2 target 55–80 mm Hg; SpO2 88–93%); ventilation adjusted to maintain pH 7.25–7.35 (typically using the highest feasible oscillatory frequency to minimise tidal volume).
    • Weaning/transition: After 24 hours, conventional ventilation could be resumed if mean airway pressure ≤24 cm H2O for 12 hours; mandatory transition once mean airway pressure reached 20 cm H2O; HFOV could be resumed within 48 hours if oxygenation deteriorated (FiO2 >0.4 or PEEP >14 for >1 hour).
  • Comparison:
    • Conventional ventilation: Pressure-control ventilation; tidal volume target 6 mL/kg predicted body weight (range 4–8); plateau pressure ≤35 cm H2O.
    • Oxygenation strategy: PEEP and FiO2 titrated using a prespecified high-PEEP oxygenation algorithm.
    • Ventilatory targets: Respiratory rate up to 35 breaths/min; pH target 7.30–7.45 (permissive hypercapnia allowed within protocol constraints).
    • Recruitment manoeuvres: Permitted and protocolised.
  • Blinding: Unblinded (ventilator strategy not feasible to mask); primary outcome (in-hospital mortality) objective, reducing risk of detection bias, but co-intervention differences remained possible.
  • Statistics: A total of 1200 patients were required to provide ≥80% power to detect a 20% relative reduction in in-hospital mortality (from an expected 45% in the control group) with a two-sided alpha of 0.05; primary analysis by intention-to-treat with relative risks and 95% CIs; formal interim monitoring used a very conservative efficacy boundary (P=0.001), and the independent data and safety monitoring committee could recommend stopping for harm.
  • Follow-Up Period: Until hospital discharge (primary outcome); key secondary outcomes through day 28; time-to-death analysis out to day 60 (patients discharged alive assumed alive at day 60).

Key Results

This trial was stopped early. Recruitment was halted following an interim analysis when 548 patients had undergone randomisation because mortality was higher in the HFOV group and there were concerns about harm (including greater haemodynamic support requirements).

Outcome HFOV (n=275) Conventional ventilation (n=273) Effect p value / 95% CI Notes
Death in hospital (primary) 129 (47%) 96 (35%) RR 1.33 95% CI 1.09 to 1.64; P=0.005 All-cause; analysed by randomised group
Death in ICU 123 (45%) 84 (31%) RR 1.45 95% CI 1.17 to 1.81; P=0.001 Objective outcome; susceptible to discharge practices only indirectly
Death before day 28 111 (40%) 83 (30%) RR 1.41 95% CI 1.09 to 1.83; P=0.004 Short-term mortality signal consistent with primary outcome
Refractory hypoxaemia 17 (6%) 34 (12%) RR 0.50 95% CI 0.29 to 0.84; P=0.02 Less refractory hypoxaemia despite higher mortality
Death after refractory hypoxaemia 11 (4%) 25 (9%) RR 0.44 95% CI 0.22 to 0.87; P=0.01 Suggests oxygenation failure was not the dominant mortality pathway in HFOV arm
New barotrauma 50 (18%) 39 (14%) RR 1.27 95% CI 0.87 to 1.86; P=0.24 Not statistically different
Refractory acidosis 20 (7%) 20 (7%) RR 0.99 95% CI 0.54 to 1.80; P=1.00 No signal of differential severe hypercapnic acidosis
Days of mechanical ventilation among survivors 11 (6–21) 10 (5–18) Not reported P=0.59 Median (IQR), among survivors only
ICU length of stay among survivors (days) 14 (9–28) 14 (9–26) Not reported P=0.77 Median (IQR), among survivors only
Midazolam dose (mg/day) 199 (100–382) 141 (68–240) Not reported P<0.001 Median (IQR) daily dose (in those receiving the drug)
Fentanyl equivalents (µg/day) 2980 (1258–4800) 2400 (1140–4430) Not reported P=0.06 Median (IQR) daily opioid dose (in those receiving the drug)
  • Mortality was higher with early HFOV (in-hospital death 47% vs 35%), leading to early trial termination.
  • HFOV reduced refractory hypoxaemia (6% vs 12%) but did not translate into improved survival, highlighting dissociation between oxygenation endpoints and patient-centred outcomes.
  • Subgroups (in-hospital mortality):
    • Recruitment intensity tertile: ≤4 OR 1.14 (0.28 to 4.68); 5–16 OR 1.37 (0.68 to 2.73); ≥17 OR 1.79 (1.18 to 2.70); interaction P=0.71.
    • BMI quartile: ≤25 OR 1.52 (0.88 to 2.64); 26–29 OR 1.85 (0.86 to 4.00); 30–34 OR 1.19 (0.55 to 2.55); ≥35 OR 2.20 (1.01 to 4.79); interaction P=0.70.
    • Baseline PaO2/FiO2 quartile: ≤86 OR 1.27 (0.65 to 2.47); 87–114 OR 1.79 (0.91 to 3.49); 115–147 OR 1.82 (0.90 to 3.64); ≥148 OR 2.12 (0.98 to 4.61); interaction P=0.44.
    • Baseline compliance quartile (mL/cm H2O): ≤20 OR 0.92 (0.51 to 1.64); 21–26 OR 2.34 (1.12 to 4.88); 27–34 OR 5.54 (2.27 to 13.55); ≥35 OR 1.13 (0.54 to 2.35); interaction P=0.23.
    • Baseline vasopressor use: Yes OR 1.70 (1.11 to 2.61); No OR 1.38 (0.75 to 2.52); interaction P=0.57.

Internal Validity

  • Randomisation and allocation:
    • Central web-based randomisation, stratified by centre, with undisclosed block sizes (2 and 4), supporting allocation concealment and minimising selection bias.
  • Dropout/exclusions:
    • At termination, 571 patients had been enrolled and 548 randomised (275 HFOV; 273 control); primary outcome data were available for the randomised cohort.
  • Performance/detection bias:
    • Open-label design was unavoidable; mortality is objective, but management co-interventions (sedation, vasoactive support, neuromuscular blockade, rescue therapies) could differ and potentially mediate or confound outcomes.
  • Protocol adherence and separation of the variable of interest:
    • Allocated strategy delivery was high: 270/275 (98%) in the HFOV group received HFOV (median 3 days, IQR 2–8).
    • Control crossover: 34/273 (12%) received HFOV as rescue (median 7 days, IQR 5–15), typically starting 2 days after randomisation; 24/34 (71%) of crossover patients died in hospital (severity confounding likely).
    • Ventilator separation (day 1 values): mean airway pressure 31.0 (SD 2.6) vs 24.0 (5.3) cm H2O; P<0.0001.
    • Oxygenation strategy separation: FiO2 day 1 0.62 (0.15) vs 0.68 (0.19); P=0.0007; PaO2/FiO2 day 1 172 (64) vs 163 (82); P=0.23.
    • Cointervention separation: any vasoactive drug use day 1 90.9% vs 83.5% (P=0.01); neuromuscular blockade use day 1 83.6% vs 62.7% (P<0.0001).
  • Baseline characteristics:
    • Groups were broadly comparable at baseline (e.g., APACHE II 29.0 ± 7.1 vs 29.1 ± 7.5; PaO2/FiO2 121 ± 45 vs 114 ± 42 mm Hg).
    • Duration of mechanical ventilation before randomisation differed (2.5 ± 3.3 vs 1.9 ± 2.9 days; P=0.003), a potential marker of baseline imbalance in timing/trajectory.
  • Heterogeneity and subgroup effects:
    • Prespecified subgroup analyses of hospital mortality did not show statistically significant interactions across recruitment intensity, BMI, baseline PaO2/FiO2, compliance, or baseline vasopressor use (all interaction P>0.2), supporting consistency of the harmful signal.
  • Timing:
    • Patients were enrolled early in ARDS (symptom onset within 2 weeks; not enrolled if eligibility met >72 hours), targeting a window where recruitment strategies might plausibly prevent progression.
  • Dose/intensity of intervention:
    • HFOV strategy used relatively high mean airway pressures (initiated at 30 cm H2O; day 1 mean 31.0 cm H2O), with concomitant increases in vasoactive support, consistent with a potentially haemodynamically stressful “dose”.
  • Outcome assessment:
    • Primary outcome (in-hospital mortality) is objective and unlikely to be misclassified; secondary outcomes included both objective outcomes (mortality) and more management-dependent endpoints (ventilator duration among survivors).
  • Statistical rigour and early stopping:
    • Trial stopped early for harm following DMC recommendation; early stopping can inflate effect sizes, but directionality (harm) was consistent across mortality endpoints.

Conclusion on Internal Validity: Overall, internal validity appears moderate-to-strong for the primary outcome given concealed central randomisation, objective mortality endpoints, and clear protocol separation; limitations arise from open-label care with differential co-interventions and early stopping, which may influence the magnitude (but not the direction) of the observed harm.

External Validity

  • Population representativeness:
    • Enrolled a severely ill, early moderate-to-severe ARDS cohort (APACHE II ~29; PaO2/FiO2 ~115–121) typical of tertiary ICUs.
    • Exclusions (e.g., severe chronic respiratory disease, high intracranial pressure risk, limitation of life support) may reduce applicability to some real-world ICU populations.
  • Applicability across settings:
    • Control arm was protocolised and relatively “high standard” (lung-protective ventilation plus a high-PEEP algorithm), which may not reflect all real-world practice at the time, potentially limiting transportability to less protocolised settings.
    • Intervention required HFOV-capable ventilators (SensorMedics 3100B) and experienced teams; generalisability is limited where HFOV expertise/equipment is unavailable.
    • Findings apply most directly to early routine HFOV using a high mean-airway-pressure open-lung strategy; they may not extrapolate to lower-pressure HFOV approaches (if used) or to paediatric/neonatal populations.

Conclusion on External Validity: Generalisability is moderate for early, invasively ventilated adult ARDS patients in well-resourced ICUs; applicability is more limited in resource-constrained environments and for HFOV protocols/devices that differ materially from OSCILLATE’s high mean-airway-pressure strategy.

Strengths & Limitations

  • Strengths:
    • Large, multicentre, international randomised trial targeting early moderate-to-severe ARDS.
    • Protocolised ventilation in both arms with explicit oxygenation/ventilation targets, improving intervention separation and interpretability.
    • Objective primary outcome with near-complete ascertainment in the randomised cohort.
    • Independent DMC with active safety monitoring and willingness to stop for harm.
  • Limitations:
    • Open-label design with clear differences in sedation, neuromuscular blockade, and vasoactive support that could mediate outcome differences.
    • Early stopping for harm may overestimate the magnitude of effect, although consistency across mortality endpoints supports a true harm signal.
    • Control arm crossover to HFOV (12%) introduces contamination, albeit largely as rescue in high-risk patients.
    • Results pertain to a specific “high mean airway pressure + recruitment” HFOV strategy and may not generalise to alternative HFOV approaches.

Interpretation & Why It Matters

  • Clinical meaning
    • Early routine HFOV (as delivered in OSCILLATE) should not be used in adult ARDS because it increased mortality compared with protocolised lung-protective conventional ventilation.
    • Improved oxygenation and fewer “oxygenation failure” events did not translate into survival benefit, reinforcing that ARDS trials must prioritise patient-centred outcomes.
  • Mechanistic insight
    • HFOV produced higher mean airway pressures and was associated with greater vasoactive support and deeper sedation/neuromuscular blockade, supporting haemodynamic compromise and downstream consequences as plausible mediators of harm.
  • Trial design lesson
    • OSCILLATE’s protocolised, high-quality conventional ventilation comparator strengthened causal inference: HFOV was tested against best practice rather than “usual care” that could be suboptimal.

Controversies & Other Evidence

  • Interpretation of harm signal (physiology vs outcomes):
    • Contemporaneous editorial commentary highlighted that, despite oxygenation improvements, HFOV in adult ARDS lacked evidence of benefit and that OSCILLATE demonstrated a credible signal of harm, shifting the field away from HFOV as an early strategy.1
  • Haemodynamic mechanism and co-interventions:
    • Correspondence focused on the plausibility that higher mean airway pressures and recruitment manoeuvres could worsen haemodynamics (necessitating more vasoactive support and sedation), and that these downstream effects may be central to explaining increased mortality rather than oxygenation failure per se.2
  • Contextualisation with OSCAR (parallel RCT):
    • OSCAR (published alongside OSCILLATE) found no mortality benefit of HFOV versus usual care, reinforcing lack of efficacy; differences in protocol intensity and comparator care between OSCILLATE and OSCAR remain central when interpreting whether harm is strategy-specific or context-dependent.3
  • Systematic reviews/meta-analyses after OSCILLATE/OSCAR:
    • Syntheses incorporating OSCILLATE and OSCAR generally show no mortality benefit for HFOV and a signal towards harm, largely driven by OSCILLATE’s excess mortality finding.45
  • Guideline impact:
    • Modern ARDS guidelines advise against routine HFOV in adult ARDS, reflecting the combined evidence base and the credible harm signal from OSCILLATE.67

Summary

  • OSCILLATE tested early HFOV with an open-lung, high mean-airway-pressure strategy against protocolised lung-protective conventional ventilation in early moderate-to-severe ARDS.
  • The trial was stopped early after 548 randomisations because mortality was higher in the HFOV arm.
  • In-hospital mortality was 47% with HFOV versus 35% with conventional ventilation (RR 1.33; 95% CI 1.09 to 1.64; P=0.005).
  • HFOV reduced refractory hypoxaemia (6% vs 12%) yet increased mortality, underscoring that improved oxygenation does not guarantee improved survival.
  • HFOV was associated with higher mean airway pressures and more vasoactive support and neuromuscular blockade, consistent with haemodynamic stress and co-intervention intensity as plausible contributors to harm.

Overall Takeaway

OSCILLATE is a landmark ARDS ventilation trial because it rigorously tested early HFOV against protocolised best-practice conventional ventilation and demonstrated increased mortality with HFOV, despite improved oxygenation. The trial strongly influenced practice and guidelines by showing that an “open-lung” high mean-airway-pressure HFOV strategy can cause net harm in adult ARDS, reinforcing the primacy of patient-centred outcomes over physiologic surrogates.

Overall Summary

  • Early routine HFOV (high mean airway pressure open-lung strategy) increased mortality in adult ARDS and is not recommended in modern practice.

Bibliography