A randomised controlled trial and cost effectiveness analysis of high frequency oscillatory ventilation against conventional artificial ventilation for adults with acute respiratory distress syndrome. The OSCAR (OSCillation in ARDS) study

Design: Unblinded randomised controlled clinical effectiveness study. Setting: Twelve large level 3 NHS general intensive care units. Target population: Adults (>16 years old) admitted to an intensive care unit with the acute respiratory distress syndrome (ARDS) predicted to require artificial ventilation for 48 hours or greater. ARDS will be defined using the American-European Consensus Committee definition of a PaO2/FiO2 ratio of less than 26.7kPa (Bernard GR et al. Intens Care Med 1994; 20: 225-32), from two blood gas analyses 12 hours apart. Exclusions are patients less than 35kg, and patients with obstructive lung pathologies due to a high risk of harm. Health technology being assessed: High frequency oscillatory (artificial) ventilation (HFOV) using a Sensormedics 3100B ventilator device (the only commercially available high frequency oscillatory ventilator) ventilating at a rate of 5-6 Hz (breaths/second) delivered by trained staff. As this will be a new technique to many centres, training will be provided to medical and nursing staff by Viasys (Sensormedics UK agents) free of charge and augmented with simulator training in critical care patient simulators in Oxford and Birmingham. The control group will receive pressure-controlled, volume-limited positive pressure ventilation with 6ml/kg tidal volume and fixed PEEP/FiO2 combinations as used in the "ARDSnet" study (The Acute Respiratory Distress Syndrome Network. New Engl J Med 2000; 342: 1301-8), the only ventilatory mode for patients with ARDS with proven benefit. Measurement of outcomes and cost: The primary efficacy outcome will be 30 day mortality (since this allows comparison with previous studies, and because most deaths in observational/interventional studies of ARDS occur within 30 days). Secondary efficacy outcomes will be mortality at ICU discharge, hospital discharge, and one year from study entry. Non-respiratory organ failure during intensive care treatment will be measured using the Critical Care Minimum Dataset definitions currently being rolled out to all ICU's. Health-related quality of life will be measured at six months and one year using self administered postal questionnaires (EQ-5D, the best measure in the ICU population. Smith S et al, Trent Research Scheme Final Report, 2003). Long-term lung damage will be assessed at one year using respiratory function questionnaires. Cognitive function will also be assessed, the exact method will be determined after a literature review. The primary measure of health care system benefit will be a within-trial cost effectiveness analysis (CEA) to estimate the cost per QALY gained for HFOV compared to conventional ventilation over 1 year. A long-term CEA will also be undertaken by constructing a decision analytic cost effectiveness model with a lifetime time horizon. The parameters for the CEA will be obtained via Bayesian synthesis of the trial evidence with evidence identified by the systematic review and a costing study in a guided sample of ICUs. The primary analyses will follow the guidance in the NICE guide to methods of health technology appraisal (A guide to the methods of Health Technology Appraisal. NICE, April 2004, London). Secondary analyses will adopt a full societal perspective. Data on non-public sector resource utilisation will be collected from the patients at 6 and 12 months. Sample size: The 30 day mortality rate in the pooled control groups of the two HFOV clinical trials to date is 31/97 (32%). The Intensive Care National Audit and Research Centre (ICNARC) identified a 42% hospital mortality in 76,900 patients with ARDS. The Scottish Intensive Care Society Audit Group (SICSAG) identified a 61% hospital mortality. Local Oxford data gave a 32% mortality for patients with ARDS. To detect a 9% absolute reduction with 80% certainty at p=0.05 requires 962 patients using an estimated 45% 30 day mortality in the control group.