Protocol. This trial protocol has been provided by the authors to give readers additional information about their work.

Protocol This trial protocol has been provided by the authors to give readers additional information about their work. Protocol for: Young D, Lamb SE, Shah S, et al. High-frequency oscillation for acute respiratory distress syndrome. N Engl J Med 2013;368:806-13. DOI: 10.1056/NEJMoa1215716

Supplemental information on protocols and statistical analysis plan for High frequency oscillation for acute respiratory distress syndrome, Young D, Lamb S, Shah S, MacKenzie I, Tunnicliffe W, Lall R, Rowan K, Cuthbertson B H, on behalf of the OSCAR collaborators. This supplement contains the following: Page Summary of changes in protocol 1 Original protocol 2 Final protocol 8 Statistical analysis plan 13 Summary of changes in protocol: The sample size was changed from 1006 to 802 after a planned interim review. Details are in the paper. The duration of the study was increased and so the end date was changed from November 2010 to July 2012. The secondary end-point, measurement of cognitive function at one year, was removed as we were unable to find a validated tool that could be used with postal questionnaires.

High Frequency OSCillation in ARDS A Collaborative Randomised Controlled Trial Comparing: Conventional positive pressure ventilation with high frequency oscillatory ventilation (HFOV) for adults with acute respiratory distress syndrome WORKING PROTOCOL Version 1 1 October 2008 ISRCTN10416500 REC reference numbers: 07/H0502/98 & 07/MRE00/73 OSCAR Trial, Kadoorie Centre for Critical Care Research and Education, John Radcliffe Hospital, Oxford OX3 9DU. Tel: 01865 220614. Email: OSCAR.trial@nda.ox.ac.uk Website: http://www.oscar-trial.org OSCAR Working Protocol Version 1-01 October 2008 Page 1 of 6

1 Background Whilst 18 randomised controlled trials of HFOV have been carried out in infants with respiratory distress syndrome, only two small RCTs in adult with acute respiratory distress syndrome (ARDS) were identified in a systematic review carried out in 2006. The inconclusive results of previous trials are largely the result of their small size. A much larger trial is required; the OSCAR Trial. OSCAR is funded by the Health Technology Assessment Programme, and Oxford University is the Sponsor. 2 The hypothesis Patients with ARDS who are treated with high frequency oscillatory ventilation (HFOV) will have a decreased mortality at 30 days (post randomisation) compared with patients treated with conventional positive pressure ventilation. 3 The intervention and control groups The study arms being compared in this trial are: Group 1: High frequency oscillatory ventilation (HFOV) versus Group 2: Conventional positive pressure ventilation The control group will receive conventional positive pressure ventilation using pressure-controlled artificial ventilation. The intervention is high frequency oscillatory ventilation (HFOV) delivered using a Vision Alpha. The management of artificial ventilation with HFOV will be based on a simple algorithm. Both groups will begin the treatment immediately following randomisation and the patients will remain on the ventilator until the start of weaning from artificial ventilation. 4 The setting Ten or more ICUs in the NHS in the United Kingdom able to care for Level 3 patients as defined by Comprehensive Critical Care. 5 Outcome measures Primary outcome measure: Secondary outcome measures: Mortality 30 days after randomisation Mortality rate at first discharge from ICU Mortality rate at first discharge from hospital Mortality rate one year after randomisation Non-pulmonary organ failures whilst treated on an intensive care unit Health-related quality of life six months after randomisation Health-related quality of life one year after randomisation Pulmonary function one year after randomisation Cognitive function one year after randomisation Ventilator-free days Antimicrobial-free days Sedative-free days OSCAR Working Protocol Version 1-01 October 2008 Page 2 of 6

6 Trial design OSCAR is a multicentre, open, randomised controlled trial. This is a pragmatic, effectiveness study, where the mode of ventilation is determined randomly, but all other treatment decisions are left to the clinicians managing the patient. 7 Population to be studied All adult intensive care patients requiring artificial ventilation who meet the inclusion criteria. 8 Eligible patients Patients are eligible for the trial if they meet the following inclusion criteria: i. Age 16 years ii. iii. iv. Weight 35 kg Receiving artificial ventilation via an endotracheal or tracheostomy tube Have acute hypoxaemic respiratory failure as defined by: Lowest recorded PaO 2 /FiO 2 ratio measured between onset of artificial ventilation and time of screening of 26.7 kpa with a positive end expiratory pressure (PEEP) 5 cm H 2 O Bilateral infiltrates on chest radiograph v. Will not be extubated by tomorrow evening (predicted by attending clinician) vi. Have been mechanically ventilated for LESS than 7 consecutive days ( 168 hours) at the point of randomisation. 9 Clinical management of the high frequency oscillatory ventilation arm As HFOV will not have been used previously in most of the intensive care units in the study, before the study starts the clinicians in these units will be trained to operate the HFOV ventilator and follow the treatment algorithm. 10 Clinical management of the conventional ventilation arm The control patients will be managed using the current best conventional ventilation strategy, which is limited tidal volume, pressure controlled artificial ventilation. 11 Patient exclusion criteria prior to trial entry Patients who could not benefit from HFOV: i. Patients with left atrial hypertension from any cause, diagnosed clinically or with echocardiography or pulmonary artery catheterisation. ii. Patients who have been mechanically ventilated for more than 7 consecutive days at the point of enrolment. Patients in whom HFOV might be hazardous: iii. Patients with moderate or severe airway disease expected to cause expiratory airflow limitation. iv. Patients who have had a lung biopsy or resection during this hospital admission. OSCAR Working Protocol Version 1-01 October 2008 Page 3 of 6

v. Patients with any other condition the clinician believes would make receiving HFOV hazardous. Administrative, practical and ethical exclusions: vi. Patients previously enrolled in the OSCAR trial. vii. Patients (or their representative*) who refuse consent. viii. Patients (or their representative*) who do not understand written or verbal information and for whom no interpreter is available. ix. Patients enrolled in another therapeutic trial in the 30 days prior to randomisation. x. Patients in whom active treatment has been withdrawn or withdrawal is planned. * Consultee (personal or nominated professional), in England and Wales; Welfare Guardian/Nearest Relative in Scotland. 12 Sample size 1006 patients (503 in each arm). 13 Planned recruitment period Recruitment started December 2007 and will continue until end November 2010. 14 Collaborating sites An Intensive Care Unit (ICU) will be considered for collaboration in the trial if it meets the following criteria: All consultants in the ICU have substantial uncertainty about the use of HFOV generally and would be prepared to enter patients into a trial comparing HFOV with conventional treatment for patients with ARDS. Consultants will attend HFOV training. The Principal Investigator will negotiate the release of all other appropriate staff for HFOV training. To ensure compliance with the trial protocol throughout a patient s stay, and to avoid cross-over after allocation, units will be signed up to the trial if they agree only to use the HFO ventilator for OSCAR patients. Centres that use the HFOV ventilator outside of the trial will have the ventilator removed. A local Principal Investigator will be identified who will ensure that locally: All staff comply with the protocol Patients in the ICU are considered promptly for the trial Trial case report forms and consent forms are completed in full The trial is conducted in accordance with the Research Governance Framework and Good Clinical Practice and fulfils all national and local regulatory requirements Access is given to source data for audit and verification. Training in the use of HFOV will be provided locally at the site or centrally in Oxford or Birmingham. 15 Ethics and local approvals National ethical approval has been obtained and covers all collaborating sites. Local approval applications for sites will be prepared by the trial office on behalf of the Principal Investigator for each site. These will be submitted to the relevant bodies and approval obtained. The Principal Investigator will be notified by the trial office when all approvals have been obtained. OSCAR Working Protocol Version 1-01 October 2008 Page 4 of 6

16 Good Clinical Practice and Research Governance The OSCAR trial does not fall under the EU Clinical Trials Directive (Directive 2001/20/EC) as it is not a medicinal product trial. It is therefore not required by law to work to ICH GCP although works to the principles outlined in ICH GCP. All HTA funded projects are expected to conduct their research in accordance with the Medical Research Council s Good Clinical Practice guidelines and the Department of Health s Research Governance Framework. 17 Publication of Results The success of the trial depends on the collaboration of nurses and doctors in the participating hospitals. Therefore chief credit for the study will be assigned to the collaborators from each participating centre and they will be named personally in the main publications. The results of the trial will be reported first to trial collaborators. Dissemination of results to patients will take place via the media, the website for the Intensive Care Society (http://www.ics.ac.uk/) and the trial website, and through relevant patient organisations. 18 Patient consent Patients will be unable to give informed consent due to alterations in conscious level caused by illness and therapeutic sedation. Consent will therefore be obtained in line with the legal requirements in England and Wales (Mental Capacity Act 2005), and in Scotland (Adults With Incapacity (Scotland) Act 2000). If a patient or their representative refuses consent the patient will receive the usual treatment as defined by the clinician responsible for the patient s care. 19 Formal trial entry and random allocation of patients Patients eligible for the trial will be randomised. Randomisation is carried out by telephoning a specialist randomisation service which is open 24-hours a day, 7 days a week. The call will take only a few minutes. Basic descriptive information will be requested and once these details have been supplied, the random allocation will be given in return. Stratification will be by recruiting ICU, age of patient and PaO 2 /FiO 2 ratio. 20 Patients not in the trial Brief details of patients initially eligible for the trial but not randomised will be recorded on a Why not in trial log at each collaborating unit. Recording this information is to establish an unbiased case selection and full reporting according to the CONSORT statement. 21 Proposed duration of treatment and weaning Patients will continue on HFOV until they have recovered sufficiently to be weaned from artificial ventilation when their FiO 2 is 0.4 or less, and the local clinician is satisfied that there are no nonpulmonary impediments to weaning. The HFOV ventilators do not allow any form of spontaneous (patient-triggered) ventilation which is normally required for weaning, so at this point the patients will be placed back on conventional ventilation and weaned according to local protocols using inspiratory pressure support. OSCAR Working Protocol Version 1-01 October 2008 Page 5 of 6

The point at which patients can be weaned from conventional artificial ventilation depends on a large number of factors that cannot be protocolised. 22 Serious adverse events (SAEs) A SAE is an untoward and unexpected occurrence that a research participant experiences which: i ii iii iv v Results in death Is life threatening Requires hospitalisation or prolongation of existing hospitalisation Results in persistent or significant disability or incapacity Consists of a congenital anomaly or birth defect. A SAE must be recorded on the appropriate trial form by the clinician caring for the patient and reported locally immediately to the Principal Investigator at that centre. The Principal Investigator will then report the SAE to the Chief Investigator of the OSCAR trial, Dr J D Young, within 3 working days of the event: Dr Duncan Young, Chief Investigator, OSCAR Trial Office, Kadoorie Centre for Critical Care Research and Education, John Radcliffe Hospital, Oxford OX3 9DU. Tel: 01865 857613, Fax: 01865 857611, Email: OSCAR.trial@nda.ox.ac.uk A confirmed, related, SAE will be submitted to a Main Research Ethics Committee within 15 days of the Chief Investigator becoming aware of the event using the NRES report of serious adverse event form. 23 Expected events Most known events related to artificial ventilation will occur equally in both groups. Exceptions that might occur more frequently in the HFOV group are: 1) Air trapping 2) Secondary effects of air trapping such as reduced carbon dioxide clearance. 24 Patient follow-up The trial office will send self-administered questionnaires to determine health related quality of life (EQ- 5D and SF-12 version 2) and specifically respiratory function (St George s Respiratory Questionnaire) to all survivors at six months and one year after randomisation. These questionnaires also include questions on social and health service use. To provide some estimate of quality of life beyond 12 months, patients recruited in the first year (December 2007 to end November 2008), will receive an additional EQ-5D questionnaire and questions concerning social and health service use at 24 and 30 months. The patients recruited in the second year, will received an additional EQ-5D questionnaire and questions concerning social and health service use at 18 months. This will allow us to model the time for health-related quality of life to return to population normal levels after ICU by group. 25 Data Analyses All analyses will be on an intention to treat basis. For the primary, and other dichotomous outcomes risk ratios and 95% confidence intervals will be calculated. Time to event outcomes such as duration of ventilation or duration of hospital stay will be analysed using survival methods and reported as hazard ratios and 95% confidence intervals. OSCAR Working Protocol Version 1-01 October 2008 Page 6 of 6

High Frequency OSCillation in ARDS A Collaborative Randomised Controlled Trial Comparing: Conventional positive pressure ventilation with high frequency oscillatory ventilation (HFOV) for adults with acute respiratory distress syndrome WORKING PROTOCOL Version 3 27 June 2011 ISRCTN10416500 REC reference numbers: 07/H0502/98 & 07/MRE00/73 OSCAR Trial, Kadoorie Centre for Critical Care Research and Education, John Radcliffe Hospital, Oxford OX3 9DU. Tel: 01865 220614. Email: OSCAR.trial@nda.ox.ac.uk Website: http://www.oscar-trial.org OSCAR Working Protocol Version 3 27 June 2011 Page 1 of 6

1 Background Whilst 18 randomised controlled trials of HFOV have been carried out in infants with respiratory distress syndrome, only two small RCTs in adult with acute respiratory distress syndrome (ARDS) were identified in a systematic review carried out in 2006. The inconclusive results of previous trials are largely the result of their small size. A much larger trial is required; the OSCAR Trial. OSCAR is funded by the Health Technology Assessment Programme, and Oxford University is the Sponsor. 2 The hypothesis Patients with ARDS who are treated with high frequency oscillatory ventilation (HFOV) will have a decreased mortality at 30 days (post randomisation) compared with patients treated with conventional positive pressure ventilation. 3 The intervention and control groups The study arms being compared in this trial are: Group 1: High frequency oscillatory ventilation (HFOV) versus Group 2: Conventional positive pressure ventilation The control group will receive conventional positive pressure ventilation using pressure-controlled artificial ventilation. The intervention is high frequency oscillatory ventilation (HFOV) delivered using a Vision Alpha. The management of artificial ventilation with HFOV will be based on a simple algorithm. Both groups will begin the treatment immediately following randomisation and the patients will remain on the ventilator until the start of weaning from artificial ventilation. 4 The setting Ten or more ICUs in the NHS in the United Kingdom able to care for Level 3 patients as defined by Comprehensive Critical Care. 5 Outcome measures Primary outcome measure: Secondary outcome measures: Mortality 30 days after randomisation Mortality rate at first discharge from ICU Mortality rate at first discharge from hospital Mortality rate one year after randomisation Non-pulmonary organ failures whilst treated on an intensive care unit Health-related quality of life six months after randomisation Health-related quality of life one year after randomisation Pulmonary function one year after randomisation Ventilator-free days Antimicrobial-free days Sedative-free days OSCAR Working Protocol Version 3 27 June 2011 Page 2 of 6

6 Trial design OSCAR is a multicentre, open, randomised controlled trial. This is a pragmatic, effectiveness study, where the mode of ventilation is determined randomly, but all other treatment decisions are left to the clinicians managing the patient. 7 Population to be studied All adult intensive care patients requiring artificial ventilation who meet the inclusion criteria. 8 Eligible patients Patients are eligible for the trial if they meet the following inclusion criteria: i. Age 16 years ii. iii. iv. Weight 35 kg Receiving artificial ventilation via an endotracheal or tracheostomy tube Have acute hypoxaemic respiratory failure as defined by: Lowest recorded PaO 2 /FiO 2 ratio measured between onset of artificial ventilation and time of screening of 26.7 kpa with a positive end expiratory pressure (PEEP) 5 cm H 2 O Bilateral infiltrates on chest radiograph v. Will not be extubated by tomorrow evening (predicted by attending clinician) vi. Have been mechanically ventilated for LESS than 7 consecutive days ( 168 hours) at the point of randomisation. 9 Clinical management of the high frequency oscillatory ventilation arm As HFOV will not have been used previously in most of the intensive care units in the study, before the study starts the clinicians in these units will be trained to operate the HFOV ventilator and follow the treatment algorithm. 10 Clinical management of the conventional ventilation arm The control patients will be managed using the current best conventional ventilation strategy, which is limited tidal volume, pressure controlled artificial ventilation. 11 Patient exclusion criteria prior to trial entry Patients who could not benefit from HFOV: i. Patients with left atrial hypertension from any cause, diagnosed clinically or with echocardiography or pulmonary artery catheterisation. ii. Patients who have been mechanically ventilated for more than 7 consecutive days at the point of enrolment. Patients in whom HFOV might be hazardous: iii. Patients with moderate or severe airway disease expected to cause expiratory airflow limitation. iv. Patients who have had a lung biopsy or resection during this hospital admission. v. Patients with any other condition the clinician believes would make receiving HFOV hazardous. OSCAR Working Protocol Version 3 27 June 2011 Page 3 of 6

Administrative, practical and ethical exclusions: vi. Patients previously enrolled in the OSCAR trial. vii. Patients (or their representative*) who refuse consent. viii. Patients (or their representative*) who do not understand written or verbal information and for whom no interpreter is available. ix. Patients enrolled in another therapeutic trial in the 30 days prior to randomisation. x. Patients in whom active treatment has been withdrawn or withdrawal is planned. * Consultee (personal or nominated professional), in England and Wales; Welfare Guardian/Nearest Relative in Scotland. 12 Sample size 802 patients (401 in each arm). 13 Planned recruitment period Recruitment started December 2007 and will continue until end July 2012. 14 Collaborating sites An Intensive Care Unit (ICU) will be considered for collaboration in the trial if it meets the following criteria: All consultants in the ICU have substantial uncertainty about the use of HFOV generally and would be prepared to enter patients into a trial comparing HFOV with conventional treatment for patients with ARDS. Consultants will attend HFOV training. The Principal Investigator will negotiate the release of all other appropriate staff for HFOV training. To ensure compliance with the trial protocol throughout a patient s stay, and to avoid cross-over after allocation, units will be signed up to the trial if they agree only to use the HFO ventilator for OSCAR patients. Centres that use the HFOV ventilator outside of the trial will have the ventilator removed. A local Principal Investigator will be identified who will ensure that locally: All staff comply with the protocol Patients in the ICU are considered promptly for the trial Trial case report forms and consent forms are completed in full The trial is conducted in accordance with the Research Governance Framework and Good Clinical Practice and fulfils all national and local regulatory requirements Access is given to source data for audit and verification. Training in the use of HFOV will be provided locally at the site or centrally in Oxford or Birmingham. 15 Ethics and local approvals National ethical approval has been obtained and covers all collaborating sites. Local approval applications for sites will be prepared by the trial office on behalf of the Principal Investigator for each site. These will be submitted to the relevant bodies and approval obtained. The Principal Investigator will be notified by the trial office when all approvals have been obtained. OSCAR Working Protocol Version 3 27 June 2011 Page 4 of 6

16 Good Clinical Practice and Research Governance The OSCAR trial does not fall under the EU Clinical Trials Directive (Directive 2001/20/EC) as it is not a medicinal product trial. It is therefore not required by law to work to ICH GCP although works to the principles outlined in ICH GCP. All HTA funded projects are expected to conduct their research in accordance with the Medical Research Council s Good Clinical Practice guidelines and the Department of Health s Research Governance Framework. 17 Publication of Results The success of the trial depends on the collaboration of nurses and doctors in the participating hospitals. Therefore chief credit for the study will be assigned to the collaborators from each participating centre and they will be named personally in the main publications. The results of the trial will be reported first to trial collaborators. Dissemination of results to patients will take place via the media, the website for the Intensive Care Society (http://www.ics.ac.uk/) and the trial website, and through relevant patient organisations. 18 Patient consent Patients will be unable to give informed consent due to alterations in conscious level caused by illness and therapeutic sedation. Consent will therefore be obtained in line with the legal requirements in England and Wales (Mental Capacity Act 2005), and in Scotland (Adults With Incapacity (Scotland) Act 2000). If a patient or their representative refuses consent the patient will receive the usual treatment as defined by the clinician responsible for the patient s care. 19 Formal trial entry and random allocation of patients Patients eligible for the trial will be randomised. Randomisation is carried out by telephoning a specialist randomisation service which is open 24-hours a day, 7 days a week. The call will take only a few minutes. Basic descriptive information will be requested and once these details have been supplied, the random allocation will be given in return. Stratification will be by recruiting ICU, age of patient and PaO 2 /FiO 2 ratio. 20 Patients not in the trial Brief details of patients initially eligible for the trial but not randomised will be recorded on a Why not in trial log at each collaborating unit. Recording this information is to establish an unbiased case selection and full reporting according to the CONSORT statement. 21 Proposed duration of treatment and weaning Patients will continue on HFOV until they have recovered sufficiently to be weaned from artificial ventilation when their FiO 2 is 0.4 or less, and the local clinician is satisfied that there are no nonpulmonary impediments to weaning. The HFOV ventilators do not allow any form of spontaneous (patient-triggered) ventilation which is normally required for weaning, so at this point the patients will be placed back on conventional ventilation and weaned according to local protocols using inspiratory pressure support. OSCAR Working Protocol Version 3 27 June 2011 Page 5 of 6

The point at which patients can be weaned from conventional artificial ventilation depends on a large number of factors that cannot be protocolised. 22 Serious adverse events (SAEs) A SAE is an untoward and unexpected occurrence that a research participant experiences which: i ii iii iv v Results in death Is life threatening Requires hospitalisation or prolongation of existing hospitalisation Results in persistent or significant disability or incapacity Consists of a congenital anomaly or birth defect. A SAE must be recorded on the appropriate trial form by the clinician caring for the patient and reported locally immediately to the Principal Investigator at that centre. The Principal Investigator will then report the SAE to the Chief Investigator of the OSCAR trial, Dr J D Young, within 3 working days of the event: Dr Duncan Young, Chief Investigator, OSCAR Trial Office, Kadoorie Centre for Critical Care Research and Education, John Radcliffe Hospital, Oxford OX3 9DU. Tel: 01865 857613, Fax: 01865 857611, Email: OSCAR.trial@nda.ox.ac.uk A confirmed, related, SAE will be submitted to a Main Research Ethics Committee within 15 days of the Chief Investigator becoming aware of the event using the NRES report of serious adverse event form. 23 Expected events Most known events related to artificial ventilation will occur equally in both groups. Exceptions that might occur more frequently in the HFOV group are: 1) Air trapping 2) Secondary effects of air trapping such as reduced carbon dioxide clearance. 24 Patient follow-up The trial office may send self-administered questionnaires to determine health related quality of life (EQ- 5D and SF-12 version 2) and specifically respiratory function (St George s Respiratory Questionnaire) to all survivors at six months and one year after randomisation. These questionnaires also include questions on social and health service use. To provide some estimate of quality of life beyond 12 months, patients recruited in the first year (December 2007 to end November 2008), will receive an additional EQ-5D questionnaire and questions concerning social and health service use at 24 and 30 months. The patients recruited in the second year, will received an additional EQ-5D questionnaire and questions concerning social and health service use at 18 months. This will allow us to model the time for health-related quality of life to return to population normal levels after ICU by group. 25 Data Analyses All analyses will be on an intention to treat basis. For the primary, and other dichotomous outcomes risk ratios and 95% confidence intervals will be calculated. Time to event outcomes such as duration of ventilation or duration of hospital stay will be analysed using survival methods and reported as hazard ratios and 95% confidence intervals. OSCAR Working Protocol Version 3 27 June 2011 Page 6 of 6

High Frequency OSCillation in ARDS STATISTICAL ANALYSIS PLAN FOR THE OSCAR TRIAL Author: Ranjit Lall Completion date: January 2010 Approved by: Trial Steering Committee Data Monitoring Ethics Committee O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 2 of 56 CONTENT Page 1. Section 1: Design and aims of the trial 3 2. Section 2: Statistical Monitoring of the trial 5 3. Section 3: Formal Interim analysis 9 4. Section 4: Analysis Populations 13 5. Section 5: Analysis Datasets 13 6. Section 6: Non-adherence to protocol 14 7. Section 7: Withdrawals 14 8. Section 8: Explanatory and Outcome variables 15 9. Section 9: Statistical Analysis and presentation of the results 22 Template tables and figures 26 Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 3 of 56 SECTION 1: DESIGN AND AIMS OF THE TRIAL 1.1 Trial design OSCAR is a UK multicentre, open randomised controlled trial. 1.2 The hypothesis It is hypothesised that patients with acute respiratory distress syndrome (ARDS) who are treated with high frequency ventilation (HFOV) will have a decreased mortality at 30 days (post randomisation) compared with patients treated with conventional positive pressure ventilation. The following aspects of the trial will be assessed. 1.3 Clinical Effectiveness The clinical effectiveness of conventional positive pressure ventilation versus high frequency oscillatory ventilation (HFOV) will be assessed using mortality as an endpoint at 30 days (after randomisation). 1.4 Quality of Life The quality of life of patients will be assessed at 6 and 12 months after randomisation and will be compared for the two types of ventilation treatment. 1.5 Harm (Safety) data Serious Adverse Events associated with either ventilation treatment will be compared. Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 4 of 56 This statistical analysis plan is divided into two main parts: the monitoring of the trial and the main statistical analysis. Section 2 and 3 detail the monitoring aspects of the trial and sections 4 to 9 detail the final statistical analysis which will be carried out when all the patients have been recruited and the database is complete, clean and frozen. Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 5 of 56 MONITORING OF THE TRIAL Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 6 of 56 SECTION 2: STATISTICAL MONITORING OF THE TRIAL Monitoring of the trial is a continual process, from the start to the end of the trial. The objectives of the statistical input during trial monitoring are to: overview the recruitment and follow-up examine the quality of the data ensure the protocol is being followed assess the randomisation sequence conduct formal interim analyses early detection of issues that could introduce bias so that corrective action can be taken (e.g. missing data, losses to follow-up, non-compliance, fraud). The Data Monitoring and Ethics Committee (DMEC) is given the responsibility of monitoring the accumulating data. Statistical reports will be produced to cover the following that provide oversight on the quality of the trial. 2.1 Recruitment and follow-up 2.1.1 Overall recruitment i. The recruitment in the trial will be summarised. Recruitment is continuously assessed by checking whether actual accrual is meeting projected targets, overall and by each centre. ii. iii. The number and type of patients who do not enter the trial will be summarised. The number of patients (deaths/survivors) entering the trial will be summarised at each stage- Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 7 of 56 from admission to hospital to randomisation randomised patients in ICU; randomised patients and discharged from ICU up to end of discharge from hospital; during follow-up 2.1.2 Follow-up rates The follow-up rates are based on the return of questionnaires at follow-up. %Sent rate = No.of patients sent Q No.who are alive and were randomised x 100 %Return rate = No.of patients who return Q (not blank) No.who are alive and a Q sent out x 100 The follow-up rates will be computed at the following time-points: Follow-up at 6 months Follow-up at 12 months The 6 and 12 months response rates will be computed on all randomised patients. The template table (Table A) given in Template tables and figures section will be used to display the follow-up rates. Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 8 of 56 2.2 Quality of Data The quality of the data in terms of accuracy, completeness, patterns of missing data and number of missing observations, out-of-range values and checking that dates are consecutive from hospital admission through to ICU, hospital discharge and follow-up will be summarised to ensure that the database remains clean and accurate. 2.3 Ensuring the trial is conducted according to the protocol Patients who do not adhere to the protocol and those who withdraw during the trial will be listed. Section 6 defines non-adherence to the protocol and section 7 lists the reasons for withdrawal. 2.4 Randomisation The randomisation procedure will be assessed to ensure that the allocation is correctly stratifying patients in the block strata (centre, age and PF ratio) over the two ventilation treatments. 2.5 Statistical Reports The summaries on the monitoring of the trial in the statistical reports which will be provided for the DMEC, will not give any details of ventilation allocation. Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 9 of 56 SECTION 3: FORMAL INTERIM ANALYSES The DMEC is responsible for assuring that OSCAR trial patients are not exposed to unnecessary or unreasonable risk, and that the trial is being conducted according to high scientific and ethical standards. This section details the following: Frequency of the interim analyses Stopping guidelines Analyses required for monitoring purposes 3.1 Frequency of the Interim analyses At the request of the DMEC: Two definite equally spaced formal interim analyses will be carried out (one approximately 1/3 of the way and another 2/3 of the way through the trial). Thus, the first interim is planned when there are approximately 270 randomised patients in the trial and the second will be carried out when there are approximately 530 randomised patients (taking account of the new revised sample size of 802). If any of the stopping guidelines (detailed below) are met at the first interim analysis, then the DMEC will decide whether a closer monitoring of the trial is required. In this case, another interim analysis may be planned ½ way in between the two interims (i.e. when approximately 400 have been randomised). At the conclusion of each meeting, the DMEC will make recommendations (formally to the TSC) as to whether the: Evidence from other trials that shows the intervention is ineffective or harmful; Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 10 of 56 Study should continue without any modifications; Study should continue with the recommended modifications; Study should be stopped for safety/efficacy. 3.2 Stopping guidelines It is not appropriate to specify fixed stopping guidelines, but the p-values given below are likely to be used as guidance for the primary end-point, to justify stopping or modifying the trial prematurely. Table B below illustrates the p-values if we look at the data 3 times (at equal time intervals), or if we look at the data 4 times (with unequal time spacing). The reasons for stopping/modifying the trial will be if: Treatments are convincingly different in terms of mortality at 30 days; There are an unacceptable number of Serious Adverse Events. A Serious Adverse event is an untoward or unexpected occurrence that a research participant experiences which is (a) life threatening; (b) requires hospitalisation or prolongation of existing hospitalisation; (c) results in persistent or significant disability or incapacity or/and (d) consists of a congential anomaly. The main criterion for early stopping of the trial by the TSC upon suggestions from the DMEC will be that evidence from the trial and from other sources suggests (a) proof beyond all reasonable doubt that for all, or for some type of patients, the use of one of the ventilation treatments is clearly indicated or contra-indicated, and (b) evidence that might reasonably be expected to influence routine clinical practice. After the 1 st interim, and as the data matures, the DMEC may decide to modify or institute its own monitoring boundaries and stopping rules. For the primary end-point, the DMEC will decide whether the stopping boundaries are appropriate at the end of each closed session in the meeting, for the next review. Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 11 of 56 Table B: Stopping guidelines for the planned interim analyses Number of planned interim analysis Interim analysis O Brien-Fleming Alpha spending function (Lan- DeMets) 3 1 0.00021-2 0.01210-3 (final) 0.05000-4 1-0.00019 2-0.00305 3-0.01235 4 (final) - 0.05000 3.3 Analyses required for monitoring purposes The analyses required for monitoring the trial will be agreed with the trial statistician and the members of the DMEC prior to the meeting. The analyses will be produced with ventilation treatment type remaining blind and the data will only un-blinded if requested by the DMEC. Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 12 of 56 MAIN STATISTICAL ANALYSIS Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 13 of 56 SECTION 4 : ANALYSIS POPULATIONS Intention to treat Population All analyses will be based on Intention- to-treat (ITT). The patients will be analysed according to the ventilation treatment they were randomised to, irrespective of the ventilation treatment they received. All patients will be included in the analysis, regardless of whether they have or have not adhered to the protocol (as detailed in section 6). SECTION 5: ANALYSIS DATASETS There will be two datasets used for the statistical analysis: (a) observed and (b) imputed. 5.1 Observed dataset This will comprise of all the data observed (including follow-up) with missing values. The observed (intention-to-treat) dataset will be used to report the main results of the trial. 5.2 Imputed dataset Multiple imputation methods will be used to impute missing data. Prior to an imputation, the data mechanisms (MAR - missing at random; NMAR - not missing at random; MCAR - missing completely at random) will be assessed to make sure that multiple imputation is viable. Only data that is validly missing will be imputed. In the case of multivariate normal data, the multiple imputation methods assuming normality will be used. In the case where one cannot assume a distribution of the data, the ICE (imputation by chain equations- Royston, 2005) will be used. Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 14 of 56 The imputed dataset will be used for sensitivity analyses. SECTION 6: NON-ADHERENCE TO PROTOCOL There will be some patients who are likely not to adhere to the protocol or depart from the intended ventilation treatment and/or evaluation. The following will be considered as non-adherence to the protocol: Concerns with ventilator: patient taken off ventilator for any undue reason (e.g. staff not confident with oscillator, clinical concerns, ventilation problems); Withdrawals: (see below); Ineligible patients: Any patients who were ineligible but were subsequently randomised into the trial; Patients who switch ventilation treatment: Any patients who cannot comply to the randomised ventilation arm or/and are switched ventilation treatment; Incomplete follow-up: Any patients with incomplete follow-up data (at month 6 and 12 (all patients). SECTION 7: WITHDRAWALS All withdrawals from ventilation only, use of ICU data only, use of follow up data only and full withdrawals from the study will be listed and summarised by ventilation treatment. Reason for withdrawal will be specified. For this trial withdrawals will be listed as follows: Withdrawals in ICU After ICU discharge to 6 months follow-up After 6 months follow-up but before 12 month follow-up Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 15 of 56 SECTION 8: EXPLORATORY AND OUTCOME VARIABLES The following outcomes and variables will be collected in this study. 8.1 Characteristics Of Patients and Baseline Time-point Pre-randomisation Variable Exhaled minute volume Total respiratory rate Peep Plateau pressure Arterial blood gas PaO 2 Arterial blood gas PaC0 2 Arterial blood gas PaO 2 Arterial blood gas ph Randomisation Immediately postrandomisation Arterial blood gas Fi0 2 Support/organ monitoring Age Sex Pa0 2 /Fi0 2 ratio Weight Height Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 16 of 56 Statistical Analysis Plan 8.2 Primary, Secondary and Safety Outcomes Cognitive function at one year after randomisation has been removed after ethical approval. After an extensive literature review, no questionnaire was found which could be used to adequately record cognitive function at follow-up time. Type of Time-point Outcome Derived Outcome Primary Postrandomisation Mortality (all cause) 30 days Status of the patient at 30 days (dead/alive) (as recorded on Death Notification sheet/survivor sheet) Safety Throughout the Serious Adverse events trial Secondary ICU Mortality rate at first discharge from ICU Mortality rate from randomisation to first ICU discharge: As recorded by Q.36: Section 7 of the CRF: ICU Discharge Intensive care length of stay Discharged from ICU/death date (date and time recorded on Q.36 Section 7 of the CRF: ICU Discharge sheet) Date of randomisation (date and time recorded on Q.36- Randomisation Form) Number of ventilator free days (up to No. of days free of advanced respiratory support up to day 30, ICU discharge if prior to day 30 or death 30 days) (No. of days recorded as no on advanced ventilation from ICU entry up to day 30 or up to ICU discharge (if prior to day 30) or death (Section 3 of the CRF: Daily data All Patients). Advanced respiratory support free No. of no advanced respiratory support from ICU entry to ICU discharge (Section 3 of the CRF: Daily days data All Patients). Basic respiratory support free days No. of no basic respiratory support from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). Number of days on renal support No. of yes on renal support from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 17 of 56 Type of Time-point Outcome Derived Outcome Secondary ICU Number of days on gastrointestinal support No. of yes on gastrointestinal support from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). Number of days on dermatological support No. of yes on dermatological support from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). Number of days on liver support No. of yes on liver support from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). Exhaled minute volume Total respiratory rate Peep Plateau pressure Arterial blood gas PaO 2 Arterial blood gas PaC0 2 Arterial blood gas ph Hospital discharge Follow-up (6 months after randomisation) Arterial blood gas Fi0 2 Mortality rate at first discharge from hospital Length of acute hospital stay SF-12 Health survey questionnaire (version 2) Mortality rate from randomisation to first hospital discharge: As recorded in CRF: Hospital Discharge Date of discharge from hospital/death (date and time recorded on Hospital Discharge form) Date of randomisation (date and time recorded on Q.36- Randomisation Form) Scoring of the SF-12 will be carried out using the SF-12v2- How to Score version 2 of the SF- 12 Health Survey manual (Ware et al.1996) Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 18 of 56 Type of Time-point Outcome Derived Outcome Secondary Follow-up (12 Mortality rate one year after (as recorded on Death Notification sheet/survivor sheet) months after randomisation) randomisation Respiratory Function (St. George s This will be scored using the manual for the respiratory questionnaire. Respiratory Questionnaire). The EQ-5D scale Evaluates patients QOL based on 5 dimensions converted into a single summary score (range:0 (death) to 1 (perfect health state)).the ED-5D will be scored using the devised algorithm (EuroQol Group, 1990) and summarised as detailed in the User Guide (EuroQol Group, 2005). Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 19 of 56 8.3 Tertiary Outcomes and Process Variables Type of outcome/ variable Time-point Outcome/variable Derived Tertiary ICU Apache II score Apache II score will be derived using the specified algorithm (Knaus et al., 1985). Number of days antimicrobial use: received in past 24 hours Number of days antimicrobial use: treating pulmonary infection Number of days antimicrobial use: given intravenously Number of days for sedative use: received in past 24 hours Number of days for sedative use: given as an intravenous bolus dose Number of days for sedative use: given by infusion Number of days for sedative use: more than 1 class of sedative Number of days for sedative use: more than 2 class of sedative Number of days for sedative use: more than 3 class of sedative No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 20 of 56 Type of Outcome Time-point Outcome Derived Tertiary ICU Number of days for muscle relaxants use: received in past 24 hours Number of days for muscle relaxants use: given as an intravenous bolus dose Number of days for muscle relaxants use: given by infusion Number of days prone in last 24 hours Number of days patient received inhaled nitric oxide in past 24 hours Arterial blood gases: Pa0 2 Arterial blood gases: PaC0 2 Arterial blood gases: ph Arterial blood gases: Fi0 2 Process Hospital ad./icu Time from hospital admission to ICU Hospital ad. /randomisation admission Time from hospital admission to randomisation No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). No. of yes on this variable from ICU entry to ICU discharge (Section 3 of the CRF: Daily data All Patients). Time (date and time) of ICU admission (as recorded on Section 2 of the CRF: First 24 hours in ICU) - Time (date and time) of hospital admission (as recorded on Form 2: Section 1 of the CRF: data required immediately after post-randomisation) Time (date and time) of randomisation as recorded on the Randomisation Form - Time (date and time) of hospital admission (as recorded on Form 2: Section 1 of the CRF: data required immediately after post-randomisation) ICU/randomisation Time from ICU to randomisation Time (date and time) of randomisation (as recorded on Randomisation Form) - time (date and time) of ICU admission (as recorded on Section 2 of the CRF: First 24 hours in ICU). Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 21 of 56 Type of Time-point Outcome Derived Outcome Process ICU Exhaled minute volume (conventional ventilation only) Total respiratory rate volume (conventional ventilation only) Peep volume (conventional ventilation only) Plateau Pressure volume (conventional ventilation only) HFOV settings on ventilation Clinical Trial O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx

Page 22 of 56 Statistical Analysis Plan SECTION 9: STATISTICAL ANALYSIS AND PRESENTATION OF THE RESULTS All statistical tests will be two-sided and performed at the 5% significance level. The statistical analyses will be carried out using SAS (version 9.13). The statistical report for the study will follow the guidelines set by CONSORT (2001). The report will detail the design, conduct, analysis and results of the data. This report will be supported by the tables, plots and flow diagrams that have been included at the end of this section. 9.1 Recruitment throughout the trial The CONSORT diagram will illustrate the flow of patients throughout the trial. Tables 1 to 6 illustrates the number (and percentage) of patients in the trial. The total number of patients in the trial who withdrew, had non-response and survived will be compared between ventilation treatment using the chi-squared test statistics or Fisher s exact test, with odds ratios and 95% confidence intervals. 9.2 Harm data (Serious adverse events/adverse events) Table 7 illustrates the serious adverse events/adverse events. The total number (and percentage) of serious adverse events and adverse events in the trial will be compared between the two ventilation treatments using the chi-squared statistics or Fisher s exact test, with odds ratios and 95% confidence intervals. Adjusted analyses will not be performed for any harm data. 9.3 Patients not recruited into the trial The number and percentage of patients who are not in the trial will be summarised by each centre (Tables 8 and 9 below). O:\08 - Ranjit - Statistician\2010 Ranjit\Analysis Plan OSCAR trial-january_2010.docx