BMJ

Confidential: For Review Only A Pragmatic, Cluster-Randomised Cross- Over Comparison of Two Oxygen Protocols in Patients with a Suspected Acute Coronary Syndrome

Journal: BMJ

Manuscript ID BMJ-2020-060486.R1

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the 08-Oct-2020 Author:

Complete List of Authors: Stewart, Ralph; City , Green Lane Cardiovascular Service; The Jones, Peter; , and Department of Surgery, Emergency Medicine Department; The University of Auckland Dicker, Bridget; St John, Clinical Audit and Research; Auckland University of Technology, Paramedicine Jiang, Yannan; The University of Auckland National Institute for Health Innovation, Department of Statistics Smith, Tony; St John Ambulance, Clinical Audit and Research Swain, Andrew; Wellington Free Ambulance, Paramedicine; Auckland University of Technology, Paramedicine Kerr, Andrew; , Department of ; The University of Auckland, Section of Epidemiology and Biostatistics Scott, Tony; , Department of Cardiology Smyth, David; Canterbury District Health Board, Department of Cardiology Ranchord, Anil; Capital and Coast District Health Board, Department of Cardiology Edmond, John; Southern District Health Board; Faculty of Medicine Than, Martin; Hospital, Emergency Medicine Webster, Mark; Auckland City Hospital, Green Lane Cardiovascular Service; The University of Auckland White, Harvey; Auckland City Hospital, Green Lane Cardiovascular Service; The University of Auckland Devlin, Gerard; Hauroa Tairāwhiti, Gisborne and Heart Foundation of

Keywords: Acute Coronary Syndrome, Oxygen, Pragmatic Clinical Trial

https://mc.manuscriptcentral.com/bmj Page 1 of 49 BMJ

1 2 3 4 5 6 7 8 9 10 11 Confidential: For Review Only 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 2 of 49

1 2 3 1 A Pragmatic, Cluster-Randomised Cross- Over Comparison of Two Oxygen 4 5 6 2 Protocols in Patients with a Suspected Acute Coronary Syndrome 7 8 3 9 10 4 Ralph A H Stewart M.D., Peter Jones MB ChB, Bridget Dicker PhD, Yannan Jiang PhD, 11 12 5 Tony SmithConfidential: MB ChB, Andrew Swain PhD, Andrew For Kerr Review M.D., Tony Scott MB Only ChB, David Smyth M.D., 13 14 6 Anil Ranchord MB ChB, John Edmond M.D., Martin Than MBBS, Mark Webster MB ChB, 15 16 7 Harvey D White DSc, Gerard Devlin M.D 17 18 8 19 20 9 Green Lane Cardiovascular Services Auckland City Hospital (R.A.H.S, M.W., H.D.W.) 21 22 10 and University of Auckland (R.A.H.S., M.W., H.D.W.); Emergency Medicine Research, Auckland City 23 24 11 Hospital, and Department of Surgery, University of Auckland, (P.J.); St John Auckland & 25 26 12 Paramedicine Department, Auckland University of Technology (B.D.); National Institute for Health 27 28 13 Innovation, University of Auckland, Auckland (Y.J.); St John Ambulance, New Zealand (T.S.); 29 30 14 Wellington Free Ambulance, Wellington (A.S.); Department of Cardiology, Middlemore Hospital, 31 15 and Section of Epidemiology and Biostatistics, University of Auckland (A.K.); Cardiology Department, 32 33 16 Northshore Hospital (T.S.); Canterbury District Health Board, Christchurch (D.S.); 34 35 17 Cardiology Department, Capital & Coast District Health Board, Wellington Hospital (A.R.); 36 37 18 Southern District Health Board, Dunedin and Dunedin School of Medicine, University of Otago, 38 39 19 Dunedin (J.E.); Department of Emergency Medicine, (M.T.); 40 41 20 Hauroa Tairāwhiti, Gisborne and Heart Foundation of New Zealand (G.D.) 42 43 21 44 45 22 CORRESPONDING AUTHOR 46 47 23 Professor Ralph A.H. Stewart 48 49 24 Green Lane Cardiovascular Service, Auckland City Hospital 50 51 25 Private Bag 92024, 52 53 26 Auckland 1030, New Zealand 54 55 27 Phone +64-9-3074949 Ext 23668 56 57 28 Email [email protected] 58 59 29 Word count: 4476 (excluding title page, abstract, references, figures and tables) 60

1 | P a g e https://mc.manuscriptcentral.com/bmj Page 3 of 49 BMJ

1 2 3 1 ABSTRACT 4 5 2 6 7 3 Objective: To determine whether high flow supplementary oxygen influences 30 day mortality in 8 9 4 patients presenting with a suspected acute coronary syndrome (ACS). 10 11 5 Design: Pragmatic, cluster-randomized, cross-over trial. 12 Confidential: For Review Only 13 6 Setting: Two oxygen protocols were used as part of routine care in patients with suspected ACS in 14 15 7 ambulances and throughout New Zealand. All four New Zealand geographic regions were 16 17 8 randomly allocated to each oxygen protocol in 6 month blocks over 2 years. 18 19 9 Participants: 40,872 patients with suspected ACS included in the All NZ ACS-Quality Improvement 20 21 10 register or ambulance ACS pathway during the study periods, 4159 (10%) had a final diagnosis of 22 23 11 STEMI and 10,218 (25%) non-STEMI. 24 25 12 Interventions: The “high” oxygen protocol recommended oxygen at 6-8 litres/minute by face mask for 26 13 ischaemic symptoms or electrocardiographic changes, irrespective of the oxygen saturation (SpO ). 27 2 28 14 The “low” oxygen protocol recommended oxygen only if SpO was <90%, with a target SpO <95%. 29 2 2 30 15 Main outcome measure: 30 day all cause mortality determined from linkage to administrative data. 31 32 16 Results: Demographic and clinical characteristics of patients managed under the high (n=20,304) 33 34 17 and the low oxygen (n=20,568) protocols were well matched. For all patients with suspected ACS 30 35 36 18 day mortality for the high and low oxygen groups respectively was 613 (3.0%) versus 642 (3.1%), 37 38 19 odds ratio [OR] 0.97, 95% CI 0.86, 1.08. In patients with STEMI 30 day mortality for the high and low 39 40 20 protocols was 178 (8.8%) and 225 (10.6%) respectively, OR 0.81, 95% CI 0.66, 1.00, and for patients 41 42 21 with non-STEMI high 187 (3.6%) and low 176 (3.5%), OR 1.05, 95% CI 0.85, 1.29. 43 44 22 Conclusion: High flow oxygen did not increase or decrease 30 day mortality in a large patient cohort 45 46 23 presenting with suspected ACS. 47 48 24 49 25 Trial Registration: ANZ Clinical Trials Website (ACTRN12616000461493); 50 51 26 https://www.anzctr.org.au/Default.aspx 52 53 27 Keywords: acute coronary syndromes, oxygen, pragmatic clinical trial 54 55 28 Abstract Word Count: 276 56 57 58 59 60

2 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 4 of 49

1 2 3 1 INTRODUCTION 4 5 2 Oxygen has been given to patients with acute myocardial infarction for over 50 years, despite limited 6 7 3 evidence that this improves outcomes.(1) Supplementary oxygen can correct or reduce hypoxemia, 8 9 4 which is common in patients with acute coronary syndromes. However there is some evidence that 10 11 5 arterial oxygen saturation levels above normal could be harmful by causing coronary vasoconstriction 12 Confidential: For Review Only 13 6 or increasing oxidative stress.(1, 2) The possibility of harm from supplementary oxygen was 14 15 7 supported by a meta-analysis of clinical trials which compared liberal with conservative oxygen 16 17 8 strategies in predominantly normoxemic patients with a range of critical medical conditions.(3) The 18 19 9 only previous large, randomised trial in patients with suspected myocardial infarction, ‘DETO2X-AMI’, 20 21 10 reported no difference in one year mortality in patients given 12 hours of high flow oxygen, compared 22 23 11 to limited oxygen.(4) Current clinical practice guidelines recommend that oxygen is not given to 24 25 12 patients with ST elevation myocardial infarction (STEMI) or non-STEMI who are not hypoxemic.(5-9) 26 13 27 28 14 There are limitations with the available evidence. First, patients included in randomised trials 29 30 15 generally have a lower risk and may not be representative of all patients managed in usual care. 31 32 16 Second, all previous trials including DETO2X-AMI, had insufficient power to identify a small, but 33 34 17 clinically relevant, 1-2% absolute difference in mortality with oxygen treatment. Third, most patients 35 36 18 included in previous trials had normal oxygen saturation levels. A benefit from oxygen may require the 37 38 19 presence of hypoxemia. The saturation threshold for starting oxygen (e.g. <95% or <90%), and the 39 40 20 target oxygen saturation level when on oxygen, are uncertain. Fourth, it is possible the effects of 41 42 21 oxygen depend upon the diagnosis. For example, patients having an acute STEMI typically have 43 44 22 prolonged and severe myocardial ischemia, and may show greater benefit than those with other 45 46 23 conditions. 47 48 24 49 25 Recommendations for treating hypoxaemia would be better informed if there was clear evidence on 50 51 26 whether there is harm from high flow oxygen, suggesting that high SpO need to be avoided, or 52 2 53 27 evidence there was benefit from giving oxygen to correct modest reductions in SpO . To determine 54 2 55 28 whether and when oxygen may be indicated in patients with a suspected ACS we undertook a 56 57 29 pragmatic, randomised comparison of ‘high’ and ‘low’ oxygen delivery strategies as part of usual care 58 59 60

3 | P a g e https://mc.manuscriptcentral.com/bmj Page 5 of 49 BMJ

1 2 3 1 in a large cohort of patients presenting to ambulances and acute cardiac care units thoughout New 4 5 2 Zealand (NZ) over 2 years. 6 7 3 8 9 4 10 11 5 METHODS 12 Confidential: For Review Only 13 6 This pragmatic, cluster-randomised, cross-over trial evaluated two oxygen protocols as part of the 14 15 7 routine care in patients presenting with suspected ACS throughout New Zealand over 2 years (Figure 16 17 8 1). 18 19 9 20 21 10 Study population 22 23 11 The study population included all suspected ACS patients in 2 national registries during the defined 24 25 12 study periods: the ANZACS-QI Registry, which includes ~98% of patients admitted to hospital in NZ 26 27 13 with ACS who have coronary angiography with or without coronary revascularisation,(10, 11), and/or 28 29 14 the Ambulance ACS Pathway, which includes all patients with suspected ACS managed by St John 30 31 15 ambulance service, which covers ~90% of the NZ population. For patients with multiple ACS 32 33 16 assessments, only the first presentation during the study was included. 34 35 17 36 18 Oxygen protocols 37 38 19 The high oxygen protocol recommended all patients with ischaemic chest pain or dyspnoea and 39 40 20 ischaemic electrocardiographic (ECG) changes receive oxygen with a flow of 6 to 8 l/minute by face 41 42 21 mask or 4 l/minute by nasal prongs, irrespective of SpO2. The oxygen flow rate could be increased if 43 44 22 needed to achieve an oxygen saturation >95%. Oxygen was recommended to be continued in the 45 46 23 ambulance until hospital admission, and in hospital until clinical evidence that myocardial ischaemia 47 48 24 had resolved. The low oxygen protocol recommended to only give oxygen in patients with ischaemic 49 50 25 chest pain or dyspnoea with ECG changes if SpO2 was <90%, with the flow rate adjusted to maintain 51 52 26 SpO2 between 90% and 94%. (see Appendix for full protocols). SpO2 was routinely monitored using 53 54 27 pulse oximeters. 55 56 28 57 58 29 The study aims and oxygen protocols were discussed with all doctors, nurses and paramedics who 59 30 care for patients with suspected or proven ACS across all geographic regions in NZ. Posters were 60

4 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 6 of 49

1 2 3 1 displayed on walls, and reminder stickers on monitors in ambulances, emergency departments, acute 4 5 2 cardiac care units and cardiac catheterization laboratories. 6 7 3 8 9 4 Cluster Cross Over Design 10 11 5 This was a pragmatic, two-arm, cluster-randomised, cross-over trial. At any one time the same 12 Confidential: For Review Only 13 6 oxygen protocol was used by all ambulances and hospitals across the pathway of care within each of 14 15 7 the 4 NZ regions (Northern, Midlands, Central, Southern). These geographic regions were chosen 16 17 8 as the clusters because all transfers between hospitals during an episode of ACS care occur within 18 9 these regions. The independent study statistician used computer generated random numbers to 19 20 10 allocate the 4 regions to the order of oxygen use, with 2 regions randomised to each oxygen protocol 21 22 11 in each cross-over period. The Northern and Southern Regions started with the high oxygen protocol, 23 24 12 and the Midland and Central Regions the low oxygen protocol, for seven months from 11 July 2016 to 25 26 13 7 February 2017. The oxygen protocols were then switched for the next 12 months (8 February 2017 27 28 14 to 13 February 2018) and changed back to the original protocols for the last five months (14 February 29 30 15 2018 to 10 July 2018) (Figure 1). The crossover dates were delayed from January to February to 31 32 16 avoid the national summer holidays, which resulted in more patients being included in the first 33 34 17 compared to the last study period. Thus each hospital and ambulance service used both the high and 35 36 18 low oxygen protocols for approximately one calendar year, and therefore similar seasons. The 37 38 19 number of cross-overs was limited to 2 to minimise practical challenges related to changing protocols 39 40 20 used as part of routine care. Two rather than one cross-over better accounted for possible temporal 41 21 trends not related to the oxygen protocols. To ensure familiarisation, protocols were introduced 1 to 42 43 22 3 months before the start of the study. To reduce the risk of early switching, clinicians and paramedics 44 45 23 were informed less than 5 days before each protocol change. The analysis plan prespecified that 46 47 24 patients admitted during the first two weeks after each protocol cross-over were to be excluded, 48 49 25 because protocol adherence may have been less during this time.” 50 51 26 52 53 27 Clincal and demographic characteristics of study population 54 55 28 All NZ residents have a unique National Health Identifier (NHI) number. For study patients, the 56 57 29 encrypted NHI (eNHI) was linked to the matched eNHI in NZ administrative data to determine all- 58 59 30 cause mortality up to 2 August 2019 with no known missing data. Demographic data and all hospital 60

5 | P a g e https://mc.manuscriptcentral.com/bmj Page 7 of 49 BMJ

1 2 3 1 admissions with discharge diagnoses based on ICD-10 codes were obtained from the National 4 5 2 Minimum Dataset included age, sex, ethnicity (NZ European, Māori, Pacific Islander, Asian, other) 6 7 3 and socioeconomic status from the NZ Deprivation Index 2013.(12) Prior myocardial infarction, prior 8 9 4 heart failure, and the Charlson co-morbidity score (13) excluding myocardial infarction and heart 10 11 5 failure, were also determined from ICD-10 codes for the past 5 years. Current smoking, diabetes and 12 Confidential: For Review Only 13 6 Killip class on admission were only available for patients in the ANZACS-QI registry. 14 15 7 16 17 8 For all patients in the ANZACS-QI Registry the discharge diagnosis was recorded by clinical staff.(10, 18 9 11) For patients managed using the St John Ambulance ACS Pathway, the incidence date was used 19 20 10 to identify the relevant hospital admission in the national hospitalization records, and discharge 21 22 11 diagnosis from ICD codes used to identify the type of ACS. Patients with diagnosis ‘unknown’ did not 23 24 12 have an ICD-10 discharge code. This group included patients discharged from hospital within three 25 26 13 hours and those transferred to healthcare centres or hospitals which do not routinely care for patients 27 28 14 with ACS. 29 30 15 31 32 16 Study end points 33 34 17 The primary outcome was all-cause mortality 30 days after first presentation with suspected ACS. The 35 36 18 pre-specified subgroup analyses were 30 day mortality in patients with a final diagnosis of STEMI and 37 38 19 non-STEMI. Secondary outcomes were one year all cause mortality, and length of hospital stay for 39 40 20 patients discharged alive at the index admission. 41 21 42 43 22 Audit of protocol adherence 44 45 23 A research nurse completed the audit form for 779 of 870 (90%) ANZACS-QI ACS registry patients 46 47 24 admitted on 28 randomly allocated days throughout the study (about 1 day each month). This form 48 49 25 recorded whether oxygen was used, the presence of ischemic symptoms or dynamic 50

51 26 electrocardiographic changes, SpO2 before starting and on oxygen, and contraindications to oxygen 52 53 27 use in each of 4 locations- ambulance, , cardiac cathetisation laboratory and 54 55 28 acute cardiac care unit. The duration of oxygen use was not recorded. 56 57 29 58 59 60

6 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 8 of 49

1 2

3 1 The Ambulance ACS pathway also recorded oxygen usage, the first SpO2 measured after ambulance 4

5 2 arrival, and the last SpO2 recorded in the ambulance for most participants. 6 7 3 8 9 4 Statistical analysis 10 11 5 A two year trial was chosen for pragmatic reasons. Based on historical registry records, we estimated 12 Confidential: For Review Only 13 6 that a study with nearly 30,000 patients would have 80% power at 5% level of significance (two- 14 15 7 sided) to detect a 20% relative and 0.6% absolute difference in all-cause 30 day mortality between the 16 17 8 two oxygen protocols, assuming the 30 day mortality of 3% for all patients with suspected ACS 18 9 patients. Historical data suggested that the intra-cluster correlation coefficient (ICC) was very small (< 19 20 10 0.0001). With only 4 regional networks served as the clusters, the power calculation did not account 21 22 11 for the cluster effect which was considered as a fixed effect in final analysis. The potential cluster 23 24 12 effect was estimated using random effects mixed model, and the estimated ICC was consistent with 25 26 13 the original trial design. 27 28 14 29 30 15 An intention to treat analysis was performed for all eligible patients identified from the two trial 31 32 16 registries, irrespective of oxygen received. Baseline characteristics and clinical outcomes of patients 33 34 17 were first summarised descriptively for each treatment period. The primary outcome was compared 35 36 18 between the two oxygen protocols using generalised linear regression model with a binomial 37 38 19 distribution and logit link. Both unadjusted and adjusted analyses were considered to evaluate 39 40 20 potential confounding effects of individual baseline characteristics associated with mortality. The 41 21 regression model adjusted for the following pre-specified baseline variables: region (Northern, 42 43 22 Midland, Central, Southern), age at admission (in years), sex, ethnicity (Māori, Pacific, Asian, NZ 44 45 23 European or other), socioeconomic status (NZ deprivation decile) and season. For the primary 46 47 24 analysis there was no known missing data for vital status, or for baseline variables except for 48 49 25 socioeconomic status, which was not known in 286 (0.7%) patients as part of the national data 50 51 26 collections. No imputation for missing data was performed. 52 53 27 54 55 28 Unadjusted and adjusted odds ratios (OR) are reported with the 95% confidence interval (CI) and 56 57 29 associated p-value. Relative risks (RR) were also estimated using Poisson regression models with 58 59 60

7 | P a g e https://mc.manuscriptcentral.com/bmj Page 9 of 49 BMJ

1 2 3 1 robust error estimates and the results were similar since the mortality rates were low in our study 4 5 2 cohort (not reported). 6 7 3 8 9 4 The same regression model was used on one year mortality. The Kaplan-Meier plot is presented for 10 11 5 the primary outcome to illustrate deaths by duration of follow-up, together with the hazard ratios and 12 Confidential: For Review Only 13 6 95% confidence intervals using Cox proportional hazard models. Cardiovascular disease as the 14 15 7 cause of death was only available to the end of 2017, and is reported for patients admitted to hospital 16 17 8 by November 2017. 18 9 19 20 10 Prespecified subgroup analysis explored the consistency of treatment effects in patients with a final 21 22 11 diagnosis of STEMI and non-STEMI, and in post hoc analyses for subgroups by region and sex, using 23 24 12 the same method as for the primary outcome measure. The interaction effect between oxygen 25 26 13 protocol and ACS diagnosis was tested in the main model by adding an interaction term between 27 28 14 treatment group and ACS type. 29 30 15 31 32 16 The Ambulance ACS pathway was implemented just before starting the trial, so data from this source 33 34 17 was not available at the time of trial design. Post hoc subgroup analyses were conducted for 35 36 18 patients managed in the ambulance ACS pathway only with first SpO2 recorded (<95% vs ≥95%). 37 38 19 These cut levels are chosen because they 95% is considered the lower end of the normal range for 39 40 20 SpO2, and patients with SpO2 below this level were excluded from many previous trials which 41 21 evaluated liberal versus conservative oxygen therapy in normoxemic subjects.(3, 14) 42 43 22 44 45 23 For patients discharged alive from hospital within 8 weeks, length of hospital stay was reported as 46 47 24 mean (SD) and median (interquartile range) for days in hospital. The mean difference in length of 48 49 25 stay and 95% confidence interval for patients managed under the high and low oxygen protocols are 50 51 26 also reported. Patients with a length of stay longer than 8 weeks (n=64) were excluded from this 52 53 27 analysis. 54 55 28 56 57 58 59 60

8 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 10 of 49

1 2 3 1 Statistical tests are two-sided at 5% significance level. Statistical analysis was conducted using SAS 4 5 2 version 9.4 (SAS Institute Inc., Cary, NC, USA). No interim efficacy analysis was conducted during the 6 7 3 trial. 8 9 4 10 11 5 Study Oversight 12 Confidential: For Review Only 13 6 The study was approved by the National Health and Disability Ethics Committee (15/NTA/117) and 14 15 7 Māori Research Review Committee without requiring individual patient consent. It was agreed that 16 17 8 obtaining consent may not be appropriate during a medical emergency for a simple low risk 18 9 intervention. The trial protocol was registered on the ANZ Clinical Trials Website 19 20 10 (ACTRN12616000461493). Study conduct and progress were reviewed by the Data Monitoring 21 22 11 Committee of the Health Research Council (HRC) of NZ (see Appendix). 23 24 12 25 26 13 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

9 | P a g e https://mc.manuscriptcentral.com/bmj Page 11 of 49 BMJ

1 2 3 1 RESULTS 4 5 2 A total of 40,872 patients with suspected or confirmed ACS were enrolled; 20,304 patients were 6 7 3 managed using the high oxygen protocol and 20,568 using the low oxygen protocol (Figure 1). This 8 9 4 included patients in the ANZACS-QI registry (high n=9726, low n=9840) and patients on the 10 11 5 ambulance ACS pathway who were not in the ANZACS-QI registry (high n=10,578 low n=10,728). 12 Confidential: For Review Only 13 6 The demographic and clinical characteristics of patients managed using the two protocols were well 14 15 7 matched (Table 1). 16 17 8 TABLE 1: BASELINE CHARACTERISTICS OF STUDY POPULATION 18 High Oxygen Low Oxygen 19 20 Number % Number % 21 22 All suspected ACS 20,304 49.7 20,568 50.3 23 24 Age mean years (SD) 66.5 (14.4) .. 66.4 (14.6) .. 25 26 Sex (male) 11,777 58.0 11,943 58.1 27 28 Ethnic group 29 30 Māori 2,298 11.3 2,370 11.5 31 32 Pacific 919 4.5 958 4.7 33 34 Asian 1,404 6.9 1,495 7.3 35 36 NZ European# 15,683 77.2 15,745 76.6 37 Clinical characteristics 38 39 Prior myocardial infarction 2343 11.5 2349 11.4 40 41 Prior heart failure 1723 8.5 1741 8.5 42 43 Co-morbidity score 0 15373 75.7 15648 76.1 44 45 1-2 3812 18.8 3854 18.7 46 47 3+ 1119 5.5 1066 5.2 48 49 Current Smoker~ 1661 17.1 1725 17.5 50 51 Diabetes~ 2283 23.5 2323 23.6 52 53 Killip class ~ 1 8632 88.7 8787 89.3 54 55 2 770 7.9 724 7.4 56 57 3 or 4 324 3.3 328 3.3 58 59 9 60

10 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 12 of 49

1 2 3 Socioeconomic Status in Quintiles+ 4 5 1 (least deprived) 3,234 15.9 3,203 15.6 6 7 2 3,512 17.3 3,554 17.3 8 9 3 3,953 19.5 4,034 19.6 10 11 4 4,717 23.2 4,770 23.2 12 Confidential: For Review Only 13 5 (most deprived) 4,753 23.4 4,856 23.6 14 15 Not recorded 135 0.7 151 0.7 16 17 Geographic region 18 19 Northern 6,913 34.1 7,109 34.6 20 21 Midland 4,815 23.7 4,916 23.9 22 23 Central 3,658 18.0 3,836 18.7 24 25 Southern 4,918 24.2 4,707 22.9 26 27 Final diagnosis 28 29 STEMI 2,031 10.0 2,128 10.4 30 31 Non-STEMI 5,152 25.4 5,066 24.6 32 33 Unstable angina 1,606 7.9 1,678 8.2 34 Non-ACS condition 9,681 47.7 9,838 47.8 35 36 Not classified 1,834 9.0 1,858 9.0 37 38 1 39 40 2 Legend: 41 42 3 + Socioeconomic status determined from NZDEP2013 census data in quintiles of the New Zealand general 43 4 population. NZDEP2013 was missing for 286 (0.7%) patients. 44 45 5 # Other ethnic groups were <2% of the study population, and are included in the New Zealand European (also 46 47 6 known as Pakeha) group. 48 49 7 ~data on smoking, diabetes and Killip class on hospital admission were recorded only for patients in ANZACS-QI 50 8 register (for high protocol in 9726, and low protocol in 9840 patients). Killip class refers to clinical evidence of 51 52 9 heart failure, 1= no heart failure, 2= rales or raised jugular venous pressure, 3= pulmonary oedema, 4= 53 54 10 cardiogenic shock. 55 56 11 57 12 58 59 13 60

11 | P a g e https://mc.manuscriptcentral.com/bmj Page 13 of 49 BMJ

1 2 3 1 Abbreviations: 4 5 2 ACS - Acute Coronary Syndromes; STEMI- ST-elevation myocardial infarction; NSTEMI – Non ST-elevation 6 3 myocardial infarction; SD – standard deviation. Co-morbidity score = Charleton Co-morbidity Score excluding 7 8 4 cardiac conditions. 9 10 5 11 12 6 OutcomesConfidential: for all patients with suspected ForACS Review Only 13 14 7 For all patients 30-day mortality for the high and low oxygen protocols was 613 (3.0%) and 642 15 16 8 (3.1%) respectively. The unadjusted odds ratio (OR) was 0.97, (95% CI 0.86, 1.08) and the adjusted 17 18 9 OR 0.96, (95% CI 0.86, 1.08) (Table 2). Because the mortality rate was low, results when were 19 20 10 similar when expressed as a relative risk reduction (RRR unadjusted 0.97 (95% CI 0.87-1.08), RRR 21 22 11 adjusted 0.96 (95% CI 0.87-1.07). 23 12 24 25 13 For those patients admitted to hospital by November 2017 (N=29,695) whose cause-specific mortality 26 27 14 data were available, 942 patients died in 30 days and of these 658 were cardiovascular deaths (high 28 29 15 protocol n=303/14,127, low protocol n=355/14,910). The unadjusted and adjusted odds ratios for 30- 30 31 16 day CVD mortality were 0.90 (95% CI 0.77,1.05) and 0.86 (95% CI 0.73, 1.01) respectively. One year 32 33 17 mortality for all patients managed using the high and low oxygen protocols was 1726 (8.5%) and 1682 34 35 18 (8.2%) respectively, unadjusted OR 1.04 (95% CI 0.97, 1.12), and adjusted OR 1.05 (95% CI 0.97, 36 37 19 1.13). 38 39 20 40 41 21 The Kaplan-Meier survival plot is displayed in Figure 2. Unadjusted and adjusted hazard ratios (HR) 42 43 22 on all-cause mortality between high and low oxygen protocols were 1.03 (95% CI 0.98, 1.09) and 1.05 44 45 23 (95% CI 0.99, 1.11) respectively. There were no significant differences in either 30 day or one year 46 24 mortality by age, sex or region (see supplementary table 1 and 2). Results are similar for all patients 47 48 25 in the ANZACS-QI registry, and all patients in the Ambulance ACS pathway not in ANZACS-QI 49 50 26 (supplementary table 1). 51 52 27 53 54 28 Length of hospital stay was compared between the two oxygen protocols in 36,278 patients admitted 55 56 29 to hospital and discharged alive within 8 weeks. The mean length of hospital stay (LOS) for patients 57 58 30 managed using the high and low oxygen protocols was 3.68 (SD 4.92) and 3.67 (SD 5.02) days 59 60

12 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 14 of 49

1 2 3 1 respectively. The median (interquartile range) LOS was 2 (1, 5) days in both groups. The adjusted 4 5 2 mean difference in length of stay was -0.0062 days, 95% CI -0.102, +0.090. 6 7 3 8 9 4 10 11 5 Outcomes in patients with STEMI 12 Confidential: For Review Only 6 For the 4159 STEMI patients, 10% of the total, 30-day mortality in the high and low oxygen group was 13 14 7 178 (8.8%) and 225 (10.6%), respectively. The unadjusted OR was 0.81 (95% CI 0.66, 1.00), and 15 16 8 adjusted OR 0.78 (95% CI 0.63, 0.97). There was not a statistically significant interaction effect 17 18 9 between STEMI and other ACS patients (p=0.106) or non-ACS patients (p=0.062). For patients with 19 20 10 a final diagnosis of STEMI one year mortality for the high procotol was 272 (13.4%) and the low 21 22 11 protocol 321 (15.1%), adjusted OR 0.84 (95% CI 0.70, 1.01), p=0.063. 23 24 12 25 26 13 Outcomes in patients with non-STEMI and without ACS 27 28 14 There were no significant differences in 30 day or one year mortality between the two oxygen groups 29 30 15 for patients with non-STEMI or for patients who did not have a final diagnosis of ACS (table 2, 31 32 16 supplementary table 2). 33 34 17 35 18 Outcomes by SpO level on ambulance arrival 36 2 37 19 Oxygen use and SpO levels were recorded in the ambulance ACS pathway in 20,444 patients. 38 2 39 20 Oxygen use in the ambulance was greater for patients with a final diagnosis of STEMI (high 60%, low 40 41 21 18%), and lower in patients with a final diagnosis of non-STEMI (high 40%, low 10%), and not ACS 42 43 22 (high 33%, low 7%). 44 45 23 46

47 24 30 day mortality was approximately 6 times higher when first SpO2 in the ambulance was <95% 48 49 25 compared to ≥95%. For patients with a first Sp02≥95% 30 day mortality was similar for patients 50 51 26 managed under the high and the low oxygen strategies (1.8% versus 1.9% respectively). For patients 52 53 27 with a first SpO2<95% 30 day mortality for patients managed under the high and the low oxygen 54 55 28 strategies was 10.8% and 12.1% respectively. Differences in 30 day mortality stratified by first SpO2 56 57 29 were similar for patients with a final diagnosis of STEMI, non-STEMI and other diagnoses, but 58 30 reflecting small numbers, the confidence intervals were wide (supplementary table 3). 59 60

13 | P a g e https://mc.manuscriptcentral.com/bmj Page 15 of 49 BMJ

1 2 3 1 4 5 2 TABLE 2: 30-DAY ALL-CAUSE MORTALITY BY OXYGEN PROTOCOL 6 7 High oxygen Low oxygen OR (95%CI) P OR (95%CI) P 8 9 N (%) N (%) Unadjusted value adjusted value 10 11 All patients + 20,304 20,568 ...... 12 Confidential: For Review Only 13 Mortality 613 (3.0%) 642 (3.1%) 0.97 (0.86, 1.08) 0.55 0.96 (0.86, 1.08) 0.50 14 15 STEMI * 2,031 2,128 ...... 16 17 Mortality 178 (8.8%) 225 (10.6%) 0.81 (0.66, 1.00) 0.049 0.78 (0.63, 0.97) 0.027 18 19 Non-STEMI* 5,152 5,066 ...... 20 21 Mortality 187 (3.6%) 176 (3.5%) 1.05 (0.85, 1.29) 0.67 1.02 (0.83, 1.27) 0.84 22 23 Unstable 24 1606 1678 ...... 25 angina 26 Mortality 13 (0.81%) 22 (1.31%) 0.61 (0.31, 1.22) 0.17 0.59 (0.29, 1.18) 0.14 27 28 Not ACS 9,681 9,838 ...... 29 30 Mortality 174 (1.8%) 175 (1.8%) 1.01 (0.82, 1.25) 0.92 1.04 (0.84, 1.29) 0.71 31 32 Diagnosis not 33 1834 1858 ...... 34 classified 35 36 Mortality 61 (3.33%) 44 (2.37%) 1.42 (0.95, 2.10) 0.081 1.45 (0.97, 2.18) 0.072 37 38 First SpO2 39 8,928 9,088 ...... 40 ≥95%++ 41 42 Mortality 162 (1.8%) 170 (1.9%) 0.97 (0.78, 1.21) 0.78 0.94 (0.75, 1.19) 0.62 43

44 First SpO2 45 1,220 1,208 ...... 46 <95%++ 47 48 Mortality 132 (10.8%) 146 (12.1%) 0.88 (0.69, 1.13) 0.33 0.86 (0.66, 1.12) 0.25 49 50 3 51 52 4 Legend: 53 5 + Primary outcome 54 55 6 † Age is stratified at median age for study population 56 57 7 *STEMI and non-STEMI were pre-specified secondary analyses 58 8 ++ First SpO = first oxygen saturation level measured on ambulance arrival. Data only available for 20,444 59 2 60 9 patients from the Ambulance ACS pathway.

14 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 16 of 49

1 2 3 1 .. Not applicable 4 5 2 Results for are presented by age, sex, region and for patients with unstable angina and diagnosis 6 3 Abbreviations: ACS –acute coronary syndrome; ANZACS-QI, All New Zealand Acute Coronary Syndromes 7 8 4 Quality Improvement Registry; CI – confidence level; NSTEMI – Non ST-elevation myocardial infarction; STEMI- 9 10 5 ST-elevation myocardial infarction; OR – odds ratio 11 12 6 Confidential: For Review Only 13 14 7 Protocol adherence 15 16 8 About half of all patients with suspected ACS had no indication for oxygen under either protocol, 17 9 mostly because they did not have clinical evidence of ongoing myocardial ischaemia. 42% of all 18 19 10 patients were given oxygen at least once across the care pathway under the high protocol compared 20 21 11 to 22% who received oxygen under the low oxygen protocol. For patients given oxygen, SpO was 22 2 23 12 higher for the high compared to the low oxygen protocol before starting oxygen [high median 96 24 25 13 (inter-quartile range 92, 98)% versus low 90 (86, 95) %] and when on oxygen [high 98 (96, 100) % 26 27 14 versus low 96 (94, 98) %]. 28 29 15 30 31 16 Audit of medical records for protocol adherence was completed in 779 of 812 patients admitted on 32 33 17 randomly selected days from the ANZACS-QI registry (Table 3). Reasons for non-completion of the 34 35 18 audit were not documented. Demographic and clinical characteristics for audited patients were 36 37 19 similar to the overall cohort (supplementary table 4). 8% of patients on the high oxygen protocol did 38 39 20 not receive oxygen despite having ischemic chest pain or electrocardiographic changes during 40 21 medical assessment. 15% of patients were non-adherent to the ‘low’ oxygen protocol, because they 41 42 22 were given oxygen without documentation of SpO <90% at least once during care (Table 3). 43 2 44 23 45 46 24 47 48 49 50 51 52 53 54 55 56 57 58 59 60

15 | P a g e https://mc.manuscriptcentral.com/bmj Page 17 of 49 BMJ

1 2 3 1 Table 3: 4 5 2 Audit of protocol adherence and oxygen use in a random sample of ANZACS-QI patients 6 7 High oxygen Low oxygen 8 9 protocol protocol 10 11 Number (%) Number (%) 12 Confidential: For Review Only 13 14 Patients audited 385 394 15 16 Oxygen given 162 (42%) 88 (22%) 17 18 SpO2 ≥90% before first oxygen use 138 (36%) 60 (15%)+ 19 20 SpO2 <90% before first oxygen use 24 (6%) 22 (6%) 21 22 Oxygen not given 23 24 Ischaemic symptoms 35 (8%)* 118 (30%) 25 26 No ischaemic symptoms 182 (47%) 188 (48%) 27 28 Contraindication for oxygen 6 (2%) 6 (2%) 29 SpO for patients given oxygen Median (IQR) Median (IQR) 30 2 31 Before first oxygen use 96 (92, 98) % 90 (86, 95) % 32 33 After first oxygen use 98 (96, 100) % 96 (94, 98) % 34 35 3 36 37 4 Legend: 38 39 5 Oxygen use was audited from review of medical records on 28 random days during the study. 40 41 6 Clinical characteristics of audited patients were similar to non-audited patients in the ANZACS-QI 42 43 7 cohort (supplementary table 4) 44 45 8 * Patients with ischaemic symptoms who were not given oxygen protocol were non-adherent for the 46 47 9 high oxygen protocol. 48 49 10 + Patients given oxygen but with no documented SpO2<90% were non-adherent for the low oxygen 50 51 11 protocol. 52 53 12 SpO2 = transcapillary oxygen saturation level when measured before and after first use of oxygen 54 55 13 across the care pathway. 56 14 IQR - inter-quartile range. 57 58 15 59 60

16 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 18 of 49

1 2 3 1 DISCUSSION 4 5 2 In this large study there was neither benefit nor harm from a protocol which recommended 6 7 3 administering high flow oxygen for ischemic symptoms or electrocardiographic changes as part of 8 9 4 routine care in a broad general population of patients presenting with suspected ACS. This finding is 10 11 5 consistent with previous studies suggesting that high flow oxygen is unlikely to benefit most patients 12 Confidential: For Review Only 13 6 with presumed ischemic chest pain and a normal oxygen saturation level.(4, 14, 15) Because oxygen 14 15 7 is widely used it is important to demonstrate that it does not cause harm. In the current study, as in 16 17 8 real world use, oxygen was given to patients with suspected ACS before the diagnosis was 18 19 9 confirmed, with the knowledge that many patients will not have a final diagnosis of ACS.(16) In this 20 21 10 trial almost 2/3 of the study population were subsequently found to have neither STEMI nor non- 22 23 11 STEMI. 24 25 12 26 13 It is possible STEMI patients could have more to gain from oxygen because they usually present with 27 28 14 severe prolonged myocardial ischaemia. Also STEMI can be diagnosed early after first medical 29 30 15 contact from the electrocardiogram, allowing targeted oxygen use if appropriate. This trial included a 31 32 16 larger STEMI cohort than previous trials,(4) even though STEMI patients comprised only 10% of the 33 34 17 whole study population. The cluster-randomised, all-comers design, without individual participant 35 36 18 consent, meant that all patients in the registries, including acutely unwell individuals who would be 37 38 19 unable to provide consent for a traditional study protocol. The inclusion of high risk participants is 39 40 20 demonstrated by the 30 day mortality for STEMI patients of ~10% compared to ~2% in DETO2X-AMI 41 42 21 trial.(4) However, despite a nominally lower 30 day mortality for patients with STEMI under the high 43 44 22 oxygen protocol, it is difficult to draw definite conclusions on benefit for this subgroup given the 45 46 23 neutral result for the primary outcome of the study. 47 48 24 49 25 Most previous oxygen strategy trials(3, 14) were designed to evaluate effects of maintaining high 50 51 26 SpO in normoxemic patients, and therefore excluded people with an oxygen saturation level below 52 2 53 27 94%(15) or 90%(4). These trials were therefore not designed to evaluate whether there is benefit from 54 55 28 high flow oxygen in patients with SpO2 levels below normal. Current clinical practice guidelines, 56 57 29 recommend giving oxygen to patients with STEMI or non-STEMI when the Sp02 is <94% or <90%(5- 58 59 30 9), which presumes some benefit, despite little randomised evidence. In the current trial the ‘low 60

17 | P a g e https://mc.manuscriptcentral.com/bmj Page 19 of 49 BMJ

1 2

3 1 oxygen’ strategy recommended a lower threshold for starting oxygen (SpO2<90%) and lower oxygen 4

5 2 flow rates to avoid high SpO2 levels. It was therefore less effective compared to the ‘high oxygen’ 6 7 3 protocol for preventing mild hypoxemia. However because of the study design treatment effects in 8 9 4 patients with SpO2 below normal could only be evaluated in a secondary analysis of ~2428 patients 10 11 5 who had an SpO2<95% on ambulance arrival. Compared to patients with a normal SpO2, these 12 Confidential: For Review Only 13 6 patients had a ~5 fold higher 30 day mortality, but no definite difference between protocols (low 14 15 7 oxygen 12.1% vs high oxygen 10.8%). A secondary analysis from DETO2X-AMI also reported a 16 17 8 much higher mortality for patients with SpO2<95% but no difference in mortality, reinfarction or heart 18 9 failure hospitalisation over the next 1-4 years between liberal and conservative oxygen protocols.(17) 19 20 10 21 22 11 Low statistical power for modest but clinically-important effects of oxygen is a limitation of both the 23 24 12 current and all previous studies. In a meta-analysis of the 8 studies(14) which compared liberal with 25 26 13 conservative oxygen use in a total of 7732 patients, in-hospital mortality was only 1.7% (133 deaths), 27 28 14 and a 30% relative difference in in-hospital mortality with high flow oxygen could not be excluded. 29 30 15 One year mortality was a secondary outcome in the current study, but was thought less likely than 30 31 32 16 day mortality to be influenced by early oxygen use. 33 34 17 35 36 18 The pragmatic cluster randomized design had both advantages and limitations. Advantages include 37 38 19 the ability to compare outcomes with two protocols used as part of usual care, in a representative 39 40 20 population, and at multiple locations across the acute care pathway. This made it possible to 41 21 evaluate oxygen use in a much larger population than would be possible in a conventional trial with 42 43 22 individual consent and randomisation. The pragmatic design allowed the trial to be undertaken with a 44 45 23 low administrative burden and at modest cost. Sicker, higher risk patients, who are generally less 46 47 24 likely to be enrolled in clinical trials, were included. A risk of the cluster design is imbalance between 48 49 25 study groups. The cross-over allowed the two oxygen protocols to be compared within regions 50 51 26 accounting for possible hospital, temporal and seasonal effects. Baseline characteristics of patients 52 53 27 managed under two protocols differed by <1%, suggesting that randomisation resulted in equivalent 54 55 28 groups. The risk of bias from patient selection or evaluation of outcomes was minimized by using 56 57 29 electronic linkage of all patients in the two registers admitted on the specified study dates to 58 59 60

18 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 20 of 49

1 2 3 1 independently collected administrative data for diagnosis and outcomes, with no or minimal missing 4 5 2 data. 6 7 3 8 9 4 A disadvantage of the study design was the study population included all patients in the two registries, 10 11 5 whether a patient had an indication for oxygen or not, and irrespective of how likely they were to have 12 Confidential: For Review Only 13 6 a final diagnosis of ACS. Thus many patients included in the analysis did not have ischemic 14 15 7 symptoms when seen, and had no indication for oxygen under either protocol. Most of these patients 16 17 8 had a very low 30 day mortality. An additional limitation is that information on oxygen use and 18 9 protocol adherence was not available for most participants, and was missing for 10% of audited 19 20 10 patients. We do not know the specific reasons for protocol non-adherence. They are likely to include 21 22 11 not documenting oxygen use, an active decision by clinical staff or patient not to follow the study 23 24 12 recommended oxygen protocol based on their personal judgment, or failure to ‘remember’ to use the 25 26 13 recommended protocol. Overall we believe the level of non-adherence is consistent with the 27 28 14 expected level in a pragmatic clinical trial undertaken in real world settings. Protocol non-adherence 29 30 15 would bias results towards no difference, and decrease the statistical power of the trial to detect any 31 32 16 true difference in outcomes related to the oxygen protocol. 33 34 17 35 36 18 The study protocols did not specify the duration of oxygen use, which would vary between patients 37 38 19 according to clinical circumstances, consistent with usual clinical practice. These oxygen protocols 39 40 20 contrast with most previous oxygen strategy trials where oxygen (or air) was usually administered for 41 21 a defined time irrespective of clinical status.(3, 14) The presence of hypoxemia could be an important 42 43 22 determinant of benefit from oxygen, but SpO levels were only known before starting oxygen for 44 2 45 23 20,444 patients managed on the ambulance ACS pathway, and of these only 2428 patients had a first 46 47 24 SpO2<95%. It was therefore not possible to reliably evaluate treatment effects in patients who had a 48

49 25 SpO2<95%. 50 51 26 52 53 27 In conclusion, this study supports current clinical practice guidelines (5, 6, 8, 9) which recommend 54 55 28 oxygen is not given to patients with suspected ACS who have a normal SpO2. However the role of 56 57 29 supplementary oxygen when SpO2 decreases below the normal range remains uncertain. 58 59 30 60

19 | P a g e https://mc.manuscriptcentral.com/bmj Page 21 of 49 BMJ

1 2 3 1 ETHICS APPROVAL 4 5 2 The authors obtained ethics approval from the National Health and Disability Ethics Committee 6 7 3 (15/NTA/117). The National Ethics Committees approved the study without patient consent because 8 9 4 they considered there was clinical equipoise, that study conduct did not place demands on patients or 10 11 5 reduce their autonomy, and obtaining informed consent may not be appropriate during a medical 12 Confidential: For Review Only 13 6 emergency. Information on the trial was openly available for patients and doctors, who could decide 14 15 7 whether to follow the study oxygen protocols. 16 17 8 18 19 9 ROLE OF THE FUNDING 20 21 22 10 This work was supported by the National Heart Foundation of New Zealand (grant number 1649), and 23 11 the HRC Clinical Practitioner Fellowships to R.S and M.T. The trial was undertaken as part of 24 25 12 ANZACS-QI which is funded by the Ministry of Health (ANZACS-QI 41) Data linkage for mortality was 26 27 13 obtained through the VIEW program funded by the Health Research Council of New Zealand. The 28 29 14 trial protocol was registered on the ANZ Clinical Trials Website (ACTRN12616000461493). 30 31 15 32 33 16 We confirm that all authors had full access to all of the data (including statistical reports and tables) in 34 35 17 the study and can take responsibility for the integrity of the data and the accuracy of the data analysis. 36 37 18 38 39 19 TRANSPARENCY STATEMENT 40 41 20 Dr Stewart affirms that the manuscript is an honest, accurate, and transparent account of the study 42 43 21 being reported; that no important aspects of the study have been omitted; and that any discrepancies 44 45 22 from the study as originally planned (and, if relevant, registered) have been explained. 46 47 23 48 49 24 PATIENT AND PUBLIC INVOLVEMENT STATEMENT 50 51 25 The study was considered by Research Review Committees and Maori health Advisory committees at 52 53 26 District Health Boards throughout New Zealand, and for the two national ambulance services, each of 54 55 27 whom included community representation. 56 57 28 58 59 60 29

20 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 22 of 49

1 2 3 1 DISSEMINATION TO PARTICIPANTS AND RELATED PATIENT AND PUBLIC COMMUNITIES 4 5 2 Results of the study have been reported to relevant health care providers through the national cardiac 6 7 3 clinical networks and ambulance services. Results have been made available to the public through 8 9 4 the National Heart Foundation of New Zealand (the study sponsor) and through the media. 10 11 5 12 Confidential: For Review Only 13 14 6 WHAT IS ALREADY KNOWN? 15 16 7 Previous trials have reported no benefit and possible harm from supplementary oxygen given to 17 18 8 normoxemic patients presenting with suspected or confirmed acute coronary syndromes. 19 20 9 21 22 10 However these trials were too small to reliably evaluate modest favourable or unfavourable effects of 23 24 11 oxygen on early mortality, or specific circumstances when oxygen may be beneficial. 25 26 12 27 28 13 WHAT THIS STUDY ADDS 29 30 14 Supplementary high flow oxygen given for ischaemic symptoms to patients with suspected ACS did 31 32 15 not change 30 day mortality. 33 34 16 35 36 17 The study did not determine whether or not supplementary oxygen benefits patients with suspected 37 38 18 ACS or ST elevation myocardial infarction if the oxygen saturation level decreases below the normal 39 40 19 range (SpO2<95%). 41 42 20 43 44 21 45 46 22 47 48 49 50 51 52 53 54 55 56 57 58 59 60

21 | P a g e https://mc.manuscriptcentral.com/bmj Page 23 of 49 BMJ

1 2 3 1 ACKNOWLEDGEMENTS 4 5 2 We thank study investigators (see Appendix), the St John Ambulance and Wellington Free 6 7 3 Ambulance servicee, the National Cardiac and Emergency Medicine Clinical Networks and all 8 9 4 hospitals which participate in the All NZ Acute Coronary Syndromes Quality Improvement (ANZACS- 10 11 5 QI) Registry (10), the Data Monitoring Committee (see Appendix), contributors to ANZACS-QI and the 12 Confidential: For Review Only 13 6 VIEW program, and physicians, nurses and paramedics who used the study oxygen protocols. We 14 15 7 also acknowledge patients who suffered a suspected ACS during the study period. 16 17 8 18 19 9 DECLARATION OF INTEREST 20 21 10 None declared 22 23 11 24 25 12 ATA HARING TATEMENT 26 D S S 27 28 13 Study data can be made available to other investigators on request. 29 30 14 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

22 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 24 of 49

1 2 3 1 FIGURE LEGENDS 4 5 2 6 7 3 Figure 1 8 9 4 Study flow diagram 10 11 5 12 Confidential: For Review Only 13 6 Figure 2 14 15 7 Kaplan Meier Plot showing all cause mortality for all patients with suspected ACS managed on the 16 17 8 high and low oxygen protocols. 18 19 9 20 21 10 All cause mortality was determined from national administrative data assessed on 12 August 2019 for 22 23 11 all patients. The length of follow-up after one year varied according to the date of first ACS 24 25 12 assessment during the study. 26 13 27 28 14 29 30 15 31 32 16 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

23 | P a g e https://mc.manuscriptcentral.com/bmj Page 25 of 49 BMJ

1 2 3 1 REFERENCES 4 5 2 1. Beasley R, Aldington S, Weatherall M, et al. Oxygen therapy in myocardial infarction: an 6 7 3 historical perspective. J R Soc Med. 2007;100(3):130-3. 8 9 4 2. McNulty PH, King N, Scott S, et al. Effects of supplemental oxygen administration on 10 11 5 coronary blood flow in patients undergoing cardiac catheterization. Am J Physiol Heart Circ Physiol. 12 Confidential: For Review Only 13 6 2005;288(3):H1057-62. 14 15 7 3. Chu DK, Kim LH, Young PJ, et al. Mortality and morbidity in acutely ill adults treated with 16 17 8 liberal versus conservative oxygen therapy (IOTA): a systematic review and meta-analysis. Lancet. 18 19 9 2018;391(10131):1693-705. 20 21 10 4. Hofmann R, James SK, Jernberg T, et al. Oxygen Therapy in Suspected Acute Myocardial 22 23 11 Infarction. N Engl J Med. 2017;377(13):1240-49. 24 25 12 5. Amsterdam EA, Wenger NK, Brindis RG, et al. 2014 AHA/ACC guideline for the management 26 13 of patients with non-ST-elevation acute coronary syndromes: executive summary: a report of the 27 28 14 American College of Cardiology/American Heart Association Task Force on Practice Guidelines. 29 30 15 Circulation. 2014;130(25):2354-94. 31 32 16 6. Ibanez B, James S, Agewall S, et al. 2017 ESC Guidelines for the management of acute 33 34 17 myocardial infarction in patients presenting with ST-segment elevation: The Task Force for the 35 36 18 management of acute myocardial infarction in patients presenting with ST-segment elevation of the 37 38 19 European Society of Cardiology (ESC). Eur Heart J. 2018;39(2):119-77. 39 40 20 7. O'Driscoll BR, Howard LS, Earis J, et al. BTS guideline for oxygen use in adults in healthcare 41 42 21 and emergency settings. Thorax. 2017;72(Suppl 1):ii1-ii90. 43 44 22 8. O'Gara PT, Kushner FG, Ascheim DD, et al. 2013 ACCF/AHA guideline for the management 45 46 23 of ST-elevation myocardial infarction: a report of the American College of Cardiology 47 48 24 Foundation/American Heart Association Task Force on Practice Guidelines. Circulation. 49 25 2013;127(4):e362-425. 50 51 26 9. Roffi M, Patrono C, Collet JP, et al. 2015 ESC Guidelines for the management of acute 52 53 27 coronary syndromes in patients presenting without persistent ST-segment elevation: Task force for 54 55 28 the management of acute coronary syndromes in patients presenting without persistent ST-segment 56 57 29 elevation of the European Society of Cardiology (ESC). Eur Heart J. 2016;37(3):267-315. 58 59 60

24 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 26 of 49

1 2 3 1 10. Kerr A, Williams MJ, White H, et al. The All New Zealand Acute Coronary Syndrome Quality 4 5 2 Improvement Programme: Implementation, Methodology and Cohorts (ANZACS-QI 9). The New 6 7 3 Zealand medical journal. 2016;129(1439):23-36. 8 9 4 11. Kerr AJ, Lee M, Jiang Y, et al. High level of capture of coronary intervention and associated 10 11 5 acute coronary syndromes in the all New Zealand acute coronary syndrome quality improvement 12 Confidential: For Review Only 13 6 cardiac registry and excellent agreement with national administrative datasets (ANZACS-QI 25). The 14 15 7 New Zealand medical journal. 2019;132(1492):19-29. 16 17 8 12. Atkinson J. SC, Crampton P. NZDep2013 Index of Deprivation. Wellington: Department of 18 9 Public Health, University of Otago., 2014. 19 20 10 13. Sundararajan V, Henderson T, Perry C, et al. New ICD-10 version of the Charlson 21 22 11 comorbidity index predicted in-hospital mortality. Journal of clinical epidemiology. 2004;57(12):1288- 23 24 12 94. 25 26 13 14. Sepehrvand N, James SK, Stub D, et al. Effects of supplemental oxygen therapy in patients 27 28 14 with suspected acute myocardial infarction: a meta-analysis of randomised clinical trials. Heart. 29 30 15 2018;104(20):1691-98. 31 32 16 15. Stub D, Smith K, Bernard S, et al. Air Versus Oxygen in ST-Segment-Elevation Myocardial 33 34 17 Infarction. Circulation. 2015;131(24):2143-50. 35 36 18 16. Kohn MA, Kwan E, Gupta M, et al. Prevalence of acute myocardial infarction and other 37 38 19 serious diagnoses in patients presenting to an urban emergency department with chest pain. J Emerg 39 40 20 Med. 2005;29(4):383-90. 41 21 17. James SK, Erlinge D, Herlitz J, et al. Effect of Oxygen Therapy on Cardiovascular Outcomes 42 43 22 in Relation to Baseline Oxygen Saturation. JACC Cardiovascular interventions. 2019. 44 45 23 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

25 | P a g e https://mc.manuscriptcentral.com/bmj Page 27 of 49 BMJ

1 2 3 4 Study oxygen protocols 5 6 ‘high’ or ‘low’ oxygen protocols introduced as part of usual care for all patients presenting 7 with suspected acute coronary syndromes in ambulances, emergency departments, cardiac 8 9 cathetersiation laboratories and acute cardiac care units throughout New Zealand 10 11 12 Confidential: For Review Only 13 Study population 14 15 All patients presenting with suspected ACS to ambulances and hospitals in the ANZACS-QI 16 or ambulance ACS registries during defined study periods over 2 years 17 18 or ANZACS-QI Registry during study periods 19 20 4 geographic regions randomised to use either the ‘high’ and ‘low’ oxygen protocol first 21 22 23 High oxygen protocol for 30 weeks Low oxygen protocol for 30 weeks 24 in Northern and Southern Regions in Midland and Central Regions 25 n= 7,570 n= 5,793 26 27 28 29 High oxygen protocol for 51 weeks Low oxygen protocol for 51 weeks 30 in Midland and Central Regions in Northern and Southern Regions 31 n= 8,473 32 n= 11,816 33 34 35

36 High oxygen protocol for 19 weeks Low oxygen protocol for 19 weeks 37 in Northern and Southern Regions in Midland and Central Regions 38 n= 4,261 19 weeks, n= 2,959 39 40 41 42 43 Primary outcome = 30 day mortality Primary outcome = 30 day mortality 44 for all patients in registries managed using high for all patients in registries managed using low 45 oxygen protocol oxygen protocol 46 n= 20,304 n= 20,568 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Confidential: ForBMJ Review OnlyPage 28 of 49

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 https://mc.manuscriptcentral.com/bmj 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 29 of 49 BMJ

1 2 3 4 SUPPLEMENTARY APPENDIX 5 6 7 Two oxygen protocols in patients presenting with a suspected acute 8 coronary syndrome 9 10 Stewart RAH, et al 11 12 Confidential: For Review Only 13 CONTENTS 14 15 1. Study Oxygen protocols ………………………………………………………………………....2 16 17 2. Data Monitoring Committee.……………………….……………………...... 3 18 19 3. Participating centres and investigators.…………………………………………………...... 4 20 21 4. Supplementary Figure 1: New Zealand map showing 4 regions, and hospitals 22 contributing to ANZACS-QI registry………………………………………...... 5 23 24 5. Supplementary Table 1: 30 day all-cause mortality by oxygen protocol in selected 25 groups ……………………………………………………………………………………………. 6 26 27 6. Supplementary Table 2: 1 year all-cause mortality by oxygen protocol in selected groups 28 ………………………………………………………………………………………………………7 29 30 7. Supplementary Table 3: 30 day mortality by oxygen protocol according to first SpO2 31 measured on ambulance arrival for patients managed on ambulance ACS pathway by 32 final diagnosis …………………………………………………………………………………….8 33 34 8. Supplementary Table 4:Baseline characteristics of patients randomly selected for audit of oxygen use, compared to other ANZACS-QI patients admitted during the study 35 periods …………………………………………………………………………………………….9 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

1 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 30 of 49

1 2 3 1. STUDY OXYGEN PROTOCOLS: 4 5 High oxygen protocol: 6  In patients with probable or confirmed acute coronary syndrome give oxygen for 7 ischaemic chest pain, ischaemic ECG changes or dyspnea related to myocardial 8 ischaemia irrespective of the measured oxygen saturation level. 9  Administer oxygen by face mask at ~8l/minute. If a face mask is not tolerated give 10 oxygen by nasal prongs at ~4 l/minute. Increase oxygen flow rate if necessary to 11 achieve saturation ≥95%. Continue oxygen until the patient is admitted to hospital or 12 Confidential:transferred to cardiology, or when For a doctor decidesReview it is no longer Only necessary. 13  Caution or avoid high flow oxygen in patients at risk of hypercapnia, including those 14 with possible obesity hypoventilation syndrome or chronic obstructive pulmonary 15 disease. 16 17 Low oxygen protocol: 18  In patients with a suspected or confirmed acute coronary syndrome do not give oxygen 19 for ischaemic chest pain, ischaemic ECG changes or dyspnea related to myocardial 20 ischaemia, unless the measured oxygen saturation is <90%. 21  When the oxygen saturation cannot be measured only administer oxygen if there is a 22 significant clinical concern of hypoxia. 23  If oxygen is administered by face mask or nasal prongs adjust the flow rate to achieve a 24 target oxygen saturation of 90 to 94%. 25  Discontinue oxygen when no longer needed to maintain saturation >90%. 26  Caution or avoid high flow oxygen in patients with possible obesity hypoventilation 27 syndrome or chronic obstructive pulmonary disease. 28  Oxygen can be administered in the ambulance to patients who need ventilation or 29 continuous positive airways pressure (CPAP). 30 In individual cases the oxygen protocol can be over ruled by clinician preference or clinical 31 indication. 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

2 | P a g e https://mc.manuscriptcentral.com/bmj Page 31 of 49 BMJ

1 2 3 2. DATA MONITORING COMMITTEE 4 5 The Health Research Council of New Zealand Data Monitoring Committee (DMC) monitored 6 trial conduct. The DMC meets in April and October each year. Aggregate data on oxygen 7 protocol adherence, and in-hospital and 30 day mortality from the ANZACS-QI Registry were 8 provided to the DMC once during the study, but no interim analysis was performed. The DMC 9 were also provided with copies of relevant research published during the trial. The DMC did not 10 review data from the Ambulance ACS pathway. 11 12 OXYGENConfidential: Data Monitoring committee (DMC)For -The Review Health Research Council Only of New Zealand 13 14 HRC DMC members: Associate Professor Katrina Sharples (Chair) - Department of Preventive 15 and Social Medicine, University of Otago, Dunedin; Dr Mark Jeffery - Department of , 16 Christchurch Hospital, Christchurch; Professor Ngaire Kerse - School of Population Health, 17 University of Auckland; Professor Thomas Lumley - Department of Statistics, University of Auckland; Associate Professor Andrew Moore - Department of Philosophy, University of Otago, 18 Dunedin; Associate Professor Conroy Wong - Department of Respiratory Medicine, Middlemore 19 Hospital, Auckland. 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

3 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 32 of 49

1 2 3 3. PARTICIPATING CENTRES AND INVESTIGATORS 4 5 In addition to the study authors the following investigators and study coordinators contributed to 6 the study. 7 8 Investigators and Co-coordinators at participating hospitals: 9 10 Northern Region 11 Auckland City Hospital: Ralph Stewart, Leah Howell, Catherine Patten, Peter Jones, Tracey 12 Barley,Confidential: Rebekka Costello, Karen Schimanski, For Nancy Review Mitchell Only 13 North Shore and Waitakere Hospitals: Tony Scott, Nichola Bradford, Kim Yates, Laura 14 Chapman 15 Middlemore Hospital: Andrew Kerr, Lynette Pearce, Renee Railton, Nicole Signal, Vanessa 16 Thornton, Adrienne Adams, Andrew Brainard 17 Whangarei Hospital: Samraj Nandra, Sue Vallancey, Ryan Howard 18 Midland Region 19 : Gerry Devlin, Liz Low, Roselyn Pillay, John Bonning 20 Tauranga Hospital: Jonathan Tisch, Jennifer Goodson, Alastair Maclean 21 Whakatane Hospital: Terri McBride, Martin Bentley-Smith, Adele Ferguson 22 Rotorua Hospital: Peace Tamuno, Paula Broughton, Vijay Thumma 23 Taupo Hospital: Paul Malpass 24 Gisborne Hospital: Alyssa Thompson, Joanne Patterson, Andrew Botting, Tom Palfi 25 Taranaki Base Hospital: Ian Ternouth, Jonele Woodhead, Carolyn Jackson, Donald McKee, 26 Mark Sagarin 27 Tokoroa / Te Kuiti Hospital: Shameem Safih 28 Thames Hospital: Vijaya Pera 29 30 Central Region 31 Wellington Hospital: Anil Ranchord, Bev Scott, Brad Peckler, Mark Hussey, Paul Quigley 32 Nelson Hospital: Liam Hughes, Sarah Flintoft, Liz Dalby, Lorna Johnston, Andrew Munro 33 Hutt Valley Hospital: Tim O’Meeghan, Ria Kleintjes, Jo-Anne Kovacs, Criselda Sayoc, Tracy 34 Langhorn, Ben Ross, Tanya Wilton 35 Hawkes Bay Hospital: Richard Luke, Diana Schmid, Julia Mackenzie, Craig Ellis 36 Palmerston North Hospital: Dave Tang, Lia Sinclair, Thomas Cheri 37 Wairarapa Hospital: Laura Davidson, Mary Quayle 38 Wanganui Hospital: Tom Thompson 39 40 Southern Region Dunedin Hospital: John Edmond, Shona Willers, Bruce Lambie, Deb Scott, Caroline Collins, 41 John Chambers 42 Christchurch Hospital: David Smyth, Michael Hume, Martin Than 43 Ashburton Hospital: John Lyons 44 Timaru Hospital: Matthew Hills, Robyn Carey, Bernhard Kuepper, Maria Hammond, Catherine 45 Baker, Kathy Patrick 46 47 Wellington Free Ambulance: Andrew Swain 48 St John Ambulance: Bridget Dicker, Tony Smith, Verity Todd 49 50 National Oxygen Study Mangers: 51 Darryl McGuire, Diana Gatland, Cardiovascular Research Unit, Auckland City Hospital 52 53 54 55 56 57 58 59 60

4 | P a g e https://mc.manuscriptcentral.com/bmj Page 33 of 49 BMJ

1 2 3 4 4. SUPPLEMENTARY FIGURE 1: NEW ZEALAND MAP SHOWING 4 REGIONS AND HOSPITALS 5 CONTRIBUTING TO ANZACS-QI REGISTRY 6 7 All transfers between hospitals for patients with ACS occur within these four regions 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 24 hrs PCI coverage Interventional centers (7) 46 <24 hrs PCI coverage Interventional centers (2) 47 Non interventional centers (15) 48 49 50 51 52 53 54 55 56 57 58 59 60

5 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 34 of 49

1 2 3 Supplementary Table 1: 30-Day all-cause mortality by oxygen protocol in selected 4 groups 5 6 High oxygen Low oxygen OR (95%CI) P OR (95%CI) P 7 N (%) N (%) Unadjusted value adjusted value 8 9 All patients+ 20,304 20,568 ...... 10 Mortality 613 (3.0%) 642 (3.1%) 0.97 (0.86, 1.08) 0.55 0.96 (0.86, 1.08) 0.50 11 Men 11777 11943 ...... 12 MortalityConfidential:364 (3.1%) 408 For (3.4%) Review0.90 (0.78, 1.04) Only0.16 0.89 (0.77, 1.04) 0.14 13 Women 8527 8625 ...... 14 Mortality 249 (2.9%) 234 (2.7%) 1.08 (0.90, 1.29) 0.41 1.08 (0.90, 1.31) 0.42 15 Geographic region 16 Northern 6913 7109 ...... 17 Mortality 175 (2.5%) 203 (2.9%) 0.88 (0.72, 1.09) 0.24 0.88 (0.71, 1.09) 0.23 18 Midland 4815 4916 ...... 19 Mortality 179 (3.7%) 163 (3.3%) 1.13 (0.91, 1.40) 0.28 1.12 (0.90, 1.40) 0.31 20 Central 3658 3836 ...... 21 Mortality 113 (3.1%) 138 (3.6%) 0.85 (0.66, 1.10) 0.22 0.87 (0.67, 1.13) 0.30 22 Southern 4918 4707 ...... 23 Mortality 146 (3.0%) 138 (2.9%) 1.01 (0.80, 1.28) 0.91 0.95 (0.75, 1.22) 0.71 24 25 ANZACS-QI registry 9726 9840 ...... 26 Mortality 286 (2.9%) 299 (3.0%) 0.97 (0.82, 1.14) 0.69 0.95 (0.80, 1.13) 0.58 27 Ambulance pathway+ 10,578 10,728 ...... 28 Mortality 327 (3.1%) 343 (3.2%) 0.97 (0.83, 1.13) 0.66 0.93 (0.79, 1.10) 0.38 29 30 Legend:+ patients on the ambulance ACS pathway who were not included in the ANZACS-QI 31 registry 32 33 Abbreviations: 34 ACS - Acute Coronary Syndromes; STEMI- ST-elevation myocardial infarction; NSTEMI – Non 35 ST-elevation myocardial infarction 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

6 | P a g e https://mc.manuscriptcentral.com/bmj Page 35 of 49 BMJ

1 2 3 Supplementary Table 2: One year all-cause mortality by oxygen protocol in selected 4 groups 5 6 High oxygen Low oxygen OR (95%CI) P OR (95%CI) P 7 N (%) N (%) Unadjusted value adjusted value 8 All patients 20,304 20,568 ...... 9 Mortality 1726 (8.5%) 1682 (8.2%) 1.04 (0.97, 1.12) 0.24 1.05 (0.97, 1.13) 0.21 10 ANZACS-QI 9726 9840 ...... 11 registry 12 MortalityConfidential:721 (7.4%) 710 (7.2%)For 1.03Review (0.93, 1.15) 0.60Only1.02 (0.92, 1.15) 0.68 13 Ambulance ACS 10,578 10,728 ...... 14 pathway 15 Mortality 1005 (9.5%) 972 (9.1%) 1.05 (0.96, 1.16) 0.27 1.04 (0.95, 1.15) 0.40 16 STEMI 2031 2128 ...... 17 Mortality 272 (13.4%) 321 (15.1%) 0.87 (0.73, 1.04) 0.12 0.84 (0.70, 1.01) 0.06 18 NSTEMI 5,152 5,066 ...... 19 Mortality 555 (10.8%) 512 (10.1%) 1.07 (0.95, 1.22) 0.27 1.07 (0.94, 1.23) 0.31 20 Unstable angina 1606 1678 ...... 21 Mortality 65 (4.1%) 76 (4.5%) 0.89 (0.64, 1.25) 0.50 0.85 (0.60, 1.21) 0.36 22 Not ACS 9,681 9,838 ...... 23 Mortality 702 (7.3%) 648 (6.6%) 1.11 (0.99, 1.24) 0.07 1.16 (1.03, 1.30) 0.01 24 Diagnosis not 1834 1858 ...... 25 classified 26 Mortality 132 (7.2%) 125 (6.7%) 1.08 (0.83, 1.4) 0.58 1.09 (0.83, 1.43) 0.54 27 First SpO2 ≥95% 8,928 9,088 ...... 28 Mortality 612 (6.9%) 601 (6.6%) 1.04 (0.93, 1.17) 0.52 1.04 (0.91, 1.17) 0.58 29 First SpO2 <95% 1,220 1,208 ...... 30 Mortality 334 (27.4%) 323 (26.7%) 1.03 (0.86, 1.24) 0.72 1.01 (0.83, 1.22) 0.92 31 32 33 Abbreviations: 34 ACS - Acute Coronary Syndromes; STEMI- ST-elevation myocardial infarction; NSTEMI – Non 35 ST-elevation myocardial infarction; SpO2 = transcapillary oxygen saturation level when 36 measured before and after first use of oxygen across the care pathway 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

7 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 36 of 49

1 2

3 Supplementary Table 3: 30 day mortality by oxygen protocol according to first SpO2 4 measured on ambulance arrival for patients managed on ambulance ACS pathway by 5 final diagnosis 6 7 High oxygen Low oxygen OR (95%CI) P OR (95%CI) P 8 N (%) N (%) Unadjusted value adjusted value

9 SpO2 ≥95% 10 All patients 12424 12538 ...... 11 Mortality 258 (2.1%) 236 (1.9%) 1.11 (0.93, 1.32) 0.27 1.09 (0.91, 1.31) 0.34 12 STEMIConfidential:1024 966 For Review.. Only...... 13 Mortality 74 (7.2%) 67 (6.9%) 1.05 (0.74, 1.47) 0.80 1.04 (0.72, 1.50) 0.84 14 NSTEMI 1474 1472 ...... 15 Mortality 62 (4.2%) 56 (3.8%) 1.11 (0.77, 1.61) 0.58 1.12 (0.77, 1.62) 0.56 16 Not ACS 7579 7781 ...... 17 Mortality 94 (1.2%) 86 (1.1%) 1.12 (0.84, 1.51) 0.44 1.12 (0.83, 1.50) 0.48 18 SPO2<95% 19 All patients 1656 1617 ...... 20 Mortality 168 (10.1%) 179 (11.1%) 0.91 (0.73, 1.13) 0.39 0.88 (0.70, 1.11) 0.29 21 STEMI 175 170 ...... 22 Mortality 40 (22.9%) 43 (25.3%) 0.88 (0.53, 1.43) 0.60 0.84 (0.49, 1.43) 0.52 23 NSTEMI 271 242 ...... 24 Mortality 32 (11.8%) 38 (15.7%) 0.72 (0.43, 1.19) 0.20 0.69 (0.41, 1.16) 0.16 25 Not ACS 995 1013 ...... 26 Mortality 65 (6.5%) 76 (7.5%) 0.86 (0.61, 1.22) 0.40 0.92 (0.64, 1.30) 0.62 27 28 Abbreviations: 29 ACS - Acute Coronary Syndromes; STEMI- ST-elevation myocardial infarction; NSTEMI – Non 30 ST-elevation myocardial infarction; SpO2 = transcapillary oxygen saturation level when 31 measured before and after first use of oxygen across the care pathway 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

8 | P a g e https://mc.manuscriptcentral.com/bmj Page 37 of 49 BMJ

1 2 3 Supplementary Table 4: Baseline characteristics of patients randomly selected for audit 4 of oxygen use, compared to other ANZACS-QI patients admitted during the study 5 periods. 6 7 ANZACS-QI Audit cohort 8 9 Number % Number % 10 11 All suspected ACS 18754 100.0 812 100.0 12 Age <68Confidential: years For9819 Review52.4 Only420 51.7 13 Age ≥68 years 8935 47.6 392 48.3 14 Sex, male 12403 66.1 533 65.6 15 Ethnic group 16 Māori 2295 12.2 94 11.6 17 Pacific 913 4.9 40 4.9 18 Asian 1463 7.8 60 7.4 19 NZ European or Other 14083 75.1 618 76.1 20 Clinical characteristics 21 Prior myocardial infarction 2363 12.6 101 12.4 22 Prior heart failure 1137 6.1 46 5.7 23 Charleson co-morbidity score (non- 24 cardiac) 25 0 14810 79 643 79.2 26 1-2 3156 16.8 139 17.1 27 3+ 788 4.2 30 3.7 28 Current Smoker 3237 17.3 149 18.4 29 Diabetes 4398 23.5 207 25.5 30 Socioeconomic Status in Quintiles+ 31 1 3061 16.3 112 13.8 32 2 3175 16.9 152 18.7 33 3 3652 19.5 142 17.5 34 4 4329 23.1 195 24.0 35 5 4347 23.2 198 24.4 36 Geographic region 37 Northern 6041 32.2 251 30.9 38 Midland 5457 29.1 238 29.3 39 Central 3641 19.4 166 20.4 40 Southern 3615 19.3 157 19.3 41 Final Diagnosis 42 STEMI 3744 20.0 159 19.6 43 NSTEMI 8991 47.9 381 46.9 44 Unstable angina 2612 13.9 118 14.5 45 Non-ACS condition 3407 18.2 154 19.0 46 Legend: 47 + Socioeconomic status determined from NZDEP2013 census data in quintiles of the New 48 Zealand general population. 49 50 Abbreviations: 51 ACS - Acute Coronary Syndromes; STEMI- ST-elevation myocardial infarction; NSTEMI – Non 52 ST-elevation myocardial infarction 53 54 55 56 57 58 59 60

9 | P a g e https://mc.manuscriptcentral.com/bmj BMJ Page 38 of 49

1 1 2 3 4 5 6 7 8 9

10 11 12 Confidential:OXYGEN THERAPY IN ACUTE For CORONARY Review SYNDROMES Only TRIAL 13 14 15 16 17 ALL NEW ZEALAND ACUTE CORONARY SYNDROMES QUALITY IMPROVEMENT (ANZACS-QI) REGISTRY TRIAL 18 TO EVALUATE TWO OXYGEN PROTOCOLS AS PART OF USUAL CARE IN PATIENTS PRESENTING WITH A 19 SUSPECTED ACUTE CORONARY SYNDROME 20 21 22 23 24 25 26 27 STATISTICAL ANALYSIS PLAN 28 29 VERSION 3. OCTOBER 2016 30 31 32 33 34 35 36 37 38 39 40

41 42 43 44 45

46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj Page 39 of 49 BMJ

2 1 2 3 Study Objectives 4 5 To determine whether giving supplemental oxygen to patients presenting with a suspected 6 7 acute coronary syndrome and clinical evidence of myocardial ischemia influences 30 day 8 mortality. 9 10 11 12 StudyConfidential: Design For Review Only 13 Cluster randomized cross over trial comparing two recommended oxygen administration 14 15 protocols as part of usual clinical care. The unit of randomization will be the four regional 16 17 cardiac clinical networks and time periods. The oxygen protocols will be recommended for 18 use in the ambulance, emergency department, CCU and cardiac catheter lab for all subjects 19 20 presenting with proven or suspected ACS. An intention to treat analysis will be performed 21 22 for all patients irrespective of oxygen received. 23 24 25 Study population 26 27 Inclusions: 28 1. All patients in New Zealand admitted to the coronary care unit, cardiac monitored 29 30 bed and/or cardiac catheter laboratory with an acute coronary syndrome (ACS) at 31 32 participating hospitals. 33 2. Patients attended by the ambulance service with a suspected ACS, who are managed 34 35 on the ACS pathway. 36 37 Exclusions: 38 1. Dead on ambulance arrival 39 40 2. Presented with an out of hospital cardiac arrest 41 42 3. Ventilated prior to admission to CCU/catheter lab 43

44 4. Documented for end of life cares 45 5. On home oxygen 46 47 6. Not admitted to CCU or catheter lab because of advanced age, co-morbidity, or 48 49 because a diagnosis other than ACS is made. (this does not exclude patient from 50 administered oxygen strategy prior to this decision being made). 51 52 53 54 About 1200 patients are admitted to NZ hospitals each month with an acute coronary 55 syndrome and included in the All New Zealand Acute Coronary Syndromes Quality 56 57 Improvement (ANZACS-QI) registry. The approximate numbers for the northern region are 58 59 400, and for midlands, central and southern regions 250 patients each month. Based on 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj BMJ Page 40 of 49

3 1 2 3 these estimates there will be ~21,000 admissions for an acute coronary syndrome to New 4 5 Zealand Hospitals over two years, with estimated 30 day mortality of 9% (n~2000). 6 7 8 9 10 Randomization 11 12 RandomizationConfidential: will be by four regional networkFor (Northern,Review Midlands, OnlyCentral and Southern). 13 This will allow oxygen administration to be consistent for patients transferred between 14 15 hospitals across the network. Any patient whose first admission is to a hospital not 16 17 participating in the trial will not be included, even if transferred to a participating hospital. 18 19 20 A cross over design is used to decrease bias which could arise because of differences in 21 22 outcomes between networks, and temporal differences not related to oxygen therapy. Two 23 networks will be randomly allocated to each of the two sequences, using computer 24 25 generated numbers. 26 27 28 It is likely there will be greater variation in the oxygen delivered during the cross over period 29 30 when the oxygen strategy is changed. Therefore patients admitted during a pre-specified 31 32 change over period (of 2-4 weeks) will not be included in the efficacy analysis. 33

34 35 For patients admitted with ACS on more than one occasion during the study, the treatment 36 37 strategy during the first admission will be assigned for the statistical analysis. 38

39 40 Timelines 41 42 Duration Two regional Two regional 43 44 networks (A+B) networks (C+D) 45 Run in phase 2-4 weeks High oxygen Conservative oxygen 46 47 First protocol 6 months High oxygen Conservative oxygen 48 49 Cross-over 2 weeks Conservative oxygen High oxygen 50 Second One year Conservative oxygen High oxygen 51 52 protocol 53 54 Cross-over 2 weeks High oxygen Conservative oxygen 55 56 First protocol 6 months High oxygen Conservative oxygen 57 58 59 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj Page 41 of 49 BMJ

4 1 2 3 Study Outcomes 4 5 Primary: 30 day all-cause mortality. 6 7 Secondary: One year all-cause mortality, length of hospital stay, Grace score on admission, 8 9 hospital readmission for cardiovascular cause and all cause at 30 days and one year. 10

11 12 The Confidential: All New Zealand Acute Coronary For Syndrome Review – Quality Improvement Only (ANZACS-QI) 13 programme is sponsored by the NZ Branch of the Cardiac Society of Australia and New 14 15 Zealand (CSANZ) and the NZ National Cardiac Clinical Network, and funded by the NZ 16 17 Ministry of Health (MOH). The goal of ANZACS-QI is to ensure that New Zealanders 18 presenting with heart disease are receiving appropriate evidence-based treatment in the 19 20 short- and long-term, regardless of age, sex, ethnicity, location, and socioeconomic status. 21 22 23 ANZACS-QI utilises two main sources of patient anonymised heart disease related data: (1) 24 25 the ANZACS-QI web-based registry, which is deployed nationally to 46 participating agencies 26 27 including 40 public hospitals and 6 private hospitals, and (2) the MOH national 28 hospitalisation and mortality datasets. For this study, all outcomes will be determined by 29 30 data linkage between the registry and national databases using encrypted NHI established 31 32 for ANZACS-QI. 33 34 35 36 37 Data Safety Monitoring Committee 38 The Health Research Council of New Zealand Data Monitoring Committee (DMC) will 39 40 monitor the trial for safety and efficacy. The DMC meets in April and October each year and 41 42 other times if necessary. Complete data on in-hospital mortality from ANZACS-QI is 43 available from hospital discharge. Data on 30 day mortality from the national mortality 44 45 registry is delayed for up to 3 months. Available data will be reviewed by the DMC at each 46 47 meeting. Outcomes data by treatment allocation will not be available to investigators until 48 49 after study completion. 50 51 52 Statistical analysis 53 54 An intention to treat analysis will be followed for all patients irrespective of oxygen received. 55 Patients admitted during a pre-specified change over period (of 2-4 weeks) will not be 56 57 included in the efficacy analysis. Multiple admissions from the same patient after first 58 59 hospitalisation during the study period will be excluded. A pre-specified subgroup analysis 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj BMJ Page 42 of 49

5 1 2 3 will evaluate outcomes for patients with ST elevation myocardial infarction and non-ST 4 5 elevation myocardial infarction. 6 7 8 Patients’ baseline characteristics will be summarized descriptively for each treatment period 9 10 and by important subgroups. Mortality rates will be presented as frequency and percentage, 11 12 and comparedConfidential: between two treatment For targets usingReview generalized linear Only regression models, 13 controlling for region and time period. The potential cluster effect will be estimated using 14 15 random effects mixed models. A similar approach will be applied to other secondary 16 17 outcomes. 18 19 20 Statistical analysis will be performed using SAS version 9.4 (SAS Institute Inc., Cary, NC) and R 21 22 version 3.3 (The R Foundation for Statistical Computing). The RCTdesign package 23 (www.rctdesign.org) will be used for group sequential design. 24 25 26 27 Interim analysis 28 For interim data monitoring on the primary outcome, a two-sided O’Brien-Fleming approach 29 30 with a symmetric stopping rule will be applied. The alpha spending function allocates the 31 32 total allowable Type I error rate (i.e. 5%) through a function based on the information 33 accrued during the trial, such as the total number of observed patients or events. That is, the 34 35 spending function depends on the fraction of patients or events observed at a particular 36 37 interim analysis out of the total number of patients or events expected or designed for. The 38 spending function that approximates O’Brien-Fleming boundary is defined as follows: 39 40 41 42  t *  tZ * )/(22)( 43 1  2/ 44 t * 45 Where denotes the information fraction. For comparing proportions, the information 46 fraction is approximated by n/N, the observed sample size n divided by the expected 47 48 maximum sample size N. For survival outcomes, this is approximated by d/D, the number of 49 50 observed deaths d divided by the total expected number of deaths D. 51 52 53 For example, if the trial will be monitored once when half of the total number of patients are 54 55 recruited with the outcome observed, the null hypothesis will be rejected at α = 0.005 (two- 56 sided). 57 58 59 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj Page 43 of 49 BMJ

6 1 2 3 Sample size 4 5 Based on recent ANZACS-QI national dataset report (Acute Coronary Syndromes in New 6 7 Zealand 2006-2013: Incidence, Coronary Intervention and Outcomes), there were 21,967 ACS 8 patients admitted to NZ hospitals in 2011 and 2012. The 28 day and one year mortality rates 9 10 are summaries below for each network and overall. 11 12 Confidential: For Review Only 13 Table 1. National ACS patients admitted to NZ hospitals in 2011-2012 14 By Region Total Northern Midlands Central Southern 15 (N=21,967) (N=7,492) (N=4,415) (N=4,840) (N=5,220) 16 n % n % n % n % n % 17 28 day mortality 2,112 9.6 717 9.6 427 9.7 475 9.8 493 9.4 18 1 year mortality 4,230 19.3 1,398 18.7 804 18.2 940 19.4 1,088 20.8 19 20 21 For the primary outcome which is a binary indicator of all-cause death within 30 days of 22 23 hospital admission, a total sample size of 28,250 would have 80% power at 5% level of 24 25 significance (two-sided) to detect a 1% difference in 30 day all-cause mortality (9.5% vs 26 10.5%) at the end of the trial. All patients admitted with ACS on more than one occasion 27 28 during the study will be assigned to the treatment strategy at the first admission in 29 30 participating hospital. With only four networks, this power calculation does not account for 31 the cluster effect, which will be estimated and adjusted as a fixed effect in final analysis. 32 33 34 35 All ACS patients registered in ANZACS-QI database, who were admitted to hospitals between 36 1st January and 31st August 2015, were linked with the latest mortality dataset from the 37 38 Ministry of Health. This includes a total of 6,269 individual patients. All-cause 30 day 39 40 mortality and in-hospital mortality rates were 3% and 2%, respectively. If we used the rates 41 in the registry database for ACS/CATHLAB patients, a total sample size of 28,250 would have 42 43 80% power at 5% level of significance (two-sided) to detect a 0.6% difference in all-cause 30 44 45 day mortality at the end of the trial. 46

47 48 49 50 51 52 53 54 55 56 57 58 59 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj BMJ Page 44 of 49

7 1 2 3 Oxygen protocols 4 5 6 7 High oxygen strategy: 8  In patients with probable or confirmed acute coronary syndrome give oxygen for 9 10 ischemic chest pain, ischemic ECG changes or dyspnea related to myocardial 11 12 Confidential:ischemia irrespective of the measured For oxygen Review saturation level. Only 13  Administer oxygen by face mask at ~8l/minute. If a face mask is not tolerated give 14 15 oxygen by nasal prongs at ~4 l/minute. Increase oxygen flow rate if necessary to 16 17 achieve saturation ≥95%. Continue oxygen until the patient is admitted to hospital 18 or transferred to cardiology, or when a doctor decides it is no longer necessary. 19 20  Caution or avoid high flow oxygen in patients at risk of hypercapnia, including 21 22 those with possible obesity hypoventilation syndrome or chronic obstructive 23 24 pulmonary disease. 25 26 27 Conservative oxygen strategy : 28 29  In patients with a suspected or confirmed acute coronary syndrome do not give 30 oxygen for ischemic chest pain, ischemic ECG changes or dyspnea related to 31 32 myocardial ischemia, unless the measured oxygen saturation is <90%. 33 34  When the oxygen saturation cannot be measured only administer oxygen if there 35 is a significant clinical concern of hypoxia. 36 37  If oxygen is administered by face mask or nasal prongs adjust the flow rate to 38 39 achieve a target oxygen saturation of 90 to 94%. 40  41 Discontinue oxygen when no longer needed to maintain saturation >90%. 42  Caution or avoid high flow oxygen in patients with possible obesity 43 44 hypoventilation syndrome or chronic obstructive pulmonary disease. 45 46  Oxygen can be administered in the ambulance to patients who need ventilation or 47 continuous positive airways pressure (CPAP). 48 49 In individual cases the oxygen protocol can be overruled by clinician preference or clinical 50 51 indication. 52 53 54 55 56 57 58 59 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj Page 45 of 49 BMJ

8 1 2 3 Documentation of oxygen delivery during the study 4 5 Recording oxygen use in the ambulance, emergency department, coronary care unit and 6 7 cardiac catheter laboratory will be encouraged as part of usual care. This will include 8 following information: 9 10 11 1. Contraindication to oxygen (eg COPD or obesity hypoventilation syndrome) 12 Confidential: For Review Only 13 2. Reason to give oxygen not by protocol (eg need for ventilation or CPAP) 14 3. Whether oxygen was given – in the ambulance, ED, Catheter Lab and CCU. 15 16 4. Last oxygen saturation before oxygen was started (x% or not known). 17 18 5. Oxygen saturation when on oxygen (x% or not known). 19 20 21 22 To encourage protocol adherence and describe differences in oxygen administered by study 23 24 group, the above data on oxygen administration and saturation levels will be audited from 25 the medical record for all patients admitted on randomly selected days (about one day 26 27 /month) with data entered to ANZACS-QI. Feedback will be provided to encourage 28 29 adherence to the specified oxygen administration regimen during the trial. 30 31 32 We would like to maintain a protocol adherence rate of 80% or higher. A sample size of 300 33 34 monitored patients will give a 95% confidence interval with 4.5% margin of error when the 35 estimated proportion of adherence is 80% (see the formula below). 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj BMJ Page 46 of 49

9 1 2 3 4 5 6 7 ACS Form CATHLAB Form - All patients admitted to hospital after 11 July 2016 - All patients admitted to hospital after 11 July 2016 8 - With Suspected ACSConfidential:- With ACS orFor suspected ACS (requestReview an ACS form) Only 9 - Type of ACS = USA / NSTEMI / STEMI - Reason for ANGIO = STEMI in 24hrs / Other suspected or known ACS 10 11 12 13 14 Oxygen Study Cohort 15 - Eligible patients with ACS and/or CATHLAB forms 16 - Completed or parked submission 17 - First episode of care (EoC) only - Region = first hospital admitted to 18 19 20 High Oxygen Strategy (Cluster A, two Regions) 21 Protocoladherence is defined for each department 22 Oxygen Audit Form (part of ACS Form) Ambulance Overall Protocol Adherence Emergency Department 23 *Wish to complete Oxygen Therapy question? (Yes = 1) Ambulance Catheter Lab - Ambulance Emergency Department 24 CCU / Cardiology Ward 25 - Emergency Department Catheter Lab - Catheter Lab CCU / Cardiology Ward 26 - CCU / Cardiology Ward 27 Conservative Oxygen Strategy (Cluster B, two Regions) - Yes = all departments ticked Yes 28 Protocoladherence is defined for each department - No = any department ticked No - Unknown = other conditions 29 *Required on randomly allocated audit dates Ambulance 30 Emergency Department 31 Catheter Lab 32 CCU / Cardiology Ward 33 34 35 36 37 38 39 40 41 42 Oxygen trial SAP version 3 October 2016

43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 47 of 49 BMJ

10 1 2 3 4 5 Protocol adherence is defined on all monitored patients as follows:

6 7 8 HIGH OXYGEN STRATEGY 9 10 The following flowchart is to be used to record data for each of the 4 departments: Ambulance/ED/Cath lab/CCU 11 12 Confidential: For Review Only

YES Was there a documented Oxygen Protocol 13 YES Oxygen Protocol reason NOT to use Oxygen? adherence 14 adherence OTHERWISE 15 (no or missing data) Was Oxygen YES Oxygen Protocol 16 saturation > 95% NA NO non-adherence 17 Was Oxygen administered? when on Oxygen? 18 Oxygen protocol UNKNOWN or adherence unknown 19 NO Oxygen protocol MISSING DATA 20 adherence unknown UNKNOWN or Oxygen Protocol 21 Ischemic symptoms MISSING DATA 22 NO or NOT adherence DOCUMENTED 23 YES 24 Oxygen protocol 25 Oxygen Protocol non-adherence MISSING DATA adherence unknown 26 27

28 Oxygen Protocol 29 adherence 30

31 32 33 CONSERVATIVE OXYGEN STRATEGY 34 The following flowchart is to be used to record data for each of the 4 departments: Ambulance/ED/Cath lab/CCU 35

36

37 YES Was there a documented Oxygen Protocol 38 reason to USE Oxygen? adherence 39 OTHERWISE (no 40 or missing data) NO/NA 41 Oxygen Protocol adherence 42 Was Oxygen administered?

43 Oxygen Protocol UNKNOWN or 44 YES adherence unknown MISSING DATA 45 YES 46 Oxygen Protocol 47 adherence 48 Was Oxygen Was Oxygen saturation YES 49 saturation Oxygen Protocol before Oxygen was started between 90-94% 50 NO non-adherence < 90%? when on Oxygen? 51 NO 52 Oxygen Protocol Oxygen Protocol 53 UNKNOWN or UNKNOWN or adherence unknown adherence unknown MISSING DATA 54 Oxygen Protocol MISSING DATA 55 non-adherence 56 57 58 59 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj BMJ Page 48 of 49

11 1 2 3 4 Appendix:New Zealand New Oxygen Zealand Study Oxygen Study 5 Use of UseOxygen of auditOxygen form audit form 6 7 Date of Audit: (given by ANZACS-QI) 8 Was there a documented reason not to use oxygen Yes No 9 (eg. COPD / obesity hypoventilation syndrome) 10 Yes No 11 Was there a documented reason to use oxygen (eg. need for ventilation / CPAP) 12 Confidential: For Review Only 13 14 1. Ambulance

15 Was oxygen administered in the Ambulance? Yes No# Unknown Not applicable+ 16 17 If yes, what was the last oxygen saturation before oxygen was started? ___% Unknown 18 (if oxygen was started before the Ambulance and continued in the 19 ambulance without stopping, record the saturation before oxygen was initially started) 20 21 If yes, what was the oxygen saturation when on oxygen? ___% Unknown 22 (Use the last recorded oxygen saturation after starting oxygen) 23 24 Ischemic symptoms Yes No or Not Documented 25 2. Emergency Department 26 Did the patient receive oxygen in ED? Yes No# Unknown Not applicable+ 27 28 If yes, what was the last oxygen saturation before oxygen was started? ___% Unknown 29 (if oxygen was started before ED and continued in ED without stopping, 30 record the saturation before oxygen was initially started)

31 If yes, what was the oxygen saturation when on oxygen? ___% Unknown 32 (Use the last recorded oxygen saturation after starting oxygen) 33 34 Ischemic symptoms Yes No or Not Documented 35 3. Catheter Lab 36 Did the patient receive oxygen when in the catheter lab? Yes No# Unknown Not applicable+ 37 38 If yes, what was the last oxygen saturation before oxygen was started? ___% Not known 39 (if oxygen was started before the Cath Lab and continued in the Cath 40 lab, record the saturation before oxygen was initially started)

41 If yes, what was the oxygen saturation when on oxygen? ___% Unknown 42 (Use the last recorded oxygen saturation after starting oxygen) 43 44 Ischemic symptoms Yes No or Not Documented 45 46 4. CCU/Cardiology ward Was oxygen given on the CCU/ward? Yes No # Unknown Not applicable+ 47 48 If yes what was the last oxygen saturation before oxygen was started? ___% Unknown 49 (if oxygen was started before CCU and continued in CCU without 50 stopping, record the saturation before oxygen was initially started, if 51 documented)

52 If yes, what was the oxygen saturation when on oxygen? ___% Unknown 53 (Use the last recorded oxygen saturation after starting oxygen) 54 55 Ischemic symptoms Yes No or Not Documented 56 57

58 59 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj Page 49 of 49 BMJ

12 1 2 3 4 User Guide: 5 6 Yes = oxygen use documented in clinical record; 7 #No = no oxygen given; 8 Unknown = not documented 9 10 +Not applicable = patient not at this location. 11 12 Confidential: For Review Only If there 13 are multiple episodes of ischemic symptoms in this location, audit the first episode of ischemic symptoms 14 -started in ED, record the last oxygen saturation before oxygen 15 was re-started. 16 17 -started in Cath lab, record the last oxygen saturation 18 before oxygen was re-started. 19 20 -started in CCU/ Cardiology ward, record the last oxygen saturation before oxygen was re-started. 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj BMJ Page 50 of 49

13 1 2 3

4 5 Changes to original statistical analysis plan after October 2016 6 7

8 9 10 1. The sample size calculation was initially estimated from 30 day mortality for all hospital 11 admissions with ACS (~9%). This was later revised to be based on 30 day mortality for patients in 12 Confidential: For Review Only 13 the ANZACS-QI registry (~3%), to be more consistent with patients included in the study 14 population 15 16 2. The electronic Ambulance ACS Pathway was implemented shortly before the study started. Data 17 from this pathway was not available when the trial was designed. The initial sample size was based 18 19 on the ANZACS-QI cohort alone. 20 3. During the study a decision was made to delay the first protocol cross-over for about 4 weeks 21 22 so protocol changes did not fall during the major summer holiday when many staff were on 23 leave. 24 25 4. The original study protocol stated that patients would be excluded if documented for end of life 26 cares, on home oxygen, or not admitted to CCU or catheter lab because of advanced age, co- 27 28 morbidity, or a diagnosis was other than ACS. However these patients could not be reliably 29 identified from registry data alone. Therefore all patients in the ANZACS-QI and Ambulance ACS 30 31 registries were included in the analysis. 32 5. Final data on cause of death was not available for this report. 33 34 6. No interim analyses were performed during the study. 35

36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Oxygen trial SAP version 3 October 2016

https://mc.manuscriptcentral.com/bmj