Blend to Limit OxygEN in ECMO: A RanDomised ControllEd Registry (BLENDER) Trial: Study Protocol and Statistical Analysis Plan

Abstract

Introduction

Critically ill patients supported with venoarterial extracorporeal membrane oxygenation (VA ECMO) are at risk of developing severe arterial hyperoxia, which has been associated with increased mortality. Lower saturation targets in this population may lead to deleterious episodes of severe hypoxia. This manuscript describes the protocol and statistical analysis plan for the Blend to Limit OxygEN in ECMO: A RanDomised ControllEd Registry (BLENDER) Trial.

Design

The BLENDER trial is a pragmatic, multicentre, registry-embedded, randomised clinical trial., registered at ClinicalTrials.gov (NCT03841084) and approved by The Alfred Hospital Ethics Committee project ID HREC/50486/Alfred-2019.

Participants and setting

Patients supported by VA ECMO for cardiogenic shock or cardiac arrest who are enrolled in the Australian national ECMO registry.

Intervention

The study compares a conservative oxygenation strategy (target arterial saturations 92–96%) with a liberal oxygenation strategy (target 97–100%).

Main Outcome Measures

The primary outcome is the number of intensive care unit (ICU)-free days for patients alive at day 60. Secondary outcomes include duration of mechanical ventilation, ICU and hospital mortality, the number of hypoxic episodes, neurocognitive outcomes, and health economic analyses. The 300-patient sample size enables us to detect a 3-day difference in ICU-free days at day 60, assuming a mean ICU-free days of 11 days, with a risk of type 1 error of 5% and power of 80%. Data will be analysed according to a predefined analysis plan. Findings will be disseminated in peer-reviewed publications.

Conclusions

This paper details the protocol and statistical analysis plan for the BLENDER trial, a registry-embedded, multicentre interventional trial comparing liberal and conservative oxygenation strategies in VA ECMO.

1. Introduction

In patients with cardiac arrest, traumatic brain injury, and stroke and general hospitalised patients, hyperoxia (defined as arterial oxygen level >100 mmHg or severe if >300 mmHg) may worsen clinical outcomes.^1,2 Hyperoxia can cause injury through multiple mechanisms, including increases in free radical production, inflammation, coronary and cerebral vasoconstriction, and by worsening ischaemia reperfusion injury.³

Patients treated with venoarterial extracorporeal membrane oxygenation (VA ECMO) for cardiogenic shock or after cardiac arrest have a high risk of organ failures and ischaemia-reperfusion injury. An oxygen fraction of 1.0 is commonly delivered into the ECMO circuit, resulting in high arterial oxygen levels. These patients are thus potentially vulnerable to adverse effects of hyperoxia. In a recent study of 1187 VA ECMO patients, the mean partial arterial pressure of oxygen (PaO₂) at 24 h was 180 (±114) mmHg (normal range 80–100 mmHg) when VA ECMO was used for cardiogenic shock and 195 (±137) mmHg when used after refractory cardiac arrest.⁴ In recent observational studies, hyperoxia has been shown to be associated with increased mortality in VA ECMO patients.[5], [6], [7]

However, the targeting of lower oxygen targets in VA ECMO patients may result in harmful episodes of hypoxaemia (PaO₂ <80 mmHg). Several recent trials in critically ill patients who did not require VA ECMO have suggested targeting lower saturations may result in no overall benefit[8], [9], [10] or potential harm.^11,12 Despite this, many centres throughout the world, including Australia and New Zealand, now routinely deliver lower percentages of oxygen into the ECMO circuit, without clear proven safety of the intervention or definitive evidence of benefit.

The BLENDER trial is a phase-2, registry-embedded, comparative effectiveness trial that compares two strategies of oxygenation in VA ECMO patients. The objective is to determine whether in patients undergoing VA ECMO, a strategy of conservative oxygen targets, as compared to a liberal oxygen strategy, improves intensive care unit (ICU)-free days at day 60.

2. Methods and analysis

2.1. Trial design and setting

The BLENDER trial is an investigator-initiated, phase-2, multicentre, parallel group, comparative effectiveness trial comparing two different oxygenation strategies in patients on VA ECMO. The first patient was randomised in September 2019, and patient recruitment is estimated to be completed by June 2023. Final collection of all 180-day outcome data is anticipated by December 2023. A total of 300 adult patients will be enrolled from Australian ECMO centres. Coenrolment in other studies is possible as long as they are observational or do not directly affect the interventional pathway.

Clinical trial registration: NCT0384108

2.2. Approvals

The trial is approved by the Alfred Ethics committee (Project 88/19) and is registered with ClinicalTrials.gov (NCT03841084).

2.3. Population

Adult patients commenced on VA ECMO and are eligible to be randomised within 6 h of ECMO initiation. For patients who transitioned from bypass to ECMO in the operating theatre, they are eligible for randomisation within 6 h of leaving the operating theatre. The complete eligibility criteria are listed in Table 1.

Table 1.

Inclusion and exclusion criteria.

Inclusion criteria

•All adult patients commenced on VA ECMO for severe cardiac failure or following refractory cardiac arrest

•Patient will be enrolled in the EXCEL registry

Exclusion criteria

•Age less than 18 years

•Patients who are suspected or confirmed to be pregnant

•The patient is already enrolled in another oxygen titration study (unless agreed by study committees)

•Patients not willing to receive blood products (e.g., Jehovah's Witness)

•Where an indication exists for a specific oxygen target as part of clinical care (e.g., carbon monoxide–poisoning)

•Six or more hours have elapsed from the time of initiation of ECMO to randomisation

•Where the treating physician deems the study is not in the patient's best interest

EXCEL: Australian national ECMO registry; VA ECMO: venoarterial extracorporeal membrane oxygenation.

2.4. Screening and recruitment

All patients who are commenced on VA ECMO at a participating trial site are considered for participation (screened). Patients are eligible if they fulfil all inclusion criteria and none of the exclusion criteria. Inclusions and exclusions (including reasons for exclusion) are reported according to the Consolidated Standards of Reporting Trials statement.¹³

2.5. Randomisation and allocation concealment

Staff members at trial sites have 24-h access to a secure, web-based central randomisation interface Research Electronic Data Capture.¹⁴ A permuted block randomisation method is used with variable block sizes, stratified by site and indication (VA ECMO for cardiogenic shock or after refractory cardiac arrest). Participants are enrolled by delegated ICU doctors, nurses, and research staff at each site, and the assigned intervention is commenced immediately. Following randomisation, each patient is assigned a unique patient study number, and their study intervention allocation is revealed.

2.6. Blinding

Staff members conducting long-term outcome measurements are blinded to allocation status, and analyses will be undertaken by blinded data managers and statisticians. Blinding of care providers and patients in the ICU is not possible, but clinicians who care for the patients following discharge from the ICU are not informed about allocation.

2.7. Interventions

Participants are randomised to either a conservative or liberal oxygen strategy while receiving VA ECMO. Once VA ECMO is discontinued, protocolised oxygen administration is ceased (Fig. 1).

Fig. 1.

Conservative arm flowchart.

FbO₂ = Inspired oxygen fraction delivered to the oxygenator; FiO₂ = inspirated fraction of oxygen in the ventilator; SpO₂ = peripheral oxygen saturations (as measured by plethysmography or arterial blood samples); ABG = arterial blood gas.

2.7.1. Conservative oxygen target strategy

Participants allocated to the conservative strategy have the inspired oxygen fraction (FbO₂₎ delivered to the oxygenator in the ECMO circuit reduced to 0.6. This is then titrated between a minimum FbO₂ 0.5 and a maximum of FbO₂ 1.0 to achieve a postoxygenator ECMO circuit blood saturation of 92–96%.¹⁵ Postoxygenator arterial blood gases are taken as required for clinical management and a minimum of twice per day as per the BLENDER protocol. The ventilator FiO₂ is set at 0.6 and is then titrated (minimum FiO₂ 0.21) to achieve arterial saturations of 92–96% on plethysmography or arterial blood samples taken from the right arm. Patients who are not receiving invasive or noninvasive mechanical ventilation while on VA ECMO have nasal or oral oxygen delivered at a flow rate that achieves right-arm saturations of 92–96% (Fig. 2).

Fig. 2.

Conservative arm – response to hypoxia flowchart.

FbO₂ = Inspired oxygen fraction delivered to the oxygenator; FiO₂ = inspirated fraction of oxygen in the ventilator; SpO₂ = peripheral oxygen saturations (as measured by plethysmography or arterial blood samples); ABG = arterial blood gas.

2.7.2. Liberal oxygen target strategy

Patients allocated to the liberal strategy have the FbO₂ set at 1.0 at all times, and postoxygenator arterial blood gases are taken twice per day. The ventilator FiO₂ is titrated to right arm saturations of ≥97–100% (but not lower than 0.5). Patients who are not receiving invasive or noninvasive mechanical ventilation while on VA ECMO have nasal or oral oxygen delivered at a flow rate that achieves right-arm saturations of ≥97–100%. (see Fig. 3, Fig. 4)

Fig. 3.

Liberal arm flowchart.

FbO₂ = Inspired oxygen fraction delivered to the oxygenator; FiO₂ = inspirated fraction of oxygen in the ventilator; SpO₂ = peripheral oxygen saturations (as measured by plethysmography or arterial blood samples); ABG = arterial blood gas.

Fig. 4.

Liberal arm response to hypoxia.

FbO₂ = Inspired oxygen fraction delivered to the oxygenator; FiO₂ = inspirated fraction of oxygen in the ventilator; SpO₂ = peripheral oxygen saturations (as measured by plethysmography or arterial blood samples); ABG = arterial blood gas.

2.8. Concomitant care

All other aspects of intensive care management are according to individual unit practice and local guidelines, with no restrictions placed on concomitant care.

2.9. Outcomes

The primary outcome is the number of days alive and free of ICU at day 60. ICU-free days is defined as the total number of days (or part days) being free of the ICU between randomization and day 60, with the exception that all patients who die before day 60 are defined as having zero ICU-free days. Details of all outcome measures are listed in Table 2. The specific elements of the composite outcomes will be reported in the primary publication.

Table 2.

Primary, secondary, and tertiary outcome measures.

Primary outcome

•Number of days alive and free of ICU at day 60

Secondary outcomes

•In-hospital mortality

•ICU mortality

•Hospital length of stay

•ICU length of stay

•Duration of mechanical ventilation

•Mortality at 6 months

•Disability at 6 months measured with the World Health Organization's Disability Assessment Schedule 2.0 12 L

•Quality of life measured at 6 months with the Euroquol 5 questionnaire

•Neurocognitive function assessed at 6 months with the Montreal Cognitive Outcomes Assessment

•Functional outcomes as assessed using the Modified Rankin Score

Tertiary outcomes

•Cost-effectiveness

ICU: intensive care unit.

2.10. Sample size calculation

The sample size calculation is derived from pilot data, where the overall mean ICU-free days till day 60 (in survivors) was 11 days (standard oxygen strategy). Using a type I error rate of 0.05 and a 3-day difference between the groups in ICU-free days, recruiting 124 patients per group will achieve 80% power to detect the specified difference. To account for the fact that ICU-free days will not be normally distributed, the sample size was inflated by 15% to 286 patients overall. This was further inflated to 300 patients to account for any drop out.

2.11. Statistical analysis plan

For the full statistical analysis plan, see online supplement.

2.12. Safety

Patients on VA ECMO experience a number of aberrations in laboratory values, signs, and symptoms due to the severity of the underlying disease and the impact of ECMO support. These do not necessarily constitute an adverse event or a serious adverse event unless they are considered to be related to study treatment or a concern in the site principal investigator's clinical judgement. Given the baseline severity of illness of patients receiving VA ECMO and an expected baseline in-hospital mortality of 50%, death is not automatically considered as an adverse event unless assessed as potentially related to the study intervention. In the BLENDER study, potential adverse events are reported and classified by their likely relationship to the study intervention (possible, probable, or definite).

2.12.1. Serious adverse events

Serious adverse events will include but not be limited to the following:

• •Seizure (diagnosed via clinical or electroencephalogram - EEG)
• •Cardiopulmonary resuscitation during ECMO
• •Any critical hypoxic episode (SpO₂ < 80% lasting for 5 or more minutes) requiring an emergency response

Serious adverse events are recorded daily in the electronic case report form during the ICU stay.

2.13. Data Monitoring and Safety Committee, medical monitor, and interim analyses

All adverse events are reviewed by an independent medical monitor, who decides if these require “out of session” notification to the Data Monitoring and Safety Committee (DMSC). Those that do not are subsequently notified at the routine DMSC review. The DSMC can recommend pausing or stopping the trial as per the DSMC charter.

The first DSMC meeting involved a planned analysis after enrolment of the first 50 patients. This was to verify adequate separation of oxygen exposure between treatment groups and satisfactory protocol compliance. A further planned interim analysis solely for safety was performed once the 60-day outcome data for 150 participants was available. Following this meeting on 26th April 2022, the DSMC approved to continue recruitment to the final target sample size.

2.14. Patient withdrawal

Patients can withdraw from the trial if

• •A clinical indication arises requiring withdrawal from the study protocol (judged by the responsible clinician or local investigator [e.g., change to a different mode of ECMO delivery such as venovenous ECMO])
• •The patient or next of kin withdraws consent

Patients who are withdrawn from the BLENDER trial will continue to have routine data collection for the purposes of the Australian national ECMO registry (EXCEL) and will still receive follow-up from the EXCEL registry (see below). In the case of withdrawal of consent to participate in the study by the patient or next of kin, the withdrawn patient's data will not be transferred for analysis in the BLENDER study. If a patient wishes to withdraw from the BLENDER trial after data transfer (from EXCEL to BLENDER), his/her data will be removed from all analyses. Patients who are withdrawn from the study at the request of clinicians will still be included in the intention-to-treat analysis.

2.15. Data management and quality

Demographic, diagnostic, illness severity, intervention, and outcome data (survival, length of stay, duration of interventions, and functional status) are collected in the EXCEL registry, a comprehensive binational research registry coordinated by the Australian and New Zealand Intensive Care Research Center, Monash University. These data are collected independently of the BLENDER trial and are made available to the BLENDER investigators under the standing agreements between both management committees. Randomisation data (including intervention allocation, serious adverse events, and protocol deviation data) are collected specifically by site research staff and entered into BLENDER-specific electronic case report forms, which are linked through encrypted patient identification numbers to the data from the EXCEL registry.

Data quality and protocol standardisation is optimised by arranging a start-up meeting and by providing an early monitoring visit and regular feedback to each centre via phone, the trial web-site, and a regular newsletter update.

2.16. Monitoring

The study will be conducted in accordance with the International Conference on Harmonisation Guidelines for Good Clinical Practice and with all relevant local and national regulations.¹⁶ The coordinating centre monitor will visit each study centre on several occasions during the recruitment phase. This will ensure that the study is conducted according to the protocol, good clinical practice guidelines, and relevant regional regulatory requirements. The study monitor will review study records for source document verification, confirm valid patient consents, data quality, and the completeness of follow-up. In addition, the study may also be audited by local or national regulatory authorities. Access to source documents and other study files will be made available at all study centres for monitoring and auditing purposes. Due to the impact of COVID, some of the site visits have been conducted remotely. The coordinating centre team will conduct regular remote monitoring on the web-based database by applying validation and consistency rules and with regular data cleaning to ensure the integrity of the study data.

3. Ethics and dissemination

This study is endorsed by the Australian and New Zealand Intensive Care Society Clinical Trials Group and ECMONet. The study was initially approved by The Alfred Hospital Research and Ethics Committee project ID HREC/50486/Alfred-2019 (Local Reference: Project 88/19). The study is also approved in other Australian States and through the respective governance committees at all participating sites, specifically, New South Wales (2020/PID00136), Queensland (Project ID 50486), South Australia (Governance Reference 13337), and Western Australia (RGS0000004380).

By virtue of the inclusion criteria, the immediate emergency nature of VA ECMO provision, and that all patients should be enrolled and randomised as soon as possible, it is not practicable to obtain informed consent before enrolment. Consent to continue in the trial is sought as soon as practical. The participant and/or their legally authorised representative are informed of the participant's inclusion in the research and are provided with an information brochure which details the nature of the research project, their participation, and the procedure to withdraw from the research and provision of contact details for those who wish to request additional information.

3.1. Collaborators

This trial has been embedded in the EXCEL registry and developed in collaboration with, and endorsed by, the Australian and New Zealand Clinical Trials Group and ECMONet,

3.2. Finance

The trial is funded by the Medical Research Future Fund (MRFF GNT1152270). The funding source will have no influences on trial design, trial conduct, data handling, data analysis, or publication.

3.3. Trial status

The BLENDER Trial began recruiting patients in September 2019, with recruitment estimated to be complete by June 2023 with a total of 300 patients. Final collection of all 6-month outcome data is expected by December 2023. The trial had 241 patients recruited till November 2022.

4. Discussion

4.1. Rationale

Hyperoxia in patients treated with VA ECMO has been associated with an increased risk of mortality. However, current practice commonly results in significant arterial hyperoxia. In addition, many VA ECMO patients have concomitant respiratory failure, so that any reduction in saturation targets may result in an unacceptably high risk of hypoxia. Currently, there is significant variation in practice in oxygen targets used at ECMO centres around the world. The question of the optimal target saturation has never been tested in a large, high-quality, randomized controlled trial.

4.2. Population

The population constitutes adult patients being treated on VA ECMO with few exclusion criteria and includes patients with cardiogenic shock and post–cardiac arrest patients (prespecified subgroups).

4.3. Intervention (conservative vs. liberal oxygen target)

The intervention in this trial was developed and modified from the ICU ROX trial,⁸ which compared two different oxygenation strategies in mechanically ventilated ICU patients. BLENDER is a comparative effectiveness trial, which investigates two different oxygenation strategies that are currently used in clinical practice. The BLENDER intervention differs from the ICU ROX protocol as the fraction of inspired oxygen delivered to the oxygenator of the ECMO circuit is adjusted, in addition to adjusting the fraction of inspired oxygen delivered by mechanical ventilation. The intervention is applied as soon as possible, but no later than 6 h after initiation of ECMO. At the midpoint review by the DSMC, the trial protocol had resulted in successful separation between the two arms, a high level of compliance, and no safety concerns.

4.4. Outcome

ICU-free days is an important phase-2 trial outcome. It has several advantages, including being already validated in the ICU population, it can capture differences in the trajectory of recovery from critical illness including long term recovery, and it has been recommended by the Australian and New Zealand Clinical Trials Group.¹⁷ It also reflects patient-level clinical and economic outcomes that have been used in other trials.¹⁸ The sample size has been inflated to account for drop out and the potentially non-normal distribution of the data. We have adequate power to detect the 3-day difference in ICU-free days, which we believe is a realistic and clinically and economically meaningful treatment effect.

4. Strengths

The BLENDER trial is the first, large, randomized controlled trial with a low risk of bias to give high-quality data on the optimal oxygen target during VA ECMO. It is multicentre and has minimal exclusion criteria, making the results generalisable. The trial is monitored according to best-practice guidelines standards, and data burden and costs are minimised through embedding in the national ECMO registry. Although the patients and clinicians cannot be blinded, all outcome assessors, statisticians, the management committee, and the authors are blinded to the group allocation. The study design is pragmatic, with routine clinical practice continuing, also improving its generalisability.

5. Weaknesses

As described, the lack of blinding potentially introduces some bias. The primary outcome measure of ICU-free days at day 60 may be not effective if there is significant discharge delay in leaving the ICU (defined as patients who are medically suitable for ward discharge but awaiting transfer). However, the time when the patient is ready for ward, rather than the actual discharge time, will be used to reduce this risk. Additionally, some patients who are commenced on ECMO at nonstudy sites who are later transferred will be excluded if they do not meet the 6-h cut-off time to enter the study. This decision was taken to ensure the intervention is commenced as soon as possible.

6. Conclusion

The BLENDER trial will provide important, high-quality data, and the results will inform clinical practice globally and potentially yield economic gains. This paper details the detailed protocol, analysis plan, and current recruitment status of the BLENDER trial. This supports the unbiased analyses of the clinical data collected.

CRediT authorship statement

Aidan Burrell: Conceptualization, Methodology, Investigation, Writing - Original Draft, Visualization, Supervision, Funding acquisition. Sze Ng: Writing - Review & Editing, Resources, Project administration. Kelly Ottosen: Resources, Project administration. Michael Bailey: Methodology, Formal analysis, Funding acquisition. Hergen Buscher: Investigation, Supervision. John Fraser: Investigation, Supervision, Funding acquisition. Andrew Udy: Investigation, Funding acquisition. David Gattas: Investigation, Funding acquisition. Richard Totaro: Investigation, Supervision. Rinaldo Bellomo: Investigation, Supervision. Paul Forrest: Investigation. Emma Martin: Investigation. Liadain Reid: Resources, Project administration. Marc Ziegenfuss: Investigation. Glenn Eastwood: Investigation. Alisa Higgins: Writing - Review & Editing, Supervision, Funding acquisition. Carol Hodgson: Writing - Review & Editing, Supervision, Funding acquisition. Edward Litton: Investigation, Supervision. Priya Nair: Investigation. Neil Orford: Investigation, Supervision. Vince Pellegrino: Investigation. Kiran Shekar: Investigation. Tony Trapani: Resources, Supervision. David Pilcher: Conceptualization, Methodology, Investigation, Writing - Original Draft, Visualization, Supervision, Funding acquisition.

Competing interests

Authors Rinaldo Bellomo, David Pilcher, Carol Hodgson, Andrew Udy and Ed Litton declare a conflict of interest as Editors of this journal.

Appendix A. Supplementary data

The following are the Supplementary data to this article: