Clinical trial protocol adherence

This protocol conforms to the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guidelines.13

Study protocol

We have designed a pragmatic randomised trial to evaluate whether HFNC is superior to usual care at reducing the risk of reintubation after adult cardiac surgery, compared with usual care. To enhance pragmatism, we have prioritised broad inclusion and exclusion criteria.14 Adult (≥18 years old) patients undergoing cardiac surgery at Vanderbilt University Medical Center who arrive at the cardiovascular intensive care unit with an endotracheal tube in place and on mechanical ventilation will be eligible for inclusion, from 1 November 2021 until enrolment targets met; we anticipate that this will take approximately 36 months. Exclusion filters have been generated to exclude cases performed in procedural areas and non-operating room areas, including the catheterisation and electrophysiology suite, as well as additional filters to exclude procedures with a total anaesthesia time of less than 3 hours.

Screening, enrolment and randomisation will be performed using tools embedded within our institution’s Epic electronic health record (Epic Systems, Verona, Wisconsin). Randomisation is performed using a random number that is generated and associated with the patient’s hospital admission medical record, but not visible to clinicians or investigators, at the documentation of ‘Anaesthesia Stop’ at the conclusion of the surgical procedure, provided patients meet inclusion criteria, using Epic’s ‘Rule Editor’ tool. There is no stratification to randomisation.

Enrolment is performed using Epic’s order set function. At the time an order is placed to extubate the patient after surgery, an order set is automatically selected based on the previously generated random number. According to hospital policy, patients may not be extubated without an extubation order. The selected order set will default to either treatment with usual care or treatment with HFNC following extubation. The clinical care team will not be blinded to treatment arm following enrolment, as blinding cannot practically be performed.

Respiratory therapists have been instructed to target 24 hours of HFNC therapy, if clinically appropriate in their judgement and that of the patient care team, with an absolute minimum of 1 hour of therapy. Respiratory therapists may use any HFNC device, at their discretion, with initial settings and titration at the discretion of the clinical team. Escalation of care beyond HFNC and transition to alternative means of oxygenation will be at the sole discretion of the treatment team. Similarly, usual care is at the discretion of the clinical team and may include HFNC, if deemed appropriate by the treating clinicians. Typical treatment modalities in usual care include, but are not limited to, nasal cannula, face mask, HFNC and bilevel positive airway pressure.

Some patients may not receive an extubation order, such as instances of early postoperative mortality with the endotracheal tube still in situ or self-extubation. We will plan to report relevant data on patients who were randomised but not enrolled as a secondary analysis to help ensure that postrandomisation exclusions do not introduce bias. All concomitant care and interventions will be permitted to continue to ensure pragmatism.

The decision to reintubate a patient will be at the sole discretion of the treatment team; there are no protocolised criteria for reintubation.

Patient and public involvement

Patients and the public will not be involved in the design, conduct or reporting plans of our research.

Sample size calculations

Assuming a 3.2% reintubation rate among the control arm1 and a 1:1 ratio of intervention patients to control patients, with 1436 patients in each group (2872 patients total), a two-sample test of binomial proportions achieves 80% power to detect a relative decrease of 50% in the 48-hour reintubation rate among the HFNC arm (ie, from 3.2% to 1.6%) at a type 1 error rate of 5%.15 Accounting for at least 10% crossover between groups, we will require 3192 patients enrolled for the full trial (1596 in each arm). The intervention and primary outcome are in hospital; as such, we do not anticipate any drop out.

Analysis plan

Statisticians and investigators will be blinded to group allocation throughout the study and subsequent analysis. Unblinding may occur at the request of the Data Safety Monitoring Board (DSMB). The dataset will be finalised and locked prior to unblinding and final analysis. The primary outcome is all-cause reintubation within 48 hours of initial extubation after surgery. Secondary outcomes include all-cause 30-day mortality, hospital length of stay, intensive care unit length of stay and ventilator-free days. Interaction analyses will be conducted to assess the differential impact of the intervention among strata defined by predicted risk of reintubation, based on our recent prognostic model of risk of reintubation after cardiac surgery.16 All data analysed, including demographic and outcome data, will be limited to data generated by clinicians in routine clinical care and captured via query of existing databases available within the electronic health record. In previous work, we have demonstrated that we are able to obtain these data without any missingness, based on extensive experience querying the electronic health record, as well as limiting our dataset to those data elements that are reliably available.1 We intend to analyse patients on a modified intention to treat basis. Patients randomised but not enrolled, for example, those who self-extubate after they have already received a randomisation number or who die prior to extubation will be added to the primary cohort and analysed in a prespecified secondary analysis.

We will summarise continuous variables using median and IQR and categorical variables using proportions and frequencies. We will quantify the difference in reintubation rates between the groups using absolute risk difference with 95% CIs. The effect of the intervention on the primary outcome will be assessed using a χ² test. Depending on event rate, a χ² or Fisher’s exact test will be used to analyse 30-day mortality. Mann-Whitney U tests will be used to compare the remaining continuous and ordinal secondary outcomes. We have not prespecified adjustment for covariates in multivariable models as our primary analysis. We will consider post-hoc adjustment, should there be any clinically relevant imbalance in baseline characteristics between groups. A two-sided p-value of <0.05 will be used to determine statistical significance for the primary outcome.

Pragmatic considerations

As this is a pragmatic study, no steps will be taken to facilitate enrolment, retention or adherence to the intervention, aside from routine updates to the respiratory therapist, nursing, physician and advance practice provider teams, on the progress of the study and availability of the principal investigator should there be any questions or concerns. Aside from the initial intervention, HFNC therapy, no additional interventions or follow-up will be performed, outside of usual care. The initial settings, titration and transition to alternative means of respiratory support will all be left to the discretion of the treatment team. All documentation and analysis will be limited to data generated by clinicians in routine patient care and available in the electronic health record. No processes will be performed to manipulate data entry or data quality.

Randomisation and enrolment targets will be tracked in real-time using a custom web-based application displaying data from the electronic health record, built using Tableau software (Tableau, Seattle, Washington). This same application will be used to track adherence to the intervention.

As this intervention was adjudicated to constitute ‘minimal risk’ by the IRB, there are no plans for post-trial care.

Data safety plan

A DSMB will act in an advisory capacity to monitor participant safety, data quality and progress. The DSMB will be responsible for reviewing composite data, separated by groups, with a focus on ensuring patient safety. Unmasked data will be provided if safety concerns arise requiring knowledge of the group assignments. At each meeting, the DSMB will consider the rationale for continuation of the study, with respect to progress of randomisation, retention, protocol adherence, data management, safety issues and make a recommendation for or against the trial’s continuation. The DSMB will adjudicate any reported instances of harm and determine appropriate corrective action, if warranted. The DSMB will consist of persons independent from investigators, with no conflict of interest in the trial. The DSMB will consist of experts in cardiovascular critical care, biomedical ethics and biostatistics, with representation from outside of Vanderbilt University Medical Center. The DSMB will meet once 500 patients have been enrolled in the study for 30 days, as well as when 1500 patients have been enrolled in the study for 30 days. An unscheduled meeting of the DSMB may be called at any time should participant safety questions or other unanticipated problems arise that require the investigators to consult with the DSMB. The primary outcome for the DSMB interim safety analysis will be all-cause mortality, with a stopping boundary of p<0.05. A secondary analysis of all-cause 48-hour reintubation will be performed, with a Haybittle-Peto stopping boundary of p<0.001. The DSMB will prepare a report at the conclusion of each meeting containing the recommendation for continuation of the study. Each meeting must include a recommendation to continue the study made by a formal DSMB majority. A recommendation to terminate the study may be made by the DSMB at any time by majority vote.