Background: Atrial fibrillation or flutter (AF/AFL) are the most common sustained arrhythmias, with significant public health and economic implications. Direct current cardioversion (DCC) is a standard treatment to restore sinus rhythm. Transoesophageal echocardiography (TOE) is indicated before DCC to exclude cardiac thrombi, if AF/AFL duration is uncertain or >24 hours and anticoagulation is insufficient. Current standard care for TOE-guided DCC involves a two-step sedation process: cardiologist-administered benzodiazepine sedation for TOE, followed by anaesthesiologist-assisted propofol sedation for DCC. In patients requiring subacute TOE-guided DCC (e.g., due to severe symptomatology), this anaesthesiology-dependent two-step model poses logistical challenges, leading to treatment delays, prolonged hospitalisations, increased costs, and reduced patient satisfaction. Alternative sedatives for DCC not requiring anaesthesiology assistance, such as benzodiazepines, may help address these issues. Some centres in Denmark and abroad have introduced a one-step cardiologist-only approach using continuous midazolam sedation for TOE-guided DCC, with uptake driven by assumed safety and cost-effectiveness.
While cardiologist-only sedation with midazolam has proven safe and effective for DCC in acute and elective settings, no trial has evaluated the outcomes of cardiologist-only midazolam sedation for subacute TOE-guided DCC. Correspondingly, expert opinion varies, with some expressing safety concerns about procedural sedation without anaesthesiology support. Thus, one-step cardiologist-only midazolam sedation for TOE-guided DCC, despite its increasing use in clinical practice, is being implemented without robust randomised evidence, creating genuine clinical equipoise and thereby providing the ethical basis for a randomised trial to address this critical gap in knowledge.
Aims and hypotheses: To conduct a randomised clinical trial (CARDIOZOLAM-1) to generate patient-centred and system-relevant evidence on sedation strategies for subacute TOE-guided DCC of AF/AFL. We hypothesise that one-step cardiologist-only midazolam sedation (intervention), compared with two-step anaesthesiologist-assisted propofol sedation (standard care), (1) reduces length of hospitalisation, (2) provides comparable efficacy and safety, (3) improves patient satisfaction, and (4) is associated with lower per-patient costs.
Material and methods
- Design: An investigator-initiated, multicentre, randomised, open-label, two-arm parallel-group, superiority trial (CARDIOZOLAM-1). The trial is pragmatic by design: usual treating clinicians deliver sedation according to flexible protocols, and baseline data and outcomes are captured primarily through existing electronic health record systems to ensure minimal staff burden, enhance feasibility and cost-efficiency, reduce bias from the open-label design, and improve generalisability.
- Population: Eligible patients, as defined under "Eligibility" section, will be enrolled from cardiology departments in the Central Denmark Region.
- Intervention and comparator: We will use minimised randomisation (by 5-year age group, sex, and arrythmia type) to compare the intervention with standard care, as outlined in "Arms and Interventions" section. As all patients will be sedated for TOE, participants will provide pre-TOE informed consent to post-TOE randomisation. Consenting patients will then be randomised to the intervention or standard care.
- Outcomes: The primary outcome is time-to-discharge after TOE, defined as mean hours-to-discharge, serving as a proxy for healthcare resource use. Secondary outcomes include time-to-shock delivery (measure of treatment delay), cardioversion rate (efficacy outcome), complication rate (safety outcome), and patient-reported outcomes. Detailed definitions are presented in "Outcome Measures" section. Patient-relevance of these outcomes has been evaluated through a dedicated Patient and Public Involvement activity.
- Statistical analysis: Assuming a between-person standard deviation in mean time-to-discharge of 2 hours, randomisation of 220 participants would provide ≥90% power (at a 2-sided significance level of 0.05) to detect a difference in mean time-to-discharge of 1 hour (minimum difference deemed clinically meaningful for patients and the healthcare system), even in the case of 10% cross-over between groups. Although we anticipate a larger reduction, a 1-hour improvement may determine whether discharge occurs within routine working hours or is deferred, thereby increasing the likelihood of overnight admission and avoidable bed occupancy at regional and national levels. To estimate treatment effects with 95% confidence intervals (CI), we will use linear regression to compare differences in means and log-binomial models for categorical outcomes to estimate risk ratios. Continuous outcomes with non-normal distribution will be log-transformed prior to analysis. Adjustment will be done for minimisation factors. Sample size-assumptions and statistical model choice have been informed by a feasibility study.
