Care guided by tissue oxygenation and haemodynamic monitoring in off-pump coronary artery bypass grafting (Bottomline-CS): assessor blind, single centre, randomised controlled trial

Trial design

The Bottomline-CS (better outcome through tissue oxygenation monitoring linked with intervention in cardiac surgery) trial was an assessor blinded, single centre, randomised trial comparing perioperative care guided by multisite tissue oxygen saturation and haemodynamic monitoring with usual care in patients having elective off-pump CABG. We selected off-pump CABG due to the resultant unique haemodynamic challenges, particularly the substantial disturbances caused by heart manipulation and positioning during the procedure. These disturbances, characterised by extreme hypotension and hypertension (supplementary figure 1), closely mimic real-world clinical scenarios where severe haemodynamic fluctuations risk poor outcomes for patients. Additionally, enrolment was restricted to patients undergoing off-pump CABG because more than half of CABG procedures in Asia are performed without cardiopulmonary bypass.28

Our trial received approval from the internal review board of Tianjin Chest Hospital on 30 March 2021. Tianjin Chest Hospital, a tertiary teaching hospital affiliated with Tianjin University, is located in Tianjin, China. The hospital performs over 3000 cardiac procedures annually, including more than 1600 off-pump CABGs, making it the third highest in China for CABG volume. Written informed consent was obtained from participants at least 24 h before surgery. The trial was registered at ClinicalTrials.gov (NCT04896736) before enrolment of the first patient. This report adheres to the Consolidated Standards of Reporting Trials (CONSORT) guidelines.

Participants

Patients aged 60 years or older who were scheduled for elective off-pump CABG were considered eligible. Exclusion criteria included preoperative dependence on respiratory support, use of an external cardiac assist device, urgent or emergent surgery, and an expected mortality within 30 days.

Protocol and interventions

One to two days before surgery, baseline measurements of SctO₂ at the left and right forehead and SstO₂ at the forearm brachioradialis were recorded for all participants. Tissue oxygen saturation was assessed using near-infrared spectroscopy cerebral and tissue oximetry (Nonin Medical, Plymouth, MN, USA). Measurements were conducted in hospital rooms with patients in a supine position, awake, calm, with eyes closed, and breathing either room air or their usual home oxygen. Concurrently, non-invasive estimations of cardiac output, stroke volume, systemic vascular resistance, heart rate, and blood pressure were obtained using a finger and arm cuffs-based haemodynamic monitor (continuous non-invasive arterial pressure module, Masimo LiDCO Haemodynamic Monitoring System, Masimo, Irvine, CA, USA).

In the operating room, patients’ haemodynamic parameters were routinely monitored, including continuous arterial blood pressure via a radial artery catheter, continuous central venous pressure via a central venous catheter inserted through the internal jugular vein, electrocardiography, and transoesophageal echocardiography when indicated. Anaesthesia was induced with either propofol or etomidate and maintained using propofol infusion or sevoflurane inhalation, or both, with the choice of anaesthetics determined by the attending anaesthesiologists. Cisatracurium or rocuronium was used for muscle relaxation. All patients were intubated and mechanically ventilated. Sufentanil served as the main analgesic intraoperatively and postoperatively, per institutional routine. Postoperative sedation, when required, was administered with propofol, midazolam, or dexmedetomidine at the discretion of the managing intensivists.

All trial participants received continuous multisite tissue oxygen saturation monitoring during and after surgery, using the same locations and technology as the preoperative measurements (Nonin Medical, Plymouth, MN, USA). Simultaneously, participants had continuous haemodynamic monitoring with a minimally invasive system (Masimo LiDCO Haemodynamic Monitoring System, Masimo, Irvine, CA, USA). This system estimated haemodynamic variables from the arterial blood pressure waveform transmitted from an intra-arterial catheter placed in the radial artery, differing from the non-invasive finger and arm cuffs-based technology used for preoperative assessment.

In patients assigned to guided care, the goal was to keep forehead SctO₂ and forearm SstO₂ within ±10% of individual preoperative baseline values, measured in hospital rooms 24-48 h before the scheduled CABG.24 Tissue oxygen saturation reflects the balance between oxygen delivery and consumption, with delivery determined by perfusion and arterial oxygen content. Managing physicians addressed deviations in tissue oxygenation by focusing on three domains: tissue perfusion, arterial oxygen content, and oxygen consumption (fig 1).

Fig 1

Multisite tissue oxygenation and haemodynamic monitoring intervention protocol. The protocol for managing patients having off-pump coronary artery bypass grafting using multisite tissue oxygenation and haemodynamic monitoring. Red boxes indicate haemodynamics, green indicate arterial blood oxygen content, and purple indicate tissue oxygen consumption and supply. The objective was to maintain SctO₂ at the left and right forehead and SstO₂ at the forearm within 10% higher or lower than the patient's baseline values from intubation until extubation or 24 h after surgery (A). Tissue oxygenation is determined by the balance between tissue oxygen consumption and supply (B). The figure illustrates the complex, multifactorial causes of deviations in SctO₂ and SstO₂ and emphasises the critical role of comprehensive differential diagnosis in clinical management (C). The protocol also underscores the hierarchical relationship and integration of haemodynamic parameters—including CO, SV, and SVR—into clinical decision making, ensuring a more holistic approach rather than focusing on isolated variables. Baseline haemodynamic and tissue oxygenation values were pivotal in diagnosing the underlying causes of deviations and guiding precise interventions. BP=blood pressure; CO=cardiac output; CO₂=carbon dioxide; HR=heart rate; PPV=pulse pressure variation; PRBC=packed red blood cell; RVR=regional vascular resistance; SctO₂=cerebral tissue oxygen saturation; SstO₂=somatic tissue oxygen saturation; SV=stroke volume; SVR=systemic vascular resistance; SVV=stroke volume variation

• Download figure
• Open in new tab
• Download powerpoint

Simultaneous haemodynamic monitoring, including cardiac output, stroke volume, and systemic vascular resistance, was used to diagnose the underlying causes of tissue oxygenation changes (fig 1). Rather than prescribing specific treatments, we recommended a bundle of potential interventions tailored to address the various causes of tissue desaturation and hypersaturation (ie, unwarranted increase or decrease). Providers managing trial participants were well trained, having practised the intervention protocol in at least 20 patients before the study. The intervention period began with anaesthesia induction and continued until extubation in the intensive care unit or up to 24 h postoperatively.

In patients randomly assigned to usual care, clinicians managed haemodynamics using standard monitoring, which included continuous arterial blood pressure via a radial artery catheter, central venous pressure via a central venous catheter inserted through the internal jugular vein, electrocardiography, and transoesophageal echocardiography when clinically indicated. Although these patients were also monitored with near-infrared spectroscopy tissue oximetry and LiDCO haemodynamic system, clinicians were blinded to the data from both monitors.

Outcomes

The primary outcome was a composite of one or more serious complications occurring within 30 days postoperatively. These were cerebral complications (ie, postoperative delirium, cognitive decline, or stroke), cardiac complications (ie, non-fatal cardiac arrest, myocardial injury, heart failure, or newly diagnosed symptomatic ventricular arrhythmia), respiratory failure, renal injury (ie, acute kidney injury stages II or III), infections (ie, deep surgical site or organ/space infection, pneumonia, confirmed bloodstream infection, uncertain source infection, or sepsis), and mortality. Each component required at least a Clavien-Dindo grade II classification (necessitating clinical intervention).29 Secondary outcomes comprised the individual components of the primary outcome, atrial fibrillation within 30 days, and the duration of hospital stay. Detailed diagnostic criteria are provided in supplementary table 1.

Patients were assessed twice daily after surgery until discharge. Investigators also interviewed family members and care givers to gather relevant information. Post-discharge outcomes were evaluated at a follow-up visit, typically on day 30, or by phone if an in-person visit was not feasible. Any uncertainties or discrepancies in outcome assessments were resolved through consensus.

Randomisation and blinding

Patients were randomised 1:1 to guided or usual perioperative care using computer generated allocations in blocks of four, accessed via a secure website on the morning of surgery. The individual responsible for generating the random allocations did not participate in intraoperative management or postoperative outcome assessment. Near-infrared spectroscopy tissue oxygenation and LiDCO haemodynamic data were accessible to clinicians managing patients in the guided care group, while these data were concealed from clinicians managing the usual care group.

Managing anaesthesiologists during surgery and intensivists after surgery were not masked to group assignments due to the study design. However, outcome assessors were independent, uninvolved in perioperative care, and masked to group assignments. Patients were not informed of their allocation. Oversight bodies, including the trial steering committee, remained masked throughout enrolment. The independent data monitoring committee reviewed interim data by group without knowing the group identities.

Sample size

The sample size calculation assumed that 30% of patients receiving usual care would have the primary composite outcome, compared with 24% in the group managed with multisite tissue oxygen saturation and haemodynamic monitoring—a 20% relative reduction deemed clinically meaningful. We used a one sided group sequential Z testing with three stages, including two interim analyses. Boundary values for efficacy were based on the O'Brien-Fleming α-spending function, while a non-binding β-spending function was used for futility.

The required sample size was 927 patients per group to achieve 80% power to detect a 6% absolute reduction in the primary outcome incidence (from 30% to 24%) at a type I error level of 0.05. Accounting for a 5% anticipated dropout or missing data rate, the target recruitment goal was set at 980 patients per group.

Statistical methods

We finalised a comprehensive statistical analysis plan before unblinding and analysis. Continuous variables were summarised as means with standard deviations (SD) or medians with interquartile ranges (IQR), depending on the distribution. Group comparisons for normally distributed data were conducted using t-tests, while Wilcoxon rank-sum tests were used for non-parametric data. Categorical variables were reported as counts and percentages, with group comparisons made using Pearson’s chi-squared test or Fisher’s exact test, as appropriate. Group differences were assessed using mean or median differences for continuous variables and differences in proportions for categorical variables.

We used a modified intention-to-treat population for the primary analysis, which included all randomly assigned patients with available primary outcome data, regardless of the care they received, unless consent was withdrawn.30 We used log-binomial regression to assess the primary outcome, with treatment allocation as the sole independent variable. Intervention effects were expressed as relative risks with 95% confidence intervals (CI). We defined statistical significance by a one sided Z value of less than −1.932 (P<0.03), consistent with the group sequential design.31 A one-sided test was chosen to evaluate whether guided care improves outcomes compared with usual care. We analysed secondary outcomes using generalised linear models for continuous data and log-binomial regression or chi-squared tests for categorical data. Safety analyses focused on adverse events potentially related to the intervention.

We used post hoc analyses to aid in interpreting the primary composite outcome. These analyses included standard approaches, such as combining events versus no events, event counting, and individual component analyses, as well as advanced methods like multivariate generalised estimating equations, the common effect test, and the average relative effect test.32

Alongside the unadjusted analyses described above, we also performed adjusted analyses for baseline imbalances discovered in the analysis using multivariable log-binomial regression models to account for potential confounders, as prespecified in the statistical analysis plan for primary and secondary outcomes. Baseline homogeneity was assessed with absolute standardised differences, with values exceeding 1.96×√(n_t+n_c)/(n_t×n_c) deemed to be imbalanced, where n_t and n_c represent the sample sizes of the treatment and control groups.33 Post hoc analyses included further adjustments for age, sex, body mass index, and history of diabetes and hypertension using multivariable log-binomial regression models.

Sensitivity analyses were conducted on the per-protocol population, comprising patients who adhered to the guided care or usual care protocol and had primary outcome data available. Adherence to the guided care protocol was defined as maintaining all three tissue oxygen saturation values within 10% higher or lower than the baseline for at least 80% of the surgery duration. Patients were excluded from these analyses if their equipment malfunctioned, they had substantial protocol deviations, they were deemed ineligible, or they underwent cardiopulmonary bypass instead of the planned off-pump surgery.

Additional sensitivity analyses were conducted to address missing primary outcome data. For data assumed to be missing at random, the Markov Chain Monte Carlo method was applied for multiple imputations. Twenty imputed datasets were generated by sampling from the Bayesian posterior distribution to create stationary distribution chains and simulate missing values. For data not assumed to be missing at random, analyses were performed using best-case and worst-case scenarios and selection models.

For the subgroup analyses of the primary outcome, we used the modified intention-to-treat population (excluding patients without primary outcome data). These analyses assessed interactions between the intervention and variables such as age, sex, body mass index, comorbidities, and smoking status. We also evaluated baseline physiological parameters, including tissue oxygen saturation, cardiac output, stroke volume, systemic vascular resistance, heart rate, and mean blood pressure.

The type I error level was set at 0.05 for all analyses. To account for multiple comparisons across the 24 secondary hypotheses, the Holm-Bonferroni method was applied. This stepwise procedure sequentially adjusts P value thresholds based on their rank, ensuring the family-wise error rate remained controlled at 5% while maintaining statistical rigour. Statistical analyses were performed using R (version 4.4.0, R Foundation for Statistical Computing, Vienna, Austria) with the following packages: tableone, dplyr, readxl, grid, forestploter, mice, lme4, and broom.34

Patient and public involvement

Patients were not involved in setting the research question, determining the outcome measures, or developing recruitment and implementation plans. They were also not consulted on the interpretation or reporting of the results. However, patient and family engagement throughout the trial was instrumental in ensuring its successful completion.