Trial overview and design

The CAffeine, Postoperative Delirium And CHange IN Outcomes after Surgery (CAPACHINOS)-2 Trial is a randomised, placebo-controlled clinical trial conducted at Michigan Medicine. The trial has been approved by the University of Michigan Medical School Institutional Review Board (HUM00218290), and written consent will be obtained from all trial participants. Additional consent provisions will also be provided for data sharing and future research. The trial was registered on ClinicalTrials.gov (NCT05574400, 10 October 2022, principal investigator (PI): PEV) and the complete trial protocol is available in online supplemental appendix 1. CAPACHINOS-2 also meets criteria put forth by the Standard Protocol Items for Randomised Trials (online supplemental appendices 2 and 3 for written informed consent document). Any required protocol amendments will be communicated to the University of Michigan Medical School Institutional Review Board, National Institute on Aging and data and safety monitoring board (DSMB). Amendments will also be logged in the full clinical trial protocol (online supplemental appendix 1).

The trial will follow a parallel arm design, with a 1:1:1 allocation ratio (placebo: 1.5 mg/kg caffeine; 3 mg/kg caffeine) (figure 1). Participants will be block-randomised, and randomisation will also be stratified by age (70–74 vs ≥75 years of age) and sex. The randomisation scheme was developed by the study statistician team (GM and EJ), and the randomisation schedule will be managed by the hospital research pharmacy. The study will follow a quadruple-blinded design: participants, researchers, clinicians and analysts will be blinded to the intervention.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Consolidated Standards of Reporting Trials flow diagram presented. The design is a three-arm parallel design with a 1:1:1 allocation ratio (placebo: low-dose caffeine: high-dose caffeine). D5W, dextrose 5% in water; EEG, electroencephalography.

Eligibility criteria

Patients ≥70 years of age presenting for major non-cardiac, non-intracranial, non-major vascular (eg, operations above the diaphragm) surgery with anticipated hospital length of stay at least 48 hours will be eligible for the trial. Exclusion criteria include the following: emergency surgery, outpatient surgery, severe cognitive impairment precluding capacity for informed consent, seizure disorder history, intolerance or allergy to caffeine, weight >130 kg (as a 3 mg/kg dose would approach the upper limit of daily intake recommended by the Food and Drug Administration), enrolment in a conflicting study, acute hepatic failure (inadequate caffeine metabolism), acute kidney injury preoperatively (which may impair caffeine clearance), diagnosis of pheochromocytoma (to avoid unsafe increases in blood pressure), severe audiovisual impairment preventing participation in cognitive function testing and non-English speaking.

Interventions

Participants will be block randomised with a 1:1:1 allocation ratio (placebo: 1.5 mg/kg caffeine: 3 mg/kg caffeine) in a three-arm parallel design as previously described. Prepared intravenous syringe solutions of dextrose 5% in water (D5W) placebo or caffeine citrate will be directly delivered to the operating room prior to the surgery of enrolled participants. For the next two postoperative mornings, the study drug will be given with scheduled, morning medications between approximately 8:30–9:30 hours as administered and overseen by a research team nurse.

Each dose was chosen based on preliminary data and literature review. First, the lower dose (1.5 mg/kg, approximately one cup of coffee24) might lower the risk of adverse side effects (eg, nausea, vomiting, anxiety).22 Second, reduced dosing may also be appropriate for older populations, as the cognitive effects of caffeine may be more pronounced with age.25 Serum caffeine concentration also increases with age, as the total volume of caffeine distribution is reduced in older patients.26 As such, higher peak serum caffeine concentrations are observed with advancing age after both intravenous and oral administration.26 The higher dose of 3 mg/kg (approximately two cups of coffee24) might provide added cognitive benefit.27 In fact, the mean weight-based dose for participants receiving caffeine and not experiencing any delirium in our preliminary trial was 3 mg/kg.16 Additive cognitive benefit has also been observed with this dose, particularly for domains relevant to delirium (eg, attention, vigilance)28–31 and after sleep deprivation,32 which is commonly experienced in the hospital setting. Nonetheless, this higher dose may also be associated with additional side effects (eg, anxiety, gastrointestinal distress, nausea/vomiting), so multiple dosing arms will help determine optimal dosing to maximise benefit-to-risk ratios.

The study drug will be administered as an intravenous infusion, using an infusion pump, over 30 (±5) min at three time points: (1) the beginning of surgical closure during the operation, (2) first postoperative morning and (3) second postoperative morning. The lower dose will consist of 1.5 mg/kg caffeine base, and higher dose will contain 3 mg/kg caffeine base; both caffeine drugs will be dissolved in 40 mL of D5W. No dose or timing changes are anticipated, and the infusion will be administered over the entire 30 (±5) min. A research team nurse will oversee the drug administration and monitor blood pressure (every 5 min), heart rate, heart rhythm (via five-lead electrocardiography), and continuous pulse oximetry for the full 30 min infusion. This monitoring will continue for 10 min after the infusion, then monitoring will continue per standard clinical protocols for each hospital unit. The study drug will be prepared, handled, and stored per standard hospital pharmacy protocols as described in the complete trial protocol (see online supplemental appendix 1). Clinical teams will be notified of any adverse drug reactions, and the infusion will be stopped for any concerns or adverse events suspected in relation to the drug infusion.

Outcomes

Timing of outcomes is presented in figure 2, which illustrates participant flow through the trial. The primary outcome is delirium, rated using the long-form Confusion Assessment Method (CAM).33 The CAM will be rated based on a brief cognitive screen that includes testing of sustained attention, short-term recall, orientation and the Delirium Symptom Interview as previously described.34–36 Signs of acute change as reported by family members, care partners and/or clinicians will also be incorporated into the CAM ratings. This cognitive testing and CAM rating will occur at preoperative baseline, 1 hour after PACU admission, and twice daily for the first three postoperative days. A daily chart review method will be performed to complement CAM screening,37 which will be particularly useful if in-person CAM screening is not possible for a given day. Research team members will undergo formal CAM training, which will entail video training (available via American Geriatrics Society CoCare website: https://help.agscocare.org) and educational sessions with PEV and AM, as our team has used this training programme for previous trials.2 38–40 After this initial training, trainees will then accompany fully trained team members during CAM interviews, and trainees will independently conduct their own CAM assessment. Trainees will need to achieve agreement on final CAM scores (ie, delirium or no delirium) for two non-delirious and two delirious participants before independently assessing trial participants with the CAM. Group inter-rater reliability assessment will then be tested every 6 months using previously described methods.41

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Participant flow through the trial is illustrated. Baseline clinical and cognitive assessments will take place during preoperative enrolment. Whole-scalp, wireless, high-density (72-channel) electroencephalography (EEG) recordings will take place immediately prior to surgery, intraoperatively and during the early stages of postanaesthesia care unit (PACU) recovery. A low-resolution four-channel system will then be used for recordings during drug infusion on the subsequent mornings. The study drug will be given during surgical closure and again during the first two postoperative mornings. A follow-up survey will then be conducted 30 days after discharge. AD8, Eight-item Informant Interview to Differentiate Ageing and Dementia; CAM, Confusion Assessment Method; IADL, Independent Activities of Daily Living; MoCA, Montreal Cognitive Assessment; PONV, Postoperative Nausea and Vomiting; PROMIS, Patient-Reported Outcomes Measurement Information System; QoR, Quality of Recovery; RASS, Richmond Agitation Sedation Scale; VAS, Visual Analogue Scale; AM, Ante Meridiem; PM, Post Meridiem.

Data management

Data collection will be the responsibility of the clinical trial staff under the supervision of the study PI (PEV). Research data will initially be reported on paper case report forms during patient interactions, and these deidentified data will be uploaded to the Research Electronic Data Capture (REDCap) application. This is an electronic database that resides on a secured, password-protected network managed by the Michigan Institute for Clinical and Health Research (National Institutes of Health-funded Clinical and Translational Science Awards institute of the University of Michigan). Quality control procedures will also be implemented beginning with the data entry system, and data quality checks will be generated. The Michigan Institute for Clinical and Health Research also performs routine, scheduled maintenance and quality control checks on the REDCap system. The REDCap system also incorporates logic that requires appropriate responses and missing/incorrect data are readily and transparently highlighted. We will also request independent audits by the Michigan Institute for Clinical & Health Research during trial operations.

Statistical analysis

Descriptive statistics will initially be calculated, with categorical outcomes presented as proportions (frequencies) and continuous data presented as means (±SD) or medians (IQR). Normality of distribution will be assessed using the Shapiro-Wilk test, and parametric or non-parametric tests will be applied as appropriate.

As mentioned previously, postoperative delirium will serve as the primary endpoint as assessed through the first three postoperative days as previously described. The primary analytical test will be a multivariable logistic regression model that will follow the generalised estimating equations (GEEs) approach. Independent variables will include the placebo group (reference), 1.5 mg/kg caffeine group, 3 mg/kg caffeine group, baseline caffeine intake and a priori variables that may independently predict delirium (eg, age, male sex, American Society of Anesthesiologists Physical Status score, baseline cognitive function via Montreal Cognitive Assessment, baseline functional status via Instrumental Activities of Daily Living, depression and epidural use46). Baseline cohort imbalances, defined by absolute standardised differences >0.20, will also be included in this model. Missing delirium data will be assessed for randomness using the Little’s Missing Completely at Random Test. If data are missing at random, imputation will be performed, and these results will be compared with results with the actual data in a supplementary appendix. If data are not missing at random, no imputation will be performed, and the Cochrane-Armitage test will be used to assess for the proportion of missing data across each arm.

For secondary endpoints, a similar GEE-based approach will be used. Within-group and between-group comparisons will be analysed based on coefficients in the model. This approach also allows for flexibility with missing data. Unstandardised beta coefficients will be presented with 95% CIs and p values (<0.05 will be considered statistically significant). All GEE models will use the empirical parameter estimates with an exchangeable correlation matrix. As indicated previously, models will adjust for baseline cohort imbalances. Additionally, for Quality of Recovery and headache severity, the differential effect of habitual caffeine consumption and group will be tested. This will be tested via interaction term of habitual caffeine consumption with caffeine group (placebo, 1.5 mg/kg caffeine, 3 mg/kg caffeine). The daily number of caffeinated beverages will be recorded for each participant, and the resulting distribution will be analysed and used to inform the most appropriate categorisation of habitual caffeine users (eg, non-users, low use, high use).

Prespecified substudy and subgroup analyses

A substudy analysis will be conducted to identify cortical dynamics associated with postoperative delirium. Based on our preliminary data, we hypothesise that delirium will reflect deviations in neural criticality, which is a postulated state of a system that is poised at the boundary of a phase transition.17 48 Proximity to neural state transitions may allow dynamic, flexible shifts in neural processes for supporting cognitive function. In our preliminary single-centre trial (CAPACHINOS-1), surrogate EEG measures of neural criticality were reduced with PACU delirium and increased with caffeine.17 As such we also hypothesise that caffeine will restore critical dynamics (ie, proximity to criticality) concurrent with reduced delirium risk. To test these hypotheses, a wireless, whole-scalp, high-density (72-channel) system (Mobile-72, CGX, San Diego, California, USA) will be used in the immediate perioperative setting (and intraoperatively). This whole-scalp, high-density system will also enable analysis of functional connectivity patterns and neuroanatomical source analyses in relation to delirium and cognitive function.

A trial subgroup analysis will also be conducted in patients meeting criteria for mild cognitive impairment at preoperative baseline (anticipated n=50)49 based on preoperative AD8 screening, Montreal Cognitive Assessment score <23,43 and preserved functional independence via cognitive independent activities in daily living. This subgroup analysis will test the hypothesis that preoperative mild cognitive impairment will be associated with deviations in baseline neural criticality, based on surrogate EEG measures.17 50 As a secondary analysis, we will also determine whether caffeine exerts a differential effect on cognitive outcomes in patients with mild cognitive impairment, based on our evidence that caffeine may optimise neural criticality, and criticality breakdowns are postulated to underlie mild cognitive impairment and Alzheimer’s disease.17 50 51 Delirium incidence, severity and Montreal Cognitive Assessment scores 1-month postdischarge will also be compared in those with and without baseline mild cognitive impairment.

Data and safety monitoring

All trial participants will be monitored by both the research staff—including direct PI oversight—and clinical teams based on standard hospital care and protocols. The research team will monitor patients daily for adverse events, which will be reported based on IRB and NIA guidelines. A licensed research nurse will also monitor participants during caffeine infusions (see full trial protocol, online supplemental appendix 1, for complete details). Additionally, a DSMB has been empanelled to act in an advisory capacity to the NIA and periodically evaluate the progress and safety of the study. Members of the DSMB have appropriate (and complementary) expertise that is suited for the trial, and each member has confirmed no conflict of interest via signed statement submitted to the NIA. The DSMB will thus be able to make independent, impartial recommendations to the NIA throughout the trial lifespan. Complete DSMB details are available in the full trial protocol (online supplemental appendix 1).

Patient and public involvement

None.

Trial strengths and limitations

Notable strengths of this study merit consideration. This is a placebo-controlled trial, and the stratified, block-randomised design will alleviate trial arm imbalance and mitigate selection bias. Multiple doses will be tested, which will help to determine the optimal dosing threshold. To enhance trial rigour, we will follow a quadruple-blinded design; team statisticians will perform analyses in a blinded manner, and the blind will only be lifted after analysis of the primary and secondary outcomes is complete. The DSMB and independent auditing services will also provide additional monitoring support. The advanced EEG analysis will also help to identify cortical dynamics underlying both delirium and caffeine administration, which will advance neurobiological understanding of delirium and cognitive function.

Important limitations also warrant consideration. Trial generalisability will be limited, given the focus on older patients and single-centre design of the study. Additionally, although different doses will be tested, it is possible that a more sophisticated, personalised or dynamic caffeine dosing strategy is required to optimise postoperative neurocognitive and clinical recovery. For example, relatively higher doses may be required for cognitive benefit in the early postoperative setting with residual anaesthesia or in those with high levels of habitual use. A future adaptive trial may also be warranted, particularly when a more comprehensive understanding of perioperative benefits and risks of caffeine are known, as well as the optimal timing of caffeine administration. While this trial will not provide definitive answers to these questions, the study will serve as an initial step because it tests different caffeine doses along multiple time points of postoperative recovery in older adults.

Ethics and dissemination

Derived from landmark declarations, codes and guidelines, Emanuel et al propose seven requirements for systematically analysing the ethical framework of a proposed clinical research study.52 These requirements are the following: social or scientific value, scientific validity, fair participant selection, favourable benefit: risk ratio, independent review, informed consent, and respect for potential and enrolled participants. This framework is applied to the current trial, with considerations outlined in box 1. As mentioned previously, the trial was also approved by the University of Michigan Medical School Institutional Review Board (HUM00218290), and written informed consent will be obtained from all participants. An expanded discussion of risks and befits is also included in the full clinical trial protocol (online supplemental appendix 1).

This trial will be presented at academic medical conferences, and trial operations and results will be disseminated via social and news media. As noted, the trial has been registered at www.ClinicalTrials.gov (NCT05574400), and updates will be made publicly available on this website. On trial completion, results will be published in medical and scientific journals.