Delirium is an acute disorder of cognition and attention that has been likened to “acute brain failure”.1 It is recognized as the most common surgical complication in older adults,2,3 affecting up to 50% of older patients after an operation.1,4 Given its association with increased hospital length of stay, readmission, cognitive and functional decline, and mortality,5,6,7,8 delirium has significant consequences for both patients and communities. In the US alone, delirium in patients over the age of 65 is estimated to cost more than $164 billion per year.9 Delirium may be preventable in 30-40% of cases,10,11 representing a target for intervention of substantial public health and political relevance.

Postoperative cardiac surgery patients are among those at greatest risk of delirium.1 Despite this, evidence regarding the prevention of delirium in this setting remains sparse. Among the modifiable precipitants of delirium identified, benzodiazepine use is a recurring association. Studies have found that both pre-12 and postoperative benzodiazepine2,13,14,15,16 use is associated with an increased risk of delirium, though the only randomized evidence comes from comparisons of benzodiazepine vs sedation with the alpha-2 agonist dexmedetomidine in the post-cardiac surgery intensive care unit.14,16,17

Minimizing perioperative benzodiazepine use is an established approach to postoperative delirium prevention. This strategy is recommended by the American Geriatric Society3 and American College of Critical Care Medicine.18 Nonetheless, because of their favourable hemodynamic profile and amnestic properties, intraoperative benzodiazepines are routinely used in cardiac surgery. No studies have evaluated the effects of benzodiazepine-free anesthetic regimes on delirium after cardiac surgery. In addition, little is known about current practice among cardiac anesthesiologists in Canada, and there are no evidence-based practice guidelines that address the issue of intraoperative benzodiazepine administration. Anecdotally, anesthesiologists express extreme and dichotomous beliefs about whether benzodiazepines should be given to cardiac surgery patients. As such, we sought to describe current practice by undertaking a survey examining benzodiazepine administration during cardiac surgery. We hypothesized that among Canadian cardiac anesthesiologists there are substantial practice pattern variations in intraoperative benzodiazepine usage, supporting the need for a large clinical trial to establish the optimal approach to cardiac anesthesia care.

Methods

This study was designed and is reported according to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.19 Before commencing data collection, we obtained local Research Ethics Board approval on 8 February 2017, which waived the need for individual, informed consent. We developed a 16-item questionnaire (E-supplement 1) that assessed five domains: (1) practitioner characteristics, (2) the proportion of cardiac surgery patients receiving benzodiazepines, (3) the specific agents and corresponding doses that were typically given, (4) the determinants of benzodiazepine administration, and (5) individual beliefs about the role of benzodiazepines in preventing intraoperative awareness. We generated questionnaire items using an iterative process that included review of pertinent literature, informal dialogue with practitioners, and a focus group of cardiac anesthesiologists practicing at Hamilton General Hospital, which is affiliated with McMaster University. A group of cardiac anesthesia staff piloted the survey to ensure clarity; we sought to develop a questionnaire that balanced the amount of information collected with ease of completion.

The target population of the survey were licensed anesthesiologists who manage adult patients undergoing cardiac surgery in university-affiliated hospitals in Canada. We chose to exclude all trainees (fellows and residents). We also excluded anesthesiologists practicing in community hospitals because 1) they represent a minority of practitioners in Canada and 2) the lack of relationship with an academic institution made it difficult to confidently identify which community hospitals were providing cardiac surgery, which would have resulted in an uncertain denominator for our sampling frame. We used a stepwise process to develop our sampling frame. First, we contacted representatives from the departments of anesthesia of the thirteen universities in Canada with a medical school and obtained a contact person for each affiliated hospital providing cardiac surgery. Second, we contacted the individual identified to 1) ascertain the number of cardiac anesthesiologists at that site, 2) establish an agreement for their site to participate in the survey, and 3) develop a plan for survey distribution to eligible faculty. Identified cardiac anesthesiologists (n = 346) formed our sampling cohort.

We distributed a link to the survey to each departmental contact, who forwarded it to eligible faculty by email. The email also included information about the purpose of the study, a statement about the voluntary nature of participation, and information about participant anonymity. Study data were collected and managed using REDCap electronic data capture tools hosted at McMaster University.20 REDCap (Research Electronic Data Capture) is a secure, web-based application designed to support research data capture, including surveys. Recruitment and study materials were distributed in English and French. The survey was initially sent on 10 April 2017. A modified Dillman approach was employed.21 We sent three follow-up emails every 2 weeks and distributed non-binding incentives (coffee cards) to the 346 potential respondents to encourage survey completion. We closed the survey 1 June 2017.

Statistical analysis

Survey responses were analyzed descriptively using counts and percentages. Response rate was calculated by dividing the number of responses by the total number of cardiac anesthesiologists identified by site contacts (n = 346). Continuous variables were described using mean (standard deviation [SD]) or median [interquartile range (IQR)] depending on distributions. Statistical analyses were conducted using Stata version 12 (StataCorpLP, College Station, TX, USA).

Results

Of 346 cardiac anesthesiologists practicing at academic centres, 70% (243/346) responded to the survey. Table 1 reports respondent characteristics. The largest proportion (32%) of respondents had been practicing anesthesia for more than 20 years. While most (61%) had completed one year of fellowship, 25% (61/243) had no additional training in the subspecialty of cardiac anesthesia. The most common response, endorsed by 41% (99/243) of respondents, to the proportion of practice devoted to cardiac anesthesia was 21-40%.

Table 1 Respondent characteristics
Full size table

Table 2 reports the survey responses to questions about benzodiazepine administration to adult patients undergoing cardiac surgery. Eleven percent (26/243) of respondents do not administer benzodiazepines during cardiac surgery. Of those who reported using benzodiazepine(s), most (214/243; 88%) anesthesiologists reported using midazolam, with a mean (SD) dose of 4.9 (3.8) mg given to an average-risk, 65-yr-old patient undergoing coronary artery bypass graft on cardiopulmonary bypass (CPB). Other benzodiazepines reportedly used included lorazepam, diazepam, and oxazepam. When asked in what proportion of cardiac surgical patients they gave benzodiazepines, the response was bimodal (see Figure), with 23% (54/243) either never giving them or giving them to less than 20% of patients and 51% (124/243) giving them to 80-100% of patients. When asked in what percentage of patients they believed it would be possible to provide cardiac anesthesia without using any benzodiazepine medication, 71% (173/243) of respondents believed this would be possible in > 80-100% of patients.

Table 2 Responses to survey items
Full size table
Figure
figure 1

Proportion of cardiac surgery patients to whom intraoperative benzodiazepines are administered as reported by 243 survey respondents

Full size image

The most common considerations that influenced the benzodiazepine use or dose were patient age, anxious/agitated patients, history of alcohol/drug/benzodiazepine use, and risk factors for intraoperative awareness. When asked which factors mandated the use of benzodiazepines during cardiac anesthesia, the majority (132/243; 54%) responded that benzodiazepines were never mandated during cardiac anesthesia. Of those who believed that benzodiazepines were sometimes mandated during cardiac anesthesia, the factors identified, in decreasing order of frequency, included: history of awareness, history of alcohol or drug use, young age, preoperative benzodiazepine use, hemodynamic instability, and emergency surgery.

For respondents using benzodiazepines, the most common point during a cardiac surgical procedure when benzodiazepines were administered was induction of anesthesia, followed by the start of CPB, when coming off CPB, while on CPB as an infusion, and at the end of the case or post-CPB.

Respondents estimated the proportion of their patients experiencing intraoperative awareness to be a median [IQR] estimate of 1 [0-5]%. Nine (4%) respondents identified having had complaints or lawsuits from cardiac surgery patients related to their experience of intraoperative awareness, 95% did not, and 1% did not want to answer this question. When asked whether they believed that benzodiazepines decrease the risk of intraoperative awareness in cardiac surgery patients, 16% of respondents were in complete agreement, 37% agreed somewhat, 17% were neutral, 12% somewhat disagreed, 17% completely disagreed, and 0.4% did not respond.

Discussion

We surveyed Canadian academic cardiac anesthesiologists about their intraoperative use of benzodiazepines. Our results confirmed a dichotomous practice with respect to intraoperative benzodiazepine administration (see Figure); 11% (26/243) of respondents did not use benzodiazepines within their cardiac anesthesia practice. Most respondents believed that benzodiazepines are never mandated during cardiac surgery and that nearly all cardiac surgery patients could be provided an anesthetic without benzodiazepines.

The most common reasons—with the exception of patient anxiety or agitation—for using intraoperative benzodiazepines during cardiac surgery were directly or indirectly related to patient risk of intraoperative awareness or postoperative delirium. Indirect factors included patient age (which may be related to concerns about awareness in younger patients and delirium in older patients) and patient history of alcohol, drug, or benzodiazepine use (a risk factor for intraoperative awareness).22 Nonetheless, non-benzodiazepine options—including opioids, anticholinergics, and alpha-2-agonists—exist for anxiolysis. Despite a recommendation from the American Society of Anesthesiologists that patients undergoing cardiac surgery should be given either a benzodiazepine or scopolamine to reduce the risk of intraoperative awareness,23 limited evidence supports the effectiveness of benzodiazepines in preventing awareness.

A systematic review examining the effectiveness of various anesthetic interventions to prevent intraoperative awareness found very low-quality evidence that benzodiazepines reduce awareness compared with thiopental, ketamine, and placebo.24 When examining the four included studies, the evidence of effect becomes questionable. The earliest study compared the effect of 2 mg·kg−1 of ketamine with 30 mg diazepam and supplemental nitrous oxide on the incidence of awareness in 26 parturients undergoing forceps delivery: 4/13 patients experienced awareness with ketamine while none did with diazepam.25 This result may relate to the benzodiazepine, the use of nitrous oxide in the diazepam group, or the questionable equivalency of the two administered regimes. Haram et al. compared the incidence of ‘unpleasant recall’ after induction of general anesthesia using either thiopental or diazepam in 82 parturients.26 In the thiopental group, 5/40 patients had recall, while none did in the diazepam group. Nonetheless, the two regimes are unlikely to be equivalent; light anesthesia may explain the observed effect. Miller et al. evaluated the influence of adding graded doses of midazolam or placebo to a standard anesthetic induction on the incidence of intraoperative recall in 90 patients undergoing arthroscopic knee surgery.27 This trial was stopped early after 4/21 patients in the placebo group and 2/69 patients in the midazolam group experienced intraoperative awareness. Chance or lighter anesthesia may explain this finding.

Studies have found an association between delirium and both preoperative12 and postoperative benzodiazepine2,13,14,15,16 use. In a nested case-control study, the odds of delirium were 3.0 (95% confidence interval [CI], 1.3 to 6.8) higher in non-cardiac surgery patients who were administered benzodiazepines post-operatively.2 In prospective observational data, for every mg of midazolam administered after cardiac surgery, patients were 7-8% more likely to develop delirium, suggesting a dose-response effect.15 The development of delirium is hypothesized to relate to imbalances in the release, synthesis, and degradation of gamma-aminobutyric acid (GABA) and other neurotransmitters.28 Because of their affinity for GABAergic receptors in the brain regions responsible for consciousness, benzodiazepine-induced decreases in arousal may cause unpredictable neurotransmission and delirium.28

Randomized controlled trials have compared benzodiazepines with dexmedetomidine for sedation in the post-cardiac surgery intensive care unit.14,16,17 In a mixed medical and surgical ICU population, sedation with dexmedetomidine resulted in more days alive without delirium (median days 7 vs 3; P = 0.01) than lorazepam.14 In another trial, 54% of patients treated with dexmedetomidine developed delirium compared with 77% of patients treated with midazolam (absolute risk increase = 23%; 95% CI, 14% to 33%; P = 0.001).16 This evidence has led to recommendations to minimize benzodiazepine use in elderly and ICU patients.3,18 Benzodiazepine administration in the ICUs in Canada is no longer considered consistent with standard of care.18,29 Nonetheless, our survey suggests that intraoperative benzodiazepine use remains common in cardiac surgery, possibly because of the absence of randomized evidence showing harm.

Current heterogeneity in practice reflects a gap in the literature and the concerns about the relationship of benzodiazepines with postoperative delirium and beliefs about their role in preventing intraoperative awareness. Our respondents endorsed giving benzodiazepines to approximately half of patients within their cardiac anesthesia practices, but believed that it was possible to give nearly all patients a cardiac anesthetic without benzodiazepines. The use of both minimized and liberal approaches to benzodiazepines by anesthesiologists during cardiac surgery confirms the existence of equipoise. This heterogeneous practice emphasizes the need for a trial to establish best practice.

Our study has several limitations. First, we surveyed Canadian academic cardiac anesthesiologists; their practice may not reflect that of all practitioners. Nonetheless, this group represents the majority of practitioners and we felt that this sampling frame would allow us to evaluate whether the dichotomous beliefs about benzodiazepine use we had anecdotally observed were reflected in practice. Second, we did not ask about what medications our respondents used as alternatives to benzodiazepines. Finally, though our response rate was excellent, 30% (103/346) of academic cardiac anesthesiologists in Canada did not participate; their practice may differ.

Despite these limitations, our survey provides important information about Canadian cardiac anesthesia practice. The substantial variation in the usage of benzodiazepine during cardiac surgery highlights the existing community equipoise and suggests competing concerns about the role of benzodiazepines in preventing intraoperative awareness and their relationship with postoperative delirium. Given the lack of data supporting either approach, further study is needed to address this evidence gap. To address this gap, we are currently conducting a two-site pilot study (clinicaltrials.gov identifier NCT03053869), whose objective is to show the feasibility of conducting a multi-centre, randomized cluster crossover trial evaluating the impact of benzodiazepine-free cardiac anesthesia on the incidence of postoperative delirium among adult cardiac surgery patients.