Seen and Ignored: Are We Undermining Studies of Brain Health Interventions Before We Start?

See Article, p e52 Intraoperative electroencephalography (EEG) provides a direct measurement of the effects of anesthetics and sedatives on the brain, and the development of processed EEG monitors has promised the practicing anesthesiologist a more simplified interpretation in the perioperative setting, where attentional resources are divided among competing needs. Particularly in light of the ongoing concerns for anesthetic contributions to perioperative neurocognitive disorders (PNDs), EEG measures are increasingly posited to provide value, not just as a tool to monitor the risk of anesthetic awareness with recall, but as a warning signal about patients with brains vulnerable to PND. Several studies have examined whether anesthetic titration in response to processed EEG monitoring could improve patient outcomes.1 While there are clearly challenges in the use of current processed EEG monitors, the recommendations for best practices for postoperative brain health2 suggest that “there is strong support for the general principle of EEG-based anesthetic titration to reduce PND rates in older adults.” In this edition, Gross et al3 offer us a glimpse into the real-world behavior of clinicians currently incorporating processed EEG in their anesthetic management. Specific intraoperative patterns of EEG, most significantly burst suppression, are associated with poor outcomes.4,5 While burst suppression does not seem to predict poor cognitive recovery in healthy volunteers,6 in surgical populations, using EEG monitors to titrate anesthetic delivery to avoid specific EEG patterns might be beneficial.1 The recent Electroencephalography Guidance of Anesthesia to Alleviate Geriatric Syndromes (ENGAGES) trial did not support the use of EEG-guided anesthetic administration for the prevention of postoperative delirium, but did find a lower 30-day mortality on patients with EEG-guided anesthetic.7 Despite multiple studies that have clearly demonstrated the potential benefit of using EEG during anesthesia care, the failure of intraoperative EEG use to reduce the incidence of postoperative delirium in ENGAGES risks the elision of previous studies altogether rather than prompting a more rigorous and nuanced debate over the practical implications of each. Thankfully, some discussion of the nuances of processed EEG measures, aging, and the vulnerable brain is developing in the literature.8,9 And yet there may be important behavioral issues beyond the limitations of the monitors themselves; for example, a recent retrospective analysis indicated that the use of bispectral index (BIS) was associated with worse outcomes, without ever articulating how the BIS values were managed.10 As with all monitors, clinical judgment is crucial to interpreting EEG physiologic data within the context of an anesthetic; any changes in outcomes from using a monitor will depend on clinician response to monitors. For there to be a change in outcome, the clinicians must incorporate information from all available monitors to optimize care for the individual patient. Patient assessment should include evaluation and correlation of EEG data with hemodynamic and other monitoring data, within the greater clinical context. That is, EEG information is additive to the patient assessment, providing clinicians with valuable signals for their practice, observation, and care. Gross et al3 performed a retrospective analysis of 138 patients undergoing a target-controlled infusion of propofol and remifentanil for robotic surgery at the Foch University Hospital in Suresnes, France, between July and December of 2017, and looked into the frequency of BIS values outside the manufacturer-recommended range during the intraoperative period. They found that, despite using the BIS monitors to guide the anesthetic, over 35% of patients had values under 40, indicating excessive anesthetic delivery. Also, burst suppression was still present in 17.8% of patients. As expected, this was more frequent in elderly patients, but also interestingly in female patients and body mass index (BMI) <26. These associations were not due to a difference in surgical procedure or detectable differences in intraoperative temperature management. Such data suggest that the clinicians in the study did not or could not adjust their anesthetic delivery to avoid the potential overdose to a subset of their patients. It is also worth noting that these deviations were much more frequent than deviations above the recommended BIS range, which would be suggestive of underdosing. The authors speculate that such behavior could be explained by several factors, including a lack of concern for the implications of low BIS values, or a lack of confidence in the BIS in a setting with otherwise normal monitoring parameters (eg, not wanting to reduce the target concentration below a certain amount out of fear of possible anesthetic awareness). Finally, it should be noted that no hemodynamic or other monitoring data were incorporated into the analysis, so it is possible that some of the instances of low BIS corresponded to periods where the BIS monitor was incongruent with the overall clinical assessment. This brief report has multiple limitations, by its very nature. It contains data from <140 anesthetics provided at a single center and hence might have limited external generalizability. The study attempts to conduct multiple comparisons to look for associations in models that are ad hoc, with minimal justification for the particular associations that they chose to investigate, versus others that could also contribute (eg, undertreated hypothyroidism). Some of this may reflect the author’s decision to present the data as a brief report rather than allowing a more extensive discussion, and the desire to contribute to hypothesis generation for future studies rather than making a more definitive statement to the field. It is interesting to consider how this glimpse into clinician behavior in response to processed EEG data should prompt future investigations into EEG-guided anesthesia. As another data point, investigators at the Cleveland Clinic found less response to automated warnings to support a borderline low blood pressure based on a processed EEG response indicative of excessive anesthetic delivery than they expected,11 suggesting that clinicians are likely to downplay the need for interventions in response to processed EEG measures that are out of range. This corresponds with our own anecdotal experience within our institution as well. We thank Gross et al3 for documenting their own experience to allow us to see how human factors impact the ongoing work by the Brain Health Initiative. The implications of an electrically silent brain during surgery are not yet fully understood. It may still prove that, for most patients, this will have no implications, but until this is fully explored, it is important to consider EEG as a physiologic variable in vulnerable patients. And, for us to conduct studies that offer meaningful answers to this question, we must be looking beyond just providing the processed EEG to the clinician delivering the anesthetic, but also considering what they are choosing to do with that information once it is available. E DISCLOSURES Name: Susana Vacas, MD, PhD. Contribution: This author helped conceive and write the article. Name: Andrew E. Hudson, MD, PhD. Contribution: This author helped conceive and write the article. This manuscript was handled by: Maxime Cannesson, MD, PhD.