We thank Mr. Kubo and Dr. Hosokawa for their interest in our study.  Perioperative neurocognitive disorders are a major public health issue, and advancing our neurobiologic understanding of these impairments represents an important step toward prevention and management. Our team welcomes the opportunity to discuss the neurophysiologic results of our study as they relate to the perioperative outcomes tested.

First, we agree that depth of anesthesia is an important factor to consider with these analyses. Our study design did account for depth and temporal course of anesthesia, as described in our supplemental methods.  Neurophysiologic data from the anesthetic maintenance phase were derived only from the time period between skin incision and first minimum alveolar concentration of 0.7 toward the end of surgery, and depth was based on age-adjusted minimum alveolar concentration. Data from other time epochs studied were anchored to the timestamped clinical event (e.g., anesthetic induction, extubation, and others) using previously described methods. It is thus unsurprising, for example, that frontal alpha power was reduced during the time window immediately before extubation compared to the plane of surgical anesthesia. Intraoperative frontal–parietal theta connectivity was only derived from the maintenance phase described, and we have previously demonstrated that these alpha- and theta-based dynamic cortical connectivity patterns persist even after accounting for anesthetic depth and agent.  We also point out that none of the identified cortical connectivity patterns were significantly associated with National Institutes of Health Toolbox scores, and we consider the theta connectivity association with delirium hypothesis-generating, given that delirium was a secondary outcome (and assessed via abbreviated instrument). 

Mr. Kubo and Dr. Hosokawa also raise important points regarding the Toolbox. We chose to use cognitive tests from the Toolbox for key reasons. First, the Toolbox offers demographically corrected normative standards which means that adjusted scores are available that account for key demographic factors and educational attainment. This allowed for statistical adjustment of additional relevant variables in the early postoperative setting as mentioned by Mr. Kubo and Dr. Hosokawa (e.g., opioid administration, pain, and others). Second, we believe the first 48 h after surgery does, in fact, represent an appropriate time to perform such cognitive analysis given that this timeframe coincides with high delirium risk.  Performing the National Institutes of Health Toolbox analyses revealed distinct cognitive deficits—working memory and processing speed—that arose in patients experiencing delirium. While we consider these results hypothesis-generating, this helps to advance our understanding beyond the arbitrary, binary distinction of delirium. These findings would not have been possible without such dedicated cognitive function testing within this timeframe. Attrition was indeed present with Toolbox testing (reasons outlined in the supplemental appendix), but perioperative Toolbox data were still available from 58 of 59 (98%) of enrolled patients, and dedicated postoperative results were available from 48 of 59 (81%) participants, encapsulating 130 observations for the generalized estimating equation approach outlined. As we discussed, the attrition present was unlikely to affect the longitudinal modeling estimates.  Finally, the limited sample size precludes dedicated subgroup analyses, but this could certainly be considered with future, larger-scale studies.

We agree that nociceptive transmission may manifest via distinct functional connectivity patterns, as assessed with techniques such as weighted phase lag index. The interplay between nociceptive processing and cortical functional connectivity patterns appears evident and this is certainly an important neuroscientific and clinical topic worth pursuing. While this type of complex analysis was not the focus of our current study, we agree that these perioperative data sets could be leveraged for generating preliminary data and hypotheses.

Finally, Mr. Kubo and Dr. Hosokawa are correct that baseline electroencephalographic data were collected in the clinical setting (preoperative holding unit). This was a prespecified design choice, as our primary hypothesis focused on connectivity patterns in the immediate perioperative environment. Preliminary data for supporting the hypothesis were derived from human volunteers in the same hospital setting. Nonetheless, a future study focused on a nonclinical, relatively artifact-free environment may be useful for identifying neurophysiologic signatures. However, one caveat to consider with this approach is that any fragile neurophysiologic findings obfuscated by artifact may not translate to the clinical environment.

Once again, I thank Mr. Kubo and Dr. Hosokawa for their interest in our study. We agree that there are additional hypotheses that merit testing in future studies, and we are grateful for their insights into these very important topics related to neurophysiologic monitoring, perioperative cognition, and pain.