Postoperative Delirium in a Randomized Trial of Intraoperative Norepinephrine versus Phenylephrine Infusion

Delirium is a common adverse postoperative event, occurring in 5 to 15% of adults after major surgery with general anesthesia.1 There is strong mechanistic reasoning that hypotension during general anesthesia may lead to cerebral hypoperfusion and thus a relative cerebral oxygen deficit, provoking delirium. Two commonly used intraoperative vasopressors, phenylephrine and norepinephrine, increase blood pressure but may differ in their effect on cardiac output.2 Norepinephrine is associated with increased cerebral oxygen saturation compared with phenylephrine when used intraoperatively,3 suggesting potential benefits to cerebral perfusion that may translate into lower rates of postoperative delirium. We retrospectively analyzed clinical delirium assessments by vasopressor assignment for patients enrolled in VEGA-1, an open-label multicenter cluster-randomized pragmatic trial comparing norepinephrine and phenylephrine by monthly randomization, in three participating University of California-San Francisco hospitals.VEGA-1 included patients older than 18 yr undergoing major noncardiac surgery of 2 h or more under general anesthesia and requiring vasopressor infusion to maintain mean arterial pressure according to clinical judgment (typically mean arterial pressure within 20% of baseline, but targets could be adjusted on an individual basis).4 The University of California-San Francisco Institutional Review Board (No. 19-29799) waived the requirement for informed consent, and the parent study was registered on clinicaltrials.gov (NCT04789330). This secondary analysis adheres to the applicable Consolidated Standards of Reporting Trials guidelines. Recruitment at the three hospitals in this delirium substudy occurred between July 1, 2021, and December 31, 2021.The dataset for the VEGA-1 trial is described elsewhere.4 The primary outcome for this delirium study was a positive delirium screen within the first 7 days of hospitalization: either a Nursing Delirium Screening Scale (Nu-DESC) score of 2 or greater5 or a positive Confusion Assessment Method for the Intensive Care Unit (CAM-ICU),6 performed according to routine clinical protocols (expected at least once per 12-h nursing shift). Additional variables were extracted from the electronic medical record for this analysis, including the use of preinduction midazolam, case urgency (American Society of Anesthesiologists [Schaumburg, Illinois] Physical Status "E" modifier, or booked as an add-on case), and preoperative delirium risk prediction calculated using the Age, World-backward, Orientation, iLlness severity, Surgical risk delirium risk prediction tool (AWOL-S).7 Missingness in preoperative AWOL-S values was addressed using multiple imputation (predictive mean matching, n = 25 imputations) based on age, sex, American Society of Anesthesiologists Physical Status, case urgency, vasopressor assignment, case room, and surgical service.We performed intention-to-treat analyses using mixed-effects logistic regression, treating hospital facility as a random effect and patient-level covariates (if any) as fixed effects. The primary analysis adjusted solely for hospital facility as a random effect. Prespecified secondary analyses included partially adjusted (by AWOL-S, case urgency, and patient age) and fully adjusted (by those variables as well as sex, preoperative medications, comorbid medical diseases, and selected intraoperative variables); results of the fully adjusted models are omitted for clarity because they did not meaningfully differ from partially adjusted findings. Unconditional relative risks of delirium were produced from the regressions using g-computation plug-in, with 95% CI produced using SEs estimated by nonparametric bootstrap. We assessed treatment effect heterogeneity through subgroup analyses across levels of prespecified, clinically motivated covariates. Analyses were performed using Stata.MP 17 (StataCorp LLC, USA) and R version 4.3.1 (https://cran.r-project.org/src/base/R-4/R-4.3.1.tar.gz; accessed June 16, 2023). All test statistics were two-sided; Benjamini–Hochberg P value significance thresholds were used to control the false discovery rate to 5% by comparing calculated P values against a progressively more stringent threshold addressing the 13 semi-independent significance tests.8Adherence to randomized vasopressor was greater than 94% at the three participating hospitals. Of 2,510 patients eligible for VEGA-1 whose surgery occurred in a participating hospital, 22 (0.9%) were excluded from this analysis, mostly due to lack of delirium screens (Supplemental Digital Content, https://links.lww.com/ALN/D695). Five hundred eighty-two (23.4%) patients were missing an AWOL-S value, which was imputed (Supplemental Digital Content, https://links.lww.com/ALN/D695). The proportions with postoperative delirium in the first 7 days after surgery were 12.2% and 13.6% in the norepinephrine and phenylephrine groups, respectively (table 1). Primary analysis adjusting for center found that rates of delirium were not different between the norepinephrine and phenylephrine groups (relative risk, 0.88 [95% CI, 0.72 to 1.08]; P = 0.23, which was above the corresponding Benjamini–Hochberg statistical significance threshold of P = 0.019), a finding that remained stable in partially adjusted (relative risk, 0.92 [95% CI, 0.76 to 1.12]) analysis after AWOL-S imputation (fig. 1).Prespecified subgroup analysis (fig. 1) yielded substantially similar conclusions. No subgroup P value was below its corresponding multiple testing–adjusted significance threshold. Minimally adjusted risk ratios for delirium in norepinephrine-treated patients age 75 yr or older and those at higher delirium risk were 0.68 (95% CI, 0.48 to 0.97; P = 0.031, above the threshold P = 0.004) and 0.78 (95% CI, 0.62 to 0.98; P = 0.032, above the threshold P = 0.008), respectively; risk ratios after multivariable adjustment were similarly not statistically significant. Full results of main and subgroup models with further multivariable adjustment are available in the Supplemental Digital Content (https://links.lww.com/ALN/D695).In conclusion, in a post hoc analysis of delirium screening data after a cluster-randomized trial of norepinephrine versus phenylephrine, no statistically significant difference in the rate of delirium was detected between the two groups. While the trial's predetermined size limited the power to detect non-zero treatment effects on this post hoc outcome, the statistically nonsignificant relative reduction in risk in the age 75 yr or older and high delirium risk subgroups, with a risk ratio of approximately 0.7 to 0.8, is consistent with other putative delirium risk interventions (e.g., avoiding excessively deep hypnosis,9 use of dexmedetomidine10). Suggestively, a recent multicenter retrospective observational study of intraoperative phenylephrine versus ephedrine, an indirect-acting sympathomimetic that promotes endogenous norepinephrine release, found 35% increased odds of delirium with phenylephrine.11 As both norepinephrine and phenylephrine are considered standards of care,2 the impact of vasopressor selection on delirium merits further study.There are important limitations to this study. This trial was performed open-label; while the delirium screening that occurred as part of clinical protocols on the hospital floors is independent of trial randomization group, we cannot definitively rule out bias in delirium ascertainment leading to these results. Pragmatic delirium screening is notorious for undercapture, which would be expected to have occurred in a balanced fashion between the two randomization groups. The statistically nonsignificant delirium risk reduction seen with norepinephrine may motivate a pragmatic, definitive follow-up trial, which, based on adherence rates seen in VEGA-1, should be feasible.Dr. Bokoch received funding support from an International Anesthesia Research Society Mentored Research Award. This analysis was supported by the University of California-San Francisco Department of Anesthesia & Perioperative Care. The VEGA-1 trial was funded by the International Anesthesia Research Society (Initiative for Multicenter Pragmatic Anesthesiology Clinical Trials (IMPACT) Award grant).Dr. Legrand has received consulting fees from Sphingotec (Hennigsdorf, Germany), La Jolla Pharmaceuticals (Waltham, Massachusetts), and Alexion (New Haven, Connecticut). Dr. Pirracchio has received consulting fees from Philips (Amsterdam, The Netherlands). The other authors declare no competing interests.Supplemental Figures and Tables, https://links.lww.com/ALN/D695