Intraoperative Use of Phenylephrine versus Ephedrine and Postoperative Delirium: A Multicenter Retrospective Cohort Study

In this multicenter retrospective study, we analyzed hospitalized adult patients who underwent non-cardiac, non-neurosurgical procedures under general anesthesia at two academic healthcare centers in Massachusetts: Beth Israel Lahey Health (Beth Israel Deaconess Medical Center, Boston, Massachusetts) and Massachusetts General Hospital (Boston, Massachusetts) between 2008 and 2020. Deidentified perioperative data from different electronic hospital management systems were used. Supplemental Digital Content 1 (section S1.1, https://links.lww.com/ALN/D310) provides details related to the data collection and sources. Ethical approval for this study (institutional review board numbers 2022P000261 at Beth Israel Deaconess Medical Center and 2022P000897 at Massachusetts General Hospital) was provided by the Committee on Clinical Investigations affiliated with the Beth Israel Lahey Health network and by the Massachusetts General Hospital Institutional Review Board, with waivers of informed consent. A data analysis and statistical plan was established before the data were accessed. This study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Supplemental Digital Content 2, https://links.lww.com/ALN/D311).¹¹ 

Adult hospitalized patients who underwent general anesthesia for noncardiac, non-neurosurgical procedures and received either phenylephrine or ephedrine intraoperatively were eligible for inclusion in this study. Patients who were administered both phenylephrine and ephedrine during surgery were not eligible for inclusion. We excluded patients with an American Society of Anesthesiologists Physical Status classification of greater than IV, a pre-existing diagnosis of delirium, dementia, or mild cognitive impairment based on a billed diagnosis within 1 yr before the index procedure,¹²  or with missing information for potential confounding variables.⁶  The complete-case approach was used for the primary analysis.

The primary exposure was intraoperative administration of phenylephrine, compared to ephedrine. The primary outcome was delirium within 7 days after surgery, defined using the International Classification of Diseases (Ninth/Tenth Revision, Clinical Modification) diagnostic codes.^4,6,13  This time frame for the development of delirium was selected based on clinical relevance and previous literature.^8,14,15  Driven by an inflammatory surge and residual effects of anesthesia,¹⁶  patients are at particularly high risk of postoperative delirium during the first 7 postoperative days.¹⁷  Further details on the definition of the primary outcome and on how delirium was diagnosed at our institution are provided in Supplemental Digital Content 1 (section S1.2, https://links.lww.com/ALN/D310).

The analyses were adjusted for a priori defined confounding variables based on literature review and clinical plausibility. Confounding variables included patient demographics such as age; sex; body mass index; American Society of Anesthesiologists Physical Status classification; and comorbidities such as cerebrovascular disease, chronic heart failure, arterial hypertension, peripheral vascular disease, chronic obstructive pulmonary disease, chronic kidney disease, diabetes mellitus, depression, schizoaffective disorders, alcohol and drug abuse,^18,19  and smoking status. We also included preoperative drug prescriptions of antipsychotics and benzodiazepines, as well as other intraoperatively administered vasoactive medications such as norepinephrine, epinephrine, and dopamine as confounding variables. Additionally, we adjusted our analyses for other surgical and anesthesia-related factors including surgical service, emergency surgery status, duration of surgery, number of packed erythrocyte units administered,²⁰  intraoperative dose of neostigmine, opioids, crystalloid and colloid infusions, nondepolarizing neuromuscular blocking agents, age-adjusted mean alveolar concentration of inhalational anesthetics,²¹  duration of hypotension measured as minutes of mean arterial pressure less than 55 mmHg,^4,22  and surgical complexity (based on work relative value units). We further adjusted for the healthcare network and controlled for the year in which the procedure was performed to address potential trends over time. Further details related to the confounding model are provided in Supplemental Digital Content 1 (section S1.3, https://links.lww.com/ALN/D310).

In the primary analysis, we studied the association of intraoperative administration of phenylephrine compared to ephedrine with postoperative delirium by applying a multivariable logistic regression analysis adjusted for a priori defined confounding variables. Model discrimination and calibration were assessed as described in detail in Supplemental Digital Content 1 (section S1.4, https://links.lww.com/ALN/D310).

The individual anesthesia provider’s preference and practice may play an important role in the selection of vasopressor agents during surgery. In a secondary analysis, we first explored this variability and subsequently addressed its potential impact on the primary association using a multivariable mixed-effects logistic regression model with all confounding variables of our primary analysis as fixed effects and the individual anesthesia providers as random effects.²³ 

We further studied a potential dose-dependent relationship of total phenylephrine dose with 7-day delirium using logistic and fractional polynomial regression modeling,²⁴  adjusted for the same variables as in the primary analysis, also including the duration of intraoperative hypotension. Last, we performed propensity score matching and inverse probability of treatment weighting. Details related to secondary analyses are described in Supplemental Digital Content 1 (section S2, https://links.lww.com/ALN/D310).

With an exploratory intent, we assessed a potential effect modification of the association between administration of phenylephrine versus ephedrine with postoperative delirium by a pre-existing diagnosis of cerebrovascular disease,²⁵  higher age as defined by the median in our cohort, and any previous diagnosis of traumatic brain injury.

To confirm the robustness of our findings, we performed several sensitivity analyses as described in detail in Supplemental Digital Content 1 (section S3, https://links.lww.com/ALN/D310). In a key-sensitivity analysis, we applied a keyword-based approach²⁶  paired with manual chart review in patients with available discharge notes to validate the use of International Classification of Diseases (Ninth/Tenth Revision, Clinical Modification) diagnostic codes for identification of delirium in this study. Subsequent sensitivity analyses included (1) multiple imputation of missing data for confounding variables; (2) 1:1 exact matching for patient age (± 4 yr); and (3) exact matching for duration (± 1 min) of intraprocedural hypotension. In addition, we conducted subgroup analyses by (4) excluding patients who received high doses of phenylephrine (more than 0.6 μg · kg⁻¹ · min⁻¹) as a surrogate for infusions; (5) excluding patients who received dopamine, epinephrine, or norepinephrine; and (6) using an established delirium screening tool, the Confusion Assessment Method assessment, for a subgroup of patients who were screened for delirium in the intensive care unit.⁶  We (7) calculated the E-value for our primary analysis to quantify unmeasured confounding. Further, we additionally adjusted our primary analysis for (8) intraoperative heart rate, (9) the occurrence of intraoperative hypoxemia or hypocapnia, and (10) sepsis within 7 days before the index procedure. We (11) conducted an analysis after reinclusion of patients with a pre-existing diagnosis of mild cognitive impairment into our study cohort. Last, we (12) added intraoperative heart rate to our inverse probability of treatment weighting model.

We applied multivariable logistic regression and fractional polynomial regression analyses adjusted for the aforementioned confounding variables. Linear combinations of the main effect and interaction terms were performed to assess potential effect modification of the primary association by different prespecified variables. The results are reported as adjusted odds ratios, adjusted absolute risk differences, or adjusted coefficients with 95% CI; α was set to 0.05. Analyses were performed using Stata (version SE 16.0, StataCorp LLC, USA) and R statistical software (version 4.2.0, Foundation for Statistical Computing, Austria).

A total of 117,954 adult hospitalized patients underwent general anesthesia for noncardiac, non-neurosurgical procedures with administration of either phenylephrine or ephedrine during the study period. After application of exclusion criteria and exclusion of patients with missing data, the final cohort consisted of 103,094 patients (fig. 1). Table 1 provides details related to patient characteristics and the distribution of variables by the administered type of vasopressor.

A total of 78,982 (76.6%) patients received phenylephrine, whereas 24,112 (23.4%) patients received ephedrine. The median [interquartile range] total intraoperative doses administered were 1.0 [0.2 to 3.3] mg for phenylephrine and 10.0 [10.0 to 20.0] mg for ephedrine. After treatment, 770 patients (0.8%) developed delirium: 685 (0.9%) patients who received phenylephrine and 85 (0.4%) patients who received ephedrine. The median [interquartile range] time with mean arterial blood pressure less than 55 mmHg was 0 [0 to 2] minutes in both groups (absolute mean standardized difference 0.126; table 1; figure S3). Figure 2 depicts the time course of mean arterial blood pressure between patients receiving phenylephrine and ephedrine. Median [interquartile range] mean blood pressure levels were higher in the phenylephrine compared to the ephedrine group (76.5 [76.0 to 77.0] vs. 74.0 [72.0 to 74.8] mmHg; P < 0.001). In adjusted analyses, intraoperative administration of phenylephrine was associated with higher adjusted odds of 7-day postoperative delirium compared to the use of ephedrine (adjusted odds ratio, 1.35; 95% CI, 1.06 to 1.71; and adjusted absolute risk difference, 0.2%; 95% CI, 0.1 to 0.3%; P = 0.015). Assessment of model calibration and discrimination indicated good model fit (Supplemental Digital Content 1, section S1.4, https://links.lww.com/ALN/D310).

The association between phenylephrine and delirium was dose-dependent. Patients who received more than 12.4 μg/kg of phenylephrine, a total dose larger than the median in our cohort, were at significantly higher risk of delirium (adjusted odds ratio, 1.21; 95% CI, 1.01 to 1.45; P = 0.035, compared to those who were administered lower doses). A Box–Tidwell test confirmed a suspected violation of the linearity assumption (P < 0.001). Therefore, we conducted fractional polynomial regression analysis adjusted for the primary confounder model including surgical duration and found that this trend was confirmed (adjusted coefficient, 0.08; 95% CI, 0.02 to 0.14; P = 0.013, per each μg/kg increase in the total phenylephrine dose; fig. 3). Propensity score matching (adjusted odds ratio, 1.53; 95% CI, 1.17 to 2.01; P = 0.002) and inverse probability of treatment weighting (adjusted odds ratio, 1.43; 95% CI, 1.08 to 1.88; P = 0.013) further confirmed the primary findings. Absolute standardized mean differences before and after inverse probability of treatment weighting are shown in table 1.

A wide variability in the preference for use of phenylephrine across all 637 individual anesthesia providers who performed anesthesia care in our study cohort was observed. The individual providers’ adjusted likelihood to use phenylephrine over ephedrine ranged from 15.6 to 95.5% across individual anesthesia providers (fig. 4). To account for this variability, we included the primary anesthesia provider who performed the case as a random effect in the mixed-effects model, yielding consistent findings with the primary analysis (adjusted odds ratio, 1.34; 95% CI, 1.05 to 1.70; P = 0.019). Details related to secondary analyses are provided in Supplemental Digital Content 1 (section S2, https://links.lww.com/ALN/D310).

There was no effect modification of the primary analysis by a pre-existing diagnosis of cerebrovascular disease (P for interaction = 0.256), higher age as defined by the median in our cohort (60 years or older, P for interaction = 0.079), or any previous diagnosis of traumatic brain injury (P for interaction = 0.964). In 45,465 patients with available discharge summaries, we observed 1,434 (3.2%) cases of postoperative delirium based on keyword search and manual chart review of discharge summaries. A key-sensitivity analysis using this outcome definition confirmed our primary findings with an even larger effect size and smaller P value (adjusted odds ratio, 1.88; 95% CI, 1.49 to 2.37; P < 0.001). Further details to this method are presented in Supplemental Digital Content 1 (section S3.1, https://links.lww.com/ALN/D310). Our findings remained robust across all subsequent sensitivity analyses, including multiple imputation of missing data (adjusted odds ratio, 1.41; 95% CI, 1.12 to 1.78; P = 0.004), exact matching for patient age (adjusted odds ratio, 1.35; 95% CI, 1.01 to 1.79; P = 0.040), and exact matching for the duration of intraprocedural hypotension (adjusted odds ratio, 1.38; 95% CI, 1.04 to 1.83; P = 0.027) as described in detail in Supplemental Digital Content 1 (section S3, https://links.lww.com/ALN/D310). Subgroup analyses excluding patients who received high doses of phenylephrine (more than 0.6 μg · kg⁻¹ · min⁻¹; adjusted odds ratio, 1.36; 95% CI, 1.07 to 1.73; P = 0.012), excluding all patients who received dopamine, epinephrine, or norepinephrine (adjusted odds ratio, 1.28; 95% CI, 1.01 to 1.63; P = 0.044), and in patients who were screened for delirium using the Confusion Assessment Method (adjusted odds ratio, 2.13; 95% CI, 1.04 to 4.38; P = 0.039) further confirmed our findings. The E-value, defined as the minimum strength of association on the risk ratio scale, that an unmeasured confounding factor would need to have with both the exposure and the outcome, conditional on the measured confounders, to fully explain away a specific exposure–outcome association^27,28  was 2.04 for the point estimate and 1.31 for the CI of our primary analysis. Therefore, considerable unmeasured confounding would be needed to explain away the observed effect estimate. Additional confounding adjustment for intraoperative heart rate (adjusted odds ratio, 1.30; 95% CI, 1.01 to 1.66; P = 0.039), the occurrence of hypoxemia or hypocapnia (adjusted odds ratio, 1.34; 95% CI, 1.06 to 1.71; P = 0.016), and preoperative sepsis (adjusted odds ratio, 1.35; 95% CI, 1.06 to 1.72; P = 0.014) yielded confirmatory results. Our findings were further confirmed after reinclusion of patients with a pre-existing diagnosis of mild cognitive impairment (adjusted odds ratio, 1.34; 95% CI, 1.05 to 1.70; P = 0.017), as well as after adding intraoperative heart rate to the inverse probability of treatment weighting model (adjusted odds ratio, 1.52; 95% CI, 1.10 to 2.08; P = 0.010).

In this multicenter retrospective cohort study in hospitalized adult patients who underwent noncardiac, non-neurosurgical procedures under general anesthesia at two academic healthcare centers, we observed that despite higher blood pressure levels, intraoperative utilization of phenylephrine, in comparison to ephedrine, was associated with higher odds of postoperative delirium within 7 days after surgery. In addition, there was a dose-dependent relationship between the intraoperatively administered phenylephrine dose and the patient’s risk of postoperative delirium. Neither excluding patients who received high doses of phenylephrine nor excluding those with concomitant use of other vasopressors affected these findings.

This study is the first to establish an association between treatment of intraprocedural hypotension with phenylephrine over ephedrine and postoperative delirium. While the etiology of postoperative delirium is multifactorial, cerebral hypoperfusion and impaired oxygenation have been proposed as important contributing factors.^4,6  With evidence linking hypotension and postoperative delirium,⁴  consensus-based guidelines on the prevention of postoperative delirium have highlighted intraoperative hypotension as a major risk factor of postoperative delirium.²⁹  Our data show, however, that while mean arterial blood pressures were slightly higher in patients receiving phenylephrine versus ephedrine, phenylephrine was associated with a higher risk of delirium.

Our findings may be explained by previous studies that reported differential effects of phenylephrine and ephedrine on cerebral circulation and oxygenation.^9,30  Phenylephrine has been associated with a reduction in cerebral oxygenation, when compared to other vasopressors,³¹  independent of the mean arterial pressure across the study groups.^9,30  However, this remains controversial.^32,33  For example, Koch et al.⁹  did not observe a decrease in cerebral blood flow or an increase in oxygen extraction fraction in the phenylephrine group. An experimental study in healthy piglets further suggested that while cerebral oxygen saturation measured using near infrared spectroscopy declined, the cerebral partial pressure of oxygen even increased during phenylephrine infusion.³⁴  To elucidate potential pharmacodynamic mechanisms for our findings, pharmacophysiological studies in a clinical setting are now needed. Based on our findings and in the context of previous studies, cerebral perfusion and oxygenation in patients undergoing general anesthesia with phenylephrine versus ephedrine could be targeted.

Our study adds to the body of evidence suggesting that phenylephrine may not be an ideal agent for treating intraoperative hypotension, with its associated higher risk of postoperative delirium. Based on our data, we suggest that clinicians could pursue strategies to preserve cerebral oxygenation in the event of intraoperative hypotension with the help of cerebral oxygen saturation monitoring. Especially if high doses of phenylephrine would be required to treat hypotension, clinicians may consider the use of other agents such as ephedrine to reduce the risk of delirium. Randomized trials comparing intraoperative administration of phenylephrine versus ephedrine are now warranted. These trials could further be enriched by including patients aged older than 65 yr. In our study, this specific subgroup of patients had an incidence of postoperative delirium (based on chart review) of 5.3%. Based on a two-sided α of 0.05, statistical power set to 80%, a sample size of 1,214 patients/group would be required in a future clinical trial to detect our observed 34% increase in delirium with phenylephrine.

This was a retrospective analysis utilizing anesthesia records and hospital registry data, and potential confounding by indication needs to be addressed. Confounding by indication related to factors such as heart rate, one of the primary drivers for using ephedrine versus phenylephrine in clinical routine, response to induction of anesthesia, or systemic inflammation needs to be ruled out. Phenylephrine is often used in patients with more extensive hypotension and higher vasopressor needs, which is reflected in our cohort. To account for this, we adjusted our analyses for a variety of patient-related, anesthesia, and procedural factors, including the duration of intraprocedural hypotension and conducted additional sensitivity analyses adjusting for heart rate and sepsis before surgery. This extensive confounder adjustment was possible due to the large sample size from two academic healthcare networks. We further conducted a propensity score–matched analyses and applied inverse probability of treatment weighting to address potential issues related to confounding by indication, and these analyses confirmed our primary findings. The E-value of 2.04 for the point estimate and 1.31 for the CI of the primary analysis further suggested that any unmeasured confounder would require a strong impact on both the exposure (phenylephrine vs. ephedrine) and outcome (delirium) variable to explain away the primary association. Finally, we observed considerable variability in the preference for either phenylephrine or ephedrine among our anesthesia providers, which was independent from patient or procedural characteristics. Furthermore, excluding patients with high (more than 0.6 μg · kg⁻¹ · min⁻¹) phenylephrine doses yielded consistent results. Another limitation is that the data analyzed were derived from a limited geographical region in the United States and reflect the practice in two large tertiary medical centers. Hence, our results cannot necessarily be generalized to other healthcare settings. Further, the use of International Classification of Diseases (Ninth/Tenth Revision, Clinical Modification) diagnostic codes to identify delirium after surgery, which is specific but has limited sensitivity in defining delirium cases,³⁵  is a limitation of our study. To address this limitation, we employed a keyword-based selection of patient discharge summaries indicative of delirium followed by manual, systematic chart review by a team of experienced anesthesiologists, intensivists, and research fellows. After this approach in a subset of patients, we observed a 3-fold higher incidence of postoperative delirium amounting to 3.2%, which is in line with previous literature.³⁶  A key-sensitivity analysis using this outcome definition confirmed our primary findings with even larger effect size, corroborating the notion that limited sensitivity of International Classification of Diseases–based postoperative delirium likely underestimated the observed effect. Moreover, our subgroup analysis in patients who were screened for delirium using the Confusion Assessment Method confirmed the primary finding. Certainly, future studies incorporating more accurate strategies to identify postoperative delirium are warranted.

In hospitalized adult patients who underwent noncardiac, non-neurosurgical procedures under general anesthesia, the intraoperative utilization of phenylephrine, in comparison to ephedrine, was associated with higher odds of postoperative delirium within 7 days after surgery. These findings deliver important information to future trials that are now warranted to investigate optimum treatment of intraoperative hypotension.

The authors are grateful to Laetitia S. Chiarella, B.Sc., and Hannah Hamdani (Center for Anesthesia Research Excellence, Beth Israel Deaconess Medical Center, Boston, Massachusetts) for making nonauthor contributions during data analyses. Elena Ahrens conducted this work in partial fulfilment of the requirements for an M.D. thesis.

Support was provided solely from institutional and/or departmental sources.

Dr. Ma received a SEAd Grant from the Society for Education in Anesthesia (Milwaukee, Wisconsin), not related to this article. Dr. Subramaniam received funding for investigator-initiated studies from Merck & Co. (Rahway, New Jersey) and the National Institutes of Health (Bethesda, Maryland) that do not pertain to this article; he is a consultant for Masimo, Inc. (Irvine, California) Dr. Rhee received funding for investigator-initiated studies from the National Institutes of Health that do not pertain to this article; he is a consultant for Takeda Pharmaceuticals, Inc. (Tokyo, Japan). Dr. Schaefer received funding for investigator-initiated studies from Merck & Co. that do not pertain to this article; he is an associate editor for BMC Anesthesiology, received honoraria for lectures from Fisher & Paykel Healthcare (Auckland, New Zealand) and Mindray Medical International Limited (Shenzen, China), and received an unrestricted philanthropic grant from Jeffrey and Judith Buzen. The other authors declare no competing interests.

Supplemental Digital Content 1: Supplemental methods and results, https://links.lww.com/ALN/D310

Supplemental Digital Content 2: STrengthening the Reporting of OBservational studies in Epidemiology Checklist, https://links.lww.com/ALN/D311