Delirium in Older Patients after Combined Epidural–General Anesthesia or General Anesthesia for Major Surgery: A Randomized Trial

This multicenter, randomized trial was conducted in five tertiary care hospitals in Beijing, China. The rationale and design of the study were reported previously.²⁵  The study protocol was approved by the Peking University Institutional Review Board (approval No. 00001052-11048; principal investigator: D.-X.W.) on July 28, 2011, and by the ethics committees of the five participating centers; changes to the trial methods and outcomes are reported in Supplemental Digital Content 1 (https://links.lww.com/ALN/C624). All participants provided written informed consent. The Peking University Clinical Research Institute was responsible for the study monitoring, data quality assessment and management, and data analysis. The study was registered with the Chinese Clinical Trial Registry (www.chictr.org.cn; identifier: ChiCTR-TRC-09000543) and ClinicalTrials.gov (identifier: NCT01661907). Long-term results will be reported in a companion article.

We enrolled patients aged 60 to 90 yr old who were scheduled for major noncardiac thoracic or abdominal surgery expected to last at least 2 h who agreed to use patient-controlled analgesia after surgery. We included patients having thoracoscopic or laparoscopic surgery when the expected incision length was at least 5 cm. We excluded patients who had severe neurologic conditions, acute myocardial infarction or stroke within 3 months, any contraindication for epidural anesthesia, severe heart dysfunction, severe liver dysfunction (Child–Pugh grade C), or renal failure.

Patients were centrally randomized using computer-generated codes with a block size of four, stratified by trial site and type of surgery (thoracic or abdominal). Participants were randomized in a 1:1 ratio to either general anesthesia with postoperative intravenous analgesia or combined epidural–general anesthesia with postoperative epidural analgesia. Allocation was concealed until shortly before anesthesia induction or epidural puncture with a 24-h interactive web system (IWRS, Brightech Clinical Information Management System, CIMS Global, USA).

Premedication was not permitted in either group, including anticholinergic drugs, sedatives, or dexmedetomidine. In patients assigned to general anesthesia alone, anesthesia was induced with midazolam (0.02 to 0.03 mg/kg), propofol, and sufentanil. Muscle relaxation was achieved using rocuronium. Anesthesia was maintained with a propofol infusion and/or the volatile anesthetic sevoflurane and/or the inhaled gas nitrous oxide. Postoperative analgesia was provided with a patient-controlled intravenous analgesia with morphine (0.5 mg/ml). The patient-controlled pump was programed to deliver 2-ml boluses with a lockout interval of 6 to 10 min and a background infusion at 1 ml/h.

In patients assigned to combined epidural and general anesthesia, the epidural catheter was inserted before induction of general anesthesia at an intervertebral space selected by the responsible anesthesiologist. Successful epidural block was confirmed by injection of 3 to 4 ml of 2% lidocaine and subsequently maintained with 0.375 to 0.5% ropivacaine during surgery. General anesthesia was induced and maintained as in the general anesthesia alone group, including administration of midazolam (0.02 to 0.03 mg/kg). Postoperative pain was treated with patient-controlled epidural analgesia using a solution of 0.12% ropivacaine and 0.5 μg/ml sufentanil. The pump was programed to deliver 2-ml boluses with a lockout interval of 20 min and a background infusion of 4 ml/h. For patients with failed epidural blocks (including failed catheterization, inadequate analgesia, blocked catheters, and accidental catheter dislodgement), general anesthesia was provided with postoperative patient-controlled intravenous analgesia.

Routine management for intraoperative hypotension included reducing anesthetic depth, fluid infusion, and administration of vasopressors such as ephedrine, phenylephrine, epinephrine, and/or norepinephrine. When indicated, clinicians were permitted to decrease or cease administration of epidural ropivacaine. Supplemental postoperative analgesia was provided at the discretion of attending surgeons or intensive care unit (ICU) physicians and could include opioids, nonsteroidal anti-inflammatory drugs, and other analgesics. Morphine equivalent doses were estimated for comparison of opioid consumption.^12,26,27  Adverse events were managed per routine.

Patients and anesthesiologists were aware of study group allocation. However, research staff who did not perform outcome assessments hid patient-controlled analgesia apparatus from investigators who performed assessments who otherwise had no knowledge of randomization and were not permitted to communicate with either patients or care providers about group assignment or treatment.

Baseline data included the Charlson Comorbidity Index.²⁸  We also evaluated activities of daily living with the Barthel Index, which ranges from 0 to 100, with higher scores indicating better activities.²⁹  Cognitive function was evaluated with the Mini-Mental State Examination with scores ranging from 0 to 30, with higher scores indicating better function.³⁰  Anxiety and depression were evaluated with the Hospital Anxiety and Depression Scale, with scores ranging from 0 to 21 for either anxiety or depression, with higher score indicating more severe symptoms. Scores greater than 7 were considered thresholds for both anxiety and depression.³¹ 

Routine intraoperative monitoring included electrocardiogram, noninvasive blood pressure, pulse oxygen saturation, end-tidal carbon dioxide, volatile anesthetic concentration, and urine output. Intraarterial pressure and central venous pressure were monitored when necessary. For patients admitted to the ICU after surgery, the electrocardiogram, intraarterial pressure, and pulse oxygen saturation were monitored continuously. For patients sent back to the general wards after surgery, electrocardiogram, noninvasive blood pressure, and pulse oxygen saturation were monitored continually through the first postoperative morning and then once or twice daily until hospital discharge. Clinicians instituted more frequent monitoring or transfer to an intensive care unit as indicated.

Our primary outcome was delirium, which was assessed dichotomously with the Confusion Assessment Method for the ICU.³²  The Chinese version of the Confusion Assessment Method for the ICU has been validated in spontaneously ventilating patients with acceptable sensitivity and specificity,³³  and we have considerable experience with the technique.^34,35  Delirium was assessed twice daily (between 8 and 10 am and between 6 and 8 pm) during the first 7 postoperative days or until hospital discharge or death if earlier. Immediately before assessing delirium, sedation or agitation was assessed using the Richmond Agitation Sedation Scale, with scores ranging from –5 (unarousable) to +4 (combative), where 0 indicates alert and calm.³⁶  For deeply sedated or unarousable patients (Richmond Agitation Sedation Scale score of –4 or –5), delirium was not assessed, and the patient was recorded as comatose.

Patients with delirium were classified into three subtypes: hyperactive (Richmond Agitation Sedation Scale score consistently positive, from +1 to +4), hypoactive (Richmond Agitation Sedation Scale score consistently neutral or negative, from –3 to 0), and mixed.³⁷  Investigators who performed follow-up and delirium assessment (G.-J.S., Q.M., Huai-Jin Li, Y. Zhao, H.K., D.H., C.-M.D., Y. Zhang, S.-T.H., P.-F.L., Y.L., and H.-Y.Z.) were trained to use the Confusion Assessment Method for the ICU by a psychiatrist (X.-Y.S.). The training program included lectures introducing delirium and the Confusion Assessment Method for the ICU, as well as simulation with actors. Initial training continued until delirium diagnoses reached 100% agreement between investigators and the psychiatrist and was repeated two to three times a year throughout data acquisition.

Secondary outcomes included ICU admission after surgery, time to onset of delirium, time to oral fluid/food intake, postoperative duration of hospitalization, and 30-day all-cause mortality. For patients admitted to the ICU after surgery, the worst Acute Physiology and Chronic Health Evaluation II (APACHE II) score within 24 h, the percentage with endotracheal intubation, the duration of mechanical ventilation (for those with endotracheal tubes), and the length of ICU stay were recorded. An additional secondary outcome was major complications other than delirium, defined as new-onset medical conditions that were deemed harmful and required therapeutic intervention (i.e., grade II or higher on the Clavien–Dindo classification).³⁸ 

Other prespecified outcomes included pain severity both at rest and with movement, which were assessed with the Numeric Rating Scale (an 11-point scale, where 0 denotes no pain and 10 the worst pain) twice daily at the time of delirium assessment during the first 3 postoperative days. After the first 7 days, evaluations were performed weekly until 30 days after surgery. Discharged patients were contacted by phone.

We recorded anesthetic-related adverse event for 3 postoperative days and thereafter recorded complications until 30 days after surgery. Among anticipated hemodynamic abnormalities, we defined intraoperative hypotension as systolic blood pressure less than 80 mmHg, intraoperative hypertension as systolic blood pressure greater than 180 mmHg, postoperative hypotension as systolic blood pressure less than 90 mmHg, and postoperative hypertension as systolic blood pressure greater than 160 mmHg.

Patients were primarily analyzed within the groups to which they were assigned, whether or not the designated treatment was received, excluding those with repeated randomizations, cancelled surgeries, or consent withdrawal before anesthesia (modified intention-to-treat population). For the primary outcome, analysis was also performed in the per-protocol population, based on the treatment received.

Our primary outcome, the incidence of postoperative delirium within 7 days, was compared by a chi-square test. A similar analysis was used for the per-protocol analysis. For patients with missing data because of early hospital discharge or death, the last delirium assessment results were considered as the final results. Exploratory analyses were performed to assess differences of the primary outcome in predefined subgroups. Treatment-by-covariate interactions were assessed separately for each subgroup factor using logistic regression.

Continuous variables were analyzed with independent-sample t tests for normally distributed data or Mann–Whitney U tests. Differences (and 95% CI) between medians were calculated with Hodges–Lehmann estimators. Categorical variables were analyzed with chi-square tests with continuity correction or Fisher exact tests. Time-to-event results were analyzed with Kaplan–Meier survival analyses with log-rank tests; patients who died within 30 days were censored at the time of death. Missing data were not replaced.

For each hypothesis, a two-sided P < 0.05 was considered statistically significant. For the treatment-by-covariate interaction in predefined subgroup analyses, a P < 0.10 was considered statistically significant. Statistical analyses were performed with SAS 9.3 software (SAS, USA) and SPSS 25.0 software (IBM SPSS, USA).

An independent data quality committee from the Peking University Clinical Research Institute monitored compliance and completeness of the data, and the Peking University Institutional Review Board reviewed the results and determined whether the trial should be suspended because of high incidence of violations or clear evidence of harm. There were no interim analyses for efficacy or futility.

Sample size was based on a cohort of patients at our facility in whom the incidence of postoperative delirium was 13.1% in older patients given general anesthesia for major abdominal surgery. We estimated that a sample size of 1,664 participants (832 per group), would provide 80% power for detecting a one-third reduction in the primary outcome, with a two-sided significance level of 0.05. We therefore planned to enroll 1,800 patients with the expectation that 7.5% would drop out.

Between November 21, 2011, and May 25, 2015, a total of 3,049 patients were screened for inclusion. Of these, 2,199 were eligible, and 1,802 were enrolled and randomly assigned to either combined epidural–general anesthesia (n = 901) or general anesthesia alone (n = 901). Among the enrolled patients, 8 were excluded because of repeated randomizations, 49 were excluded because surgery was cancelled, and 25 withdrew consent before anesthesia. A total of 1,720 patients were therefore included in the modified intention-to-treat population, with 857 given combined epidural–general anesthesia and 863 given general anesthesia alone. There was a total of 118 protocol deviations, leaving 1,602 patients included in the per-protocol analysis (776 in the epidural–general anesthesia group and 826 in the general anesthesia group; fig. 1).

Demographic and baseline variables were well balanced between the two groups except that preoperative hypertension was less common, and creatinine concentrations were lower in patients assigned to combined epidural–general anesthesia (Table 1). Patients with combined epidural–general anesthesia were given epidural lidocaine (median, 60 mg [interquartile range, 40 to 80]) and ropivacaine (median, 85 mg [interquartile range, 60 to 125]). As expected, patients in the epidural–general anesthesia group consumed less volatile anesthesia, opioids, cisatracurium, and nonsteroidal anti-inflammatory drugs; additionally, they received more artificial colloids and had lower mean arterial pressures, higher heart rates, and greater urine output. Patients assigned to epidural–general anesthesia were given more epidural sufentanil but less intravenous morphine during the first 7 postoperative days. Total perioperative morphine equivalent consumption was significantly less in the combined epidural–general anesthesia patients (mean difference, −32 mg; 95% CI, −41 to −23]; Table 2 and Table S1 in Supplemental Digital Content 2, https://links.lww.com/ALN/C625).

The incidence of postoperative delirium within 7 days was significantly lower in patients assigned to epidural–general anesthesia (15 [1.8%] of 857 patients) than in the general anesthesia group (43 [5.0%] of 863 patients; relative risk, 0.351; 95% CI, 0.197 to 0.627; P < 0.001; number needed to treat, 31; fig. 2). The per-protocol analysis showed a similar difference (11 [1.4%] of 776 patients vs. 39 [4.7%] of 826 patients; relative risk, 0.300, 95% CI, 0.155 to 0.582; P < 0.001). All three subtypes of delirium were significantly less common in the epidural–general anesthesia patients (Table 3). In subgroup analyses, we found a significant interaction for the primary outcome between treatment group and study center (center 1 vs. others; P = 0.067); there were no significant interactions between treatment group and other predefined factors. The effect of combined epidural– general anesthesia on delirium was roughly similar across all subgroups (fig. 3).

Among 339 (20%) patients admitted to the ICU after surgery, those assigned to combined epidural–general anesthesia were 33% less likely to remain intubated, and ICU duration was about 5% shorter. Among all patients, moderate- to-severe pain (Numeric Rating Scale pain score of 4 or higher) at rest was significantly less common in the epidural–general anesthesia group on the first postoperative morning (relative risk, 0.72; 95% CI, 0.55 to 0.95; P = 0.019); moderate-to-severe pain during movement was also significantly less common on the first postoperative morning (relative risk, 0.82; 95% CI, 0.73 to 0.93; P = 0.001) and afternoon (relative risk, 0.83; 95% CI, 0.74 to 0.95; P = 0.005), and the second postoperative morning (relative risk, 0.86; 95% CI, 0.74 to 0.99; P = 0.040). Other secondary outcomes including nondelirium complications within 30 days did not differ between the two groups (Table 3 and Table S2 in Supplemental Digital Content 2, https://links.lww.com/ALN/C625).

Patients randomized to epidural–general anesthesia had more intraoperative hypotension (421 [49%] vs. 288 [33%]; relative risk, 1.47; 95% CI, 1.31 to 1.65; P < 0.001), spent more time with mean arterial pressure of less than 65 mmHg (17 min [interquartile range 3 to 42] vs. 8 min [0 to 25]), and were more likely to require vasopressors (495 [58%] vs. 387 [45%]; relative risk, 1.29; 95% CI, 1.17 to 1.41; P < 0.001); in contrast, they had less hypertension (183 [21%] vs. 302 [35%]; relative risk, 0.61; 95% CI, 0.52 to 0.71; P < 0.001). Over the initial 3 postoperative days, patients assigned to epidural–general anesthesia were less likely to experience hypertension (64 [7%] vs. 161 [19%]; relative risk, 0.40; 95% CI, 0.30 to 0.53; P < 0.001) or postoperative nausea and vomiting (80 [9%] vs. 116 [13%]; relative risk, 0.69; 95% CI, 0.53 to 0.91; P = 0.007). One patient in the epidural–general anesthesia group died from pulmonary embolism on the first day after surgery, a complication that was considered unrelated to study group assignment (Table 4).

There was less delirium during the first 7 postoperative days in older patients randomized to combined epidural–general anesthesia for major thoracic and abdominal surgeries compared with general anesthesia alone. The reduction was consistent across all three motoric subtypes of delirium and similar for all predefined subgroups. The treatment effect was substantial and highly statistically significant, with the incidence being only about a third in patients assigned to combined epidural–general anesthesia. Because the incidence of delirium in the general anesthesia alone group was only 5%, the number needed to treat was 31 (the reciprocal of the absolute risk reduction). We note, however, that the same relative treatment effect would correspond to a number needed to treat of 15 at a baseline delirium risk of 10% and 10 at a baseline risk of 15%—both of which are well within reported ranges.^2,4,5 

Two systematic reviews reported that the incidence of delirium was similar after neuraxial and general anesthesia for hip fracture surgery.^10,19  However, more recent observational analyses observed less delirium in patients who had neuraxial rather than general anesthesia for hip or knee arthroplasties.^20–22  One small trial randomized 70 older patients to general anesthesia and intravenous analgesia or combined epidural–general anesthesia with epidural analgesia; that is, a protocol similar to ours. Despite better analgesia and improved mental status in the combined anesthesia patients, the frequency of postoperative delirium was similar in each group (24% vs. 26%, respectively). Because the number of events was small, only 16, CIs around the true treatment effect were large.³⁹  Our results based on more than 25 times as many enrolled patients and 4 times as many outcomes are presumably more reliable. We did not observe any statistically significant or clinically important subgroup differences, suggesting that the effects of combining epidural and general anesthesia on delirium apply broadly.

Various mechanisms may contribute to delirium sparing in patients given combined epidural–general anesthesia. First, epidural blocks reduced the consumption of general anesthetics during surgery; specifically, sevoflurane exposure was reduced by 18%, which is consistent with previous reports.²³  Previous work shows that deep general anesthesia is associated with more delirium⁴⁰  and that deep propofol sedation during spinal anesthesia promotes delirium.⁴¹  Reduced general anesthetic consumption in the combined epidural–general anesthesia group may therefore have contributed to less delirium in the combined anesthesia patients.

Epidural blocks improved postoperative analgesia; specifically, moderate-to-severe pain was reduced by 28% at rest and by 18% with movement, benefits that are consistent with previous studies.¹⁸  Severe pain is an important risk factor of postoperative delirium.¹³  Better analgesia with epidural blocks might therefore have helped to reduce delirium.

Opioids are strongly associated with delirium.^42,43  Because analgesia was better, patients randomized to combined epidural–general anesthesia were given only 28% as much intraoperative opioid and postoperatively were given a short-acting opioid rather than morphine. Reduced opioid consumption and use of sufentanil rather than morphine may therefore have reduced delirium in patients given combined anesthesia.⁴⁴ 

An additional factor is that epidural analgesia blunts the stress and inflammatory responses to surgical tissue injury.^17,24  Because inflammation is thought to promote delirium, blunted stress responses may have reduced the incidence of postoperative delirium in patients assigned to combined anesthesia.¹⁴  Finally, intubation and concomitant sedation promotes delirium.⁴⁵  Because epidural blocks decreased the proportion of patients who were admitted to the ICU with endotracheal intubation, delirium might have been reduced as well.⁴⁵ 

The incidence of delirium in our general anesthesia alone group was lower than that reported in many previous studies including ours,^{8,12,34,35,41}  but within previously reported ranges.^3,5,46  Our lower incidence presumably reflects relatively low baseline risk. For example, we enrolled patients as young as 60 yr, whereas many delirium trials restrict enrolment to patients exceeding 65 or even 70 yr. Consequently, when compared with other studies, our patients were younger, had fewer comorbidities, had and better baseline Mini-Mental State Examination scores^34,35,41 —all of which presumably reduced the incidence of delirium.¹³  We also took precautions to reduce delirium. For example, we did not allow premedication with sedatives and/or anticholinergics and used an ultra-short-acting opioid intraoperatively. Postoperative nursing care has also improved in recent years and now routinely includes early mobilization and efforts to minimize night-time disruptions.¹³  Finally, about a third of our patients had minimally invasive surgeries, which presumably reduce surgical stress and consequent inflammation, both of which are thought to contribute to delirium.¹⁴ 

Epidural blocks are considered to be safe in patients without specific contraindications.¹⁶  Our results are consistent because the incidence of severe adverse events was low, and none was attributed to epidural anesthesia and analgesia. However, epidural anesthesia significantly increased the incidence of intraoperative hypotension and the need for vasopressor treatment, which is a well known consequence of combining general and epidural anesthesia.¹⁸  In recent years intraoperative hypotension has been linked to delirium,⁴  myocardial injury,⁴⁷  acute kidney injury,⁴⁷  and even perioperative mortality,⁴⁸  although there remains limited randomized evidence of harm.⁴⁹  The benefit of combined epidural–general anesthesia thus needs to be balanced against potential risks of hypotension in individual patients.

For pragmatic reasons, participants and care providers were not masked from group assignment. However, investigators who performed postoperative follow-ups and outcome assessment did not participate in perioperative care and had no knowledge of treatment assignments, although blinding was surely imperfect. We only enrolled patients scheduled for major thoracic and abdominal surgeries, and patients with severe comorbidities were excluded. Our results can presumably be generalized to other major noncardiac operations; the benefits of combined epidural–general anesthesia may differ for minor operations or cardiac surgery, which has a higher baseline delirium incidence. For various reasons, we excluded 82 (5%) patients from our intention-to-treat analysis and 200 (11%) patients from our per-protocol analysis. Most exclusions were for technical reasons that seem unlikely to have resulted from bias. As might therefore be expected, results were similar with intention-to-treat and treatment-received analyses.

With 1,720 patients completing the trial, ours is far larger than others comparing combined epidural–general anesthesia with general anesthesia alone. However, the baseline incidence of delirium was low in the reference group and even lower in the combined epidural–general group. Consequently, the total number of delirium cases was only 58. Thus, from the perspective of outcome events, the trial is relatively small. Furthermore, a factor-of-three reduction in a complex and multifactorial outcome such as delirium seems unlikely. It is therefore plausible—and perhaps likely—that the true effect of combined epidural–general anesthesia on delirium is less than we observed. Pain evaluations were suboptimal because we assessed pain intensity just twice daily, starting the first postoperative morning; furthermore, some data were missing in some cases. The use of patient-controlled analgesia with background continuous morphine infusion in the general anesthesia group may threaten the external validity of our results. In a companion paper, we report results for an outcome at 5 yr, thus giving us two primary outcomes. We did not correct for multiplicity, but our results are robust (P < 0.001) and statistical compensation for multiple outcomes would not change our conclusions.

In summary, delirium is a common and serious postoperative complication with few if any established preventive measures. Older patients randomized to combined epidural–general anesthesia with epidural analgesia for major noncardiac surgeries had one third as much delirium compared with those assigned to general anesthesia alone with opioid analgesia. Patients given combined epidural–general anesthesia also required less opioid and experienced less nausea and vomiting—but had more hypotension. Clinicians deciding whether to use combined epidural–general anesthesia for prevention of delirium should consider baseline delirium risk (which strongly influences the number-needed-to-treat) and individual patient risk of hypotension.

The authors gratefully acknowledge Xin-Yu Sun, M.D., and Shu-Zhe Zhou, M.D. (Department of Geriatric Psychiatry, Peking University Sixth Hospital, Beijing, China) and Cui Yuan, M.N., Xiao-Jing Liu, associate B.N., Xin-Zhu Lu, B.N., and Fang Wang, B.N. (Department of Critical Care Medicine, Peking University First Hospital, Beijing, China) for their help with psychiatric consultation and training courses.

Supported by Peking University Clinical Research Program (Beijing, China) grant No. PUCRP201101.

The authors declare no competing interests.

Full protocol available at: dxwang65@bjmu.edu.cn. Raw data available at: dxwang65@bjmu.edu.cn.

Department of Anesthesiology and Critical Care Medicine, Peking University First Hospital, Beijing, China: Yuan Zeng, M.D., Dong-Liang Mu, M.D., Ya-Fei Liu, M.D., Wei-Jie Zhou, M.D., Guo-Jin Shan, associate B.S.Med.Tech., Qiong Ma, associate B.S.Med.Tech., Xue-Yi Zheng, M.D., Cong Fu, M.D., Yue Zhang, M.D., Ph.D., Xin-Quan Liang, M.D., Chao Liu, M.D., Shu-Ting He, M.D., Tong Cheng, M.D., Si-Ming Huang, M.D., Ya-Ting Du, M.D., Si-Chao Xu, M.D. Department of Anesthesiology, Beijing Shijitan Hospital, Capital Medical University, Beijing, China: Run Wang, M.D., Li Xiao, M.D., Jing Zhang, M.D., Wen-Zheng Yang, M.D.

Department of Anesthesiology, Peking University Third Hospital, Beijing, China: Wei-Ping Liu, M.D., Wen-Yong Han, M.D.

Department of Anesthesiology, Peking University People’s Hospital, Beijing, China: Yao Yu, M.D.

Department of Anesthesiology, Beijing Hospital, National Centre of Gerontology; Institute of Geriatric Medicine, Chinese Academy of Medical Sciences, Beijing, China: Zhen Hua, M.D., Jing-Jing Zhang, M.D.