Spinal Anesthesia with Targeted Sedation based on Bispectral Index Values Compared with General Anesthesia with Masked Bispectral Index Values to Reduce Delirium: The SHARP Randomized Controlled Trial

The research protocol was approved by the Mercy Medical Center (Baltimore, Maryland) Institutional Review Board (No. 2015-45). The trial was registered at ClinicalTrials.gov (NCT03133845, Principal Investigator Charles Brown). The initial protocol was released by the investigators to ClinicalTrials.gov on October 23, 2015. Due to quality control issues (in particular, the specificity of some outcomes, most notably postdischarge secondary outcomes that are not reported in this manuscript), the protocol was not formally registered and released to the public until April 2017, so the formal registration was retrospective to the start of the trial. The primary aim and outcome as reported in this manuscript have been unchanged since the initial submission to ClinicalTrials.gov on October 23, 2015. However, the secondary delirium outcomes (delirium severity and number of days of delirium) were not formally added to the trial registration until April 2017, although these outcomes were collected since the start of the trial as part of the study protocol. Other changes in enrollment criteria and sample size calculation are described below. Participants provided written informed consent. The SHaping Anesthetic techniques to Reduce Postoperative delirium (SHARP) study was conducted as a single-center prospective randomized controlled superiority trial with two parallel groups. The protocol was published near the end of the trial to summarize the conduct of the trial and provide the final statistical plan.²¹ 

Patients were approached before scheduled surgery by a research coordinator to evaluate eligibility and obtain informed consent. Inclusion criteria were (1) age 65 yr or greater; (2) undergoing lumbar spine fusion; (3) expected surgery duration less than 3 h; (4) under the care of a participating surgeon; and (5) ability to understand and comply with study procedures. Exclusion criteria were (1) contraindications to spinal anesthesia (e.g., severe aortic stenosis, anticoagulant therapy); (2) body mass index greater than 40 kg/m²; (3) previous L2–L5 full lumbar fusion; (4) communication issues precluding baseline assessments; (5) baseline dementia or Mini-Mental State Examination less than 24; (6) psychiatric disease precluding cooperation with sedation; and (7) surgeon or anesthesiologist preference for either anesthetic approach for any reason due to clinical considerations. Delirium was not formally assessed, although all patients were assessed for capacity to consent. Patients were enrolled between September 2015 to May 2019. Eligibility criteria were expanded after the study began to allow slightly younger patients, a higher body mass index, and longer duration of surgery. The specific criteria that were changed were a decrease in the lower age limit from 70 yr to 65 yr, an increase in the upper limit of body mass index (from 35 kg/m² to 40 kg/m²), and an increase in the upper limit of anticipated surgery duration (from 2 h to 3 h).

A computer-generated simple randomization list with 1:1 allocation was created by a research nurse before the study. For allocation concealment, assignments were placed in sealed opaque envelopes, which were sequentially handed to clinicians after randomization, before entering the operating room.

The intervention group received spinal anesthesia with targeted depth of anesthesia based on BIS values. The BIS monitor is approved to monitor depth of anesthesia and displays a unitless number (0 to 100) derived from processed electroencephalogram waveforms. BIS values between 40 and 60 are consistent with general anesthesia.²²  In the intervention group, spinal anesthesia was obtained using intrathecal injection of bupivacaine (10 to 15 mg) or lidocaine. Patients received sedation with propofol (25 to 150 mcg · kg^–1 · min^–1), targeted to a BIS greater than 60 to 70. However, the anesthesiologist was instructed to prioritize clinical concerns if depth of sedation needed to be increased.

In the control group, patients received general anesthesia with an endotracheal tube. Anesthesia induction was with propofol (1 to 2 mg/kg) or etomidate, maintenance with a volatile anesthetic, muscle relaxation with a nondepolarizing muscle relaxant, and analgesia with fentanyl (generally 2 to 5 mcg/kg titrated) or hydromorphone and/or morphine. Patients on baseline opioids could receive additional opioids based on clinical criteria. For patients under general anesthesia, the anesthetic provider was masked to BIS values unless there was a clinical need.

Delirium outcome assessors were masked to the intervention. Postoperative data were abstracted from the electronic medical record by staff masked to the intervention. Patients, surgeons, and anesthesiologists were not masked, because it is impossible for the anesthetic technique to be masked to treating physicians or patients. Statisticians and investigators involved in data analysis were masked.

Perioperative care was based on established clinical protocols. Patients could receive intrathecal morphine during spinal anesthesia at the discretion of the anesthesiologist, or by direct intraoperative injection at the discretion of the surgeon. Postoperative analgesia was with fentanyl or hydromorphone patient-controlled analgesia, with transition to oxycodone or other oral opioids as tolerated.

Delirium was assessed once daily during the first 3 postoperative days in the hospital using the validated Confusion Assessment Method²³  (sensitivity, 94 to 100%; specificity, 90 to 95%). For purposes of missing data, daily in-hospital assessments were not considered missing if the patient was discharged from the hospital on that day and not available for assessment. The Confusion Assessment Method included formal tests of cognition (Mini-Mental State Examination,²⁴  Calendar Reverse Months, Shortened Digit Span Forward/Reverse, and Delayed Word Recall tests) as well as questions for nurses, clinicians, and family. Patients who refused an assessment and no delirium assessment could be made were considered to not have delirium for that assessment. The primary outcome was incident delirium as defined by any positive assessment during hospitalization. A chart review for delirium was also conducted using validated methods to supplement in-person assessments.²⁵  Secondary outcomes included delirium duration and severity (Delirium Rating Scale-Revised 98).²⁶  Covariate information was collected from baseline assessments, patient report, and the medical record. Instrumental activities of daily living were measured at baseline.²⁷  Number of surgical levels included the range of involved vertebrae.

At the start of the trial, we assumed a delirium incidence of 40% in the control group (general anesthesia with masked BIS values) and a 50% reduction in the intervention group, based on previous studies.^15,18  Further, we assumed a 4 to 6% dropout or crossover. With these assumptions, 190 patients would be needed to show a difference in incidence of delirium at a 0.05 significance level with a power of 0.8. After the first year of data collection, the delirium incidence was noted to be less than predicted, and so the sample size was increased to at least 218, based on a revised assumption of delirium incidence (40% to 35% in the control arm) and similar assumptions regarding 50% reduction in delirium in the intervention group and 4 to 6% dropout.

The primary analysis was based on the intention to treat principle (patients included in the group to which randomized). For the primary outcome, incident delirium, both the absolute difference and relative change were computed. The chi-square test was used to compare proportions with the primary outcome between groups. Secondary outcomes were compared using Wilcoxon rank sum tests. Normally distributed variables are reported using mean ± SD, and nonnormally distributed variables are reported using mean and interquartile range. Adjusted analyses were conducted with multivariable logistic regression to account for potential confounding, first with prespecified adjustment for age, education, and cognitive score²⁸  and second with adjustment for additional variables associated with delirium in bivariate analyses. As-treated analyses were also conducted (patients included in the group to which they received treatment). Standard diagnostics, including goodness of fit, influence, and collinearity, were examined for all regression models. BIS data were downloaded from the monitor after surgery and were analyzed in several ways, including the mean ± SD and minutes below or above clinically relevant cutoffs (BIS less than 40 and BIS greater than 55), based on the methodology of previous studies.^11,16 

Prespecified subgroup analyses were conducted based on stratification by age (less than 75 vs. 75 yr old or greater), Charlson Comorbidity Index (0 vs. 1 or greater), and baseline cognition (Mini-Mental State Examination less than 27 vs. 27 or greater), with cutoffs chosen based on biologic relevance and/or to have anticipated sufficient number of patients in the subgroups.^16,29,30 Post hoc, we examined four subgroups identified based on differences in bivariate analyses. Relative risks were calculated within each subgroup, and 95% CIs were generated using the percentile method via a bootstrap procedure (5,000 bootstrap samples). The hypothesis that the intervention would have differential effect based on subgroups was formally tested using a P value for interaction, without adjustment for other covariates. SAS v9.4 (USA) was used. Formal interim analyses were to assess recruitment, safety events, and dropout, but not efficacy, and a Data Safety and Monitoring Board monitored study conduct and safety. There were no prespecified stopping criteria, and enrollment ceased when the target sample size was obtained. In all analyses, P < 0.05 was considered significant, and all hypothesis testing was two-tailed.

A patient flow diagram is shown in figure 1. Of 799 patients screened from September 8, 2015, to May 6, 2019, 111 patients were randomized to spinal anesthesia with targeted sedation based on BIS values, and 108 patients were randomized to general anesthesia with masked BIS values. Reasons that patients were not enrolled and randomized are listed in figure 1. Enrollment was stopped upon accrual of enrollment goals. Among patients randomized to spinal anesthesia with targeted sedation based on BIS values, an adequate level of spinal anesthesia could not be obtained in seven patients, and these patients crossed over to receive general anesthesia. Among patients randomized to general anesthesia with masked BIS values, two patients withdrew after randomization, and one patient crossed over to receive spinal anesthesia.

The median age of patients in this study was 72 yr (interquartile range, 69 to 77), 38% were male, and the median Mini-Mental State Examination score was 29 (interquartile range, 27 to 29). Patients rated their average preoperative pain as a median of 7 (interquartile range, 5 to 8) and their current pain as a median of 3 (interquartile range, 1 to 6). Patient characteristics were generally similar in the two arms of the study (table 1). However, the Charlson Comorbidity Index was slightly higher and there were more patients with a previous myocardial infarction and atrial fibrillation in the spinal anesthesia with targeted sedation based on BIS values group.

Intra- and postoperative characteristics are described in table 2 (intention to treat) and Supplemental Digital Content table 1 (https://links.lww.com/ALN/C700; as treated). Overall, the median length of surgery was 128 min (interquartile range, 106 to 159), the median number of spinal levels was 3 (interquartile range, 2 to 4), and the median estimated blood loss was 300 ml (interquartile range, 200 to 460). In the spinal anesthesia with targeted sedation based on BIS values group, the median dose of bupivacaine was 14 mg (interquartile range, 12.5 to 15), and the maximum propofol infusion rate was a median of 80 mcg · kg^–1 · min^–1 (interquartile range, 75 to 100). Among patients who received general anesthesia, desflurane was predominantly utilized. Patients in the general anesthesia with masked BIS values group received more fentanyl and less IV fluids.

The average BIS value in the spinal anesthesia with targeted sedation based on BIS values group was higher than in the general anesthesia with masked BIS values group (median of 62 [interquartile range, 53 to 70] vs. 45 [interquartile range, 41 to 50]; P < 0.001). The median duration of BIS less than 40 was substantially lower in the spinal anesthesia with targeted sedation based on BIS values group compared to the general anesthesia with masked BIS values group (3 min [interquartile range, 0 to 22] vs. 68 min [interquartile range, 22 to 102]; P < 0.001).

The overall incidence of delirium was 22% (48 of 217). Out of 544 opportunities for delirium assessments for nondischarged patients, 509 in-person assessments were completed, and 24 assessments were refused by patients. Two patients refused all assessments. In the intention to treat analysis, there was no significant difference in the incidence of delirium in the spinal anesthesia with targeted sedation based on BIS values group (25.2% [28 of 111]) compared with the general anesthesia with masked BIS values group (18.9% [20 of 106]; P = 0.259), absolute difference, 6.4% (95% CI, –4.6 to 17.4%), and relative risk, 1.22 (95% CI, 0.85 to 1.76). When a chart review delirium method was used to supplement the in-person assessments, there was no significant difference in the incidence of delirium in the spinal anesthesia with targeted sedation based on BIS values group (27.9% [31 of 111]) compared with the general anesthesia with masked BIS values group (23.6% [25 of 106]; P = 0.465). Similarly, there was no difference by group in the incidence of delirium for each individual postoperative day or in maximum delirium severity score (table 3 [intention to treat]; Supplemental Digital Content table 2, [https://links.lww.com/ALN/C700; as treated]). The incidence of delirium was also not different between groups when adjusted for variables associated with delirium in bivariate analyses (Supplemental Digital Content table 3, https://links.lww.com/ALN/C700).

Duration of recovery in the postanesthesia care unit was similar between the two groups, but pain at postanesthesia care unit discharge was lower in the spinal anesthesia with targeted sedation based on BIS values group compared with the general anesthesia with masked BIS values group (median, 4 [interquartile range, 1 to 5] vs. median, 5 [interquartile range, 3 to 7]; P = 0.004). There were two strokes in the spinal anesthesia with targeted sedation based on BIS values group, and there was one death in the general anesthesia with masked BIS values group. Other complications by randomization group are listed in table 2 (intention to treat) and Supplemental Digital Content table 1 (https://links.lww.com/ALN/C700; as treated).

There were three prespecified subgroup analyses, based on cutoffs of the Mini-Mental State Examination, the Charlson Comorbidity Index, and age, with forest plot results by the primary intention to treat analysis shown in figure 2. (The forest plot for the as treated analysis, as well as an expanded description of the numbers of events in each subgroup, are shown in Supplemental Digital Content figure 1 and Supplemental Digital Content table 4, respectively, https://links.lww.com/ALN/C700). Baseline Mini-Mental State Examination did moderate the effect of the intervention (P interaction = 0.009). Specifically, for patients with Mini-Mental State Examination less than 27, the incidence of delirium was less in the spinal anesthesia with targeted sedation based on BIS values group compared to the general anesthesia with masked BIS values group (17.7% [3 of 17] vs. 43.5% [10 of 23]). On the other hand, for patients with Mini-Mental State Examination 27 or greater, the incidence of delirium was greater in the spinal anesthesia with targeted sedation based on BIS values group versus the general anesthesia with masked BIS values group (26.6% [25 of 94] vs. 12.1% [10 of 83]). There was no difference in the effect of the intervention (i.e., no interaction) based on the other prespecified subgroups of age strata (less than 75 vs. 75 yr old or greater) or Charlson Comorbidity Index (0 vs. 1 or greater). Several other subgroup analyses were chosen post hoc (sex, education, use of short-acting opioids at baseline, and administration of intrathecal morphine during surgery; fig. 2). Intrathecal morphine did modify the effect of the intervention in the intention to treat analysis (P interaction = 0.029) but not in the as treated analysis (P interaction = 0.088). Specifically, for patients who did not receive intrathecal morphine, the incidence of delirium in the intention to treat analysis was less in the spinal anesthesia with targeted sedation based on BIS values group compared to the general anesthesia with masked BIS values group (8.8% [3 of 34] vs. 20.4% [10 of 49]). On the other hand, for patients who did receive intrathecal morphine, the incidence of delirium was greater in the spinal anesthesia with targeted sedation based on BIS values group versus the general anesthesia with masked BIS values group (32.5% [25 of 77] vs. 17.5% [10 of 57]).

In bivariate analyses, male sex, lower Mini-Mental State Examination score, higher Charlson Comorbidity Index, preoperative short-acting opioid medication, longer surgery, and increased postoperative pain were among the variables associated with delirium (Supplemental Digital Content tables 5 and 6, https://links.lww.com/ALN/C700). In adjusted models (Supplemental Digital Content table 3, https://links.lww.com/ALN/C700), only lower Mini-Mental State Examination score remained independently associated with delirium. The administration of intrathecal morphine was also associated with delirium in the adjusted model, but not in the bivariate comparison.

The results of this trial demonstrate that spinal anesthesia with targeted sedation based on BIS values compared with general anesthesia with masked BIS values does not reduce the incidence of delirium in lumbar spine surgery patients.

The results of this study add to several studies examining whether titrating depth of anesthesia and anesthetic exposure compared with usual care can reduce delirium. Early trials in general anesthesia patients suggested that a strategy to reduce anesthetic exposure based on BIS values could reduce delirium.^11,12  Based on these and other studies, delirium guidelines have recommended depth of anesthesia monitoring may be considered.¹  However, the recent large trial reported no difference in delirium in patients randomized to a strategy of avoiding excessive anesthetic exposure and burst suppression on the electroencephalogram.¹⁴  Similarly, the results of the current study demonstrate that a bundled approach to reduce anesthetic agent exposure as determined by BIS values does not reduce the incidence of delirium in older adults undergoing lumbar spine fusion surgery.

An important consideration in interpreting previous studies is that in most trials, all patients received general anesthesia. The pertinent comparisons were general anesthesia versus deeper general anesthesia, and the benefits of lighter anesthesia could not be examined. This is an important gap since critical care guidelines recommend that mechanically ventilated patients in the intensive care unit benefit from light sedation,³¹  a level of consciousness that is substantially more alert than general anesthesia. Two trials in hip fracture surgery patients under spinal anesthesia examined benefits of intraoperative “light” sedation.^15,16  However, the results of these two studies were conflicting, and moreover, the elderly, frail, and cognitively impaired populations may not be generalizable to most older adults undergoing surgery. Thus, there has been a clear need to determine whether reducing depth of anesthesia beyond general anesthesia could reduce delirium in a generalizable population of older adults. This question is highly relevant since many surgeries can be performed with neuraxial or regional approaches. The SHARP study addressed this question in a pragmatic manner and demonstrated no delirium reduction in patients treated with spinal anesthesia with targeted sedation based on BIS values compared with general anesthesia with masked BIS values.

One of three preplanned subgroup analyses showed different effects of the intervention according to baseline cognition. Specifically, for patients with Mini-Mental State Examination less than 27, there was less delirium in the spinal anesthesia with targeted sedation based on BIS values group, while for patients with Mini-Mental State Examination 27 or greater, there was less delirium in the general anesthesia with masked BIS values group. The results of this subgroup analysis are qualitatively similar to a subgroup analysis reported in a trial of depth of sedation in hip fracture surgery patients.¹⁶  In this trial in which the median Mini-Mental State Examination score was 24, the subgroup of healthy patients with a Charlson Comorbidity Index score of 0 to 1 (but not higher) had less delirium with a light versus deep sedation strategy. Thus, in both the hip fracture trial and the current trial, patients who were relatively healthy but with impaired cognition derived benefit from lighter sedation. These results need to be considered hypothesis-generating since they were subgroup analyses. One potential explanation is that cognitively impaired patients are more sensitive to anesthetic depth, perhaps due to underlying neurodegenerative disease.^32–35  On the other hand, it is not entirely clear why cognitively intact patients benefited from general anesthesia with masked BIS values. The overall risk of delirium was less in these patients, as would be expected. Future studies should examine anesthetic strategies to reduce depth of anesthesia in cognitively impaired older adults, although the logistics of enrolling a sufficient number of eligible patients would be challenging. A post hoc analysis also showed that the administration of intrathecal morphine was independently associated with delirium and modified the effects of the intervention such that in patients who received intrathecal morphine, there was less delirium in the general anesthesia with masked BIS values group. Previous work has suggested that intrathecal morphine was associated with less postoperative delirium,³⁶  while in our study, patients who received intrathecal morphine had more delirium, and the finding of this post hoc analysis should also be considered exploratory.

In the current study, the strongest and most consistent delirium risk factor was lower Mini-Mental State Examination score. These results are consistent with other studies examining risk factors for delirium³  and highlight the importance of cognitive testing for risk stratification. Overall, pain and pain treatment were important, with baseline short-acting opioids and maximum postoperative pain being associated with delirium. These results highlight the balance of treating pain while minimizing deliriogenic opioid medication.^3,37 

There are several strengths of this study. The SHARP trial used a unique study design to compare spinal anesthesia with targeted sedation based on BIS values versus general anesthesia with masked BIS values in cognitively intact older adults. The intervention was pragmatic, conducted at a community-based hospital, and achieved a separation in BIS values. The research group is experienced in assessing postoperative delirium. Although the study sample was older adults undergoing spine surgery, results are likely generalizable to a number of surgeries for which general or neuraxial/regional anesthesia is appropriate.

There are several limitations. The intervention was bundled, and it is unclear which aspect (light sedation, spinal anesthesia, or propofol) was most responsible for the subgroup effect. The doses of propofol that were used were relatively high, the sedation protocol was pragmatic, and a formal observer assessment of sedation was not used. Thus, a number of patients in the spinal anesthesia with targeted sedation based on BIS values group had BIS values below the target of 60 to 70, and this may have biased the results toward the null. Additionally, BIS may not be an accurate measure of depth of anesthesia in older adults. However, the majority of patients had BIS values that exceeded the upper limit of 55 that has been advocated to prevent awareness during general anesthesia.^38,39  The bundled approach also did not permit the use of other sedative agents, such as dexmedetomidine, and future studies are needed to examine potentially beneficial effects of dexmedetomidine in this population. The study was powered for a large effect size, based on a previous study,¹⁵  and we revised the estimate of delirium incidence due to a lower incidence than originally expected. However, the overall incidence of delirium was still below the expected incidence in the power calculation, and so the study was underpowered. Nevertheless, given the observed effect, it is unlikely that a larger study would demonstrate a benefit in the intervention group. We assessed delirium only once daily, and some cases may have been missed. Thus, imprecision of the outcome assessment and/or misclassification may have biased the results. Patients in the spinal anesthesia with targeted sedation based on BIS values had more cardiac and vascular disease at baseline, although the baseline Mini-Mental State Examination was slightly higher than the general anesthesia with masked BIS values group. Perioperative management aside from the intervention was based on established protocols, and this introduced heterogeneity into the study. There was crossover between study arms in eight patients, largely due to obtaining adequate spinal anesthesia in patients with degenerative spine disease, and this is a source of bias. However, results were similar in intention to treat and as treated analyses. The Mini-Mental State Examination is a general screen of cognition and is limited by ceiling effect and educational biases.⁴⁰  Further, the distinction between a Mini-Mental State Examination score above and less than 27 may not be clinically meaningful, and so the results of the subgroup analyses should be considered hypothesis-generating. Finally, the trial was not formally registered in ClinicalTrials.gov until 2017 due to quality control issues, although the initial protocol with the aim and primary outcome of this manuscript was submitted in October 2015.

In conclusion, the results of the SHARP study demonstrate that spinal anesthesia with targeted sedation based on BIS values does not reduce delirium in older adults undergoing lumbar spine surgery. Further studies are needed to examine optimal anesthetic strategies in cognitively impaired patients, who are at high risk for delirium.

The study was funded by the National Institutes of Health (Bethesda, Maryland; NIH K76 AG057020 and NIH K23 AG051783; to Dr. Brown) and by the International Anesthesia Research Society (San Francisco, California; to Dr. Brown).

Dr. Brown has consulted for and received grant funding from Medtronic Inc. (Minneapolis, Minnesota). Dr. Neufeld has received grant funding from Hitachi Inc. (Tokyo, Japan) and consulted for Merck Inc. (Kenilworth, New Jersey). Dr. Hogue has received payment for advisory board membership from Medtronic Inc. and Edwards Lifesciences (Irvine, California). He serves on a data safety monitoring committee for Merck Inc. Dr. Cha has consulted for Avania LLC (Marlborough, Massachusetts) and MC3 Corp (Dexter, Michigan). The other authors declare no competing interests.

Full protocol available at: cbrownv@jhmi.edu. Raw data may be available with the appropriate institutional agreements at: cbrownv@jhmi.edu.