Authors: Michael P. Schnetz (1Anesthesiology and Perioperative Medicine, School of Medicine, University of Pittsburgh; Pittsburgh, PA, USA), Brian J. Reon (1Anesthesiology and Perioperative Medicine, School of Medicine, University of Pittsburgh; Pittsburgh, PA, USA), James W. Ibinson (1Anesthesiology and Perioperative Medicine, School of Medicine, University of Pittsburgh; Pittsburgh, PA, USA; 2Clinical and Translational Science Institute, University of Pittsburgh; Pittsburgh, PA, USA), Murat Kaynar (1Anesthesiology and Perioperative Medicine, School of Medicine, University of Pittsburgh; Pittsburgh, PA, USA; 3Critical Care Medicine, School of Medicine, University of Pittsburgh; Pittsburgh, PA, USA), Aman Mahajan (1Anesthesiology and Perioperative Medicine, School of Medicine, University of Pittsburgh; Pittsburgh, PA, USA; 4Biomedical Informatics, School of Medicine, University of Pittsburgh; Pittsburgh, PA, USA; 5Bioengineering, Swanson School of Engineering, University of Pittsburgh; Pittsburgh, PA, USA), Keith M. Vogt (1Anesthesiology and Perioperative Medicine, School of Medicine, University of Pittsburgh; Pittsburgh, PA, USA; 2Clinical and Translational Science Institute, University of Pittsburgh; Pittsburgh, PA, USA; 5Bioengineering, Swanson School of Engineering, University of Pittsburgh; Pittsburgh, PA, USA; 6Center for the Neural Basis of Cognition, University of Pittsburgh; Pittsburgh, PA, USA)
Categories: Article
Source: Anesthesia and analgesia
Authors: Michael P. Schnetz, Brian J. Reon, James W. Ibinson, Murat Kaynar, Aman Mahajan, Keith M. Vogt
Though patients are commonly monitored for depth of anesthesia, it is unclear to what extent administration of intravenous anesthetic medications may affect calculated Bispectral Index^™^ (BIS index) values under general anesthesia.
In a retrospective analysis of electronic anesthesia records from an academic medical center, we examined BIS index changes associated with 14 different intravenous medications, as administered in routine practice, during volatile-based anesthesia using a novel screening approach. Discrete time windows were identified in which only a single drug bolus was administered, and subsequent changes in the BIS index, concentration of volatile anesthetic, and arterial pressure were analyzed. Our primary outcome was change in BIS index, following drug administration. Adjusted 95% confidence intervals were compared to pre-determined thresholds for clinical significance. Secondary sensitivity analyses examined the same outcomes, with available data separated according to differences in baseline volatile anesthetic concentrations, doses of the administered medications, and length of time window.
The study cohort was comprised of data from 20,170 distinct cases, 54.7% of patients were male, with a median age of 55. In the primary analysis, ketamine at a median dose of 20 mg was associated with a median (confidence limits) increase in BIS index of 3.8 (2.5 – 5.0). Midazolam (median dose 2 mg) was associated with a median decrease in BIS index of 3.0 (1.5 – 4.5). Neither of these drug administrations occurred during time periods associated with changes in volatile anesthetic concentration. Analysis for dexmedetomidine was confounded by concomitant decreases in volatile anesthetic concentration. No other medication analyzed, including propofol and common opioids, was associated with a significant change in BIS index. Secondary analyses revealed that similar BIS index changes occurred when midazolam and ketamine were administered at different volatile anesthetic concentrations and different doses, and these changes persisted 11–20 minutes post-administration.
Modest, but persistent changes in BIS index occurred following doses of ketamine (increase) and midazolam (decrease) during periods of stable volatile anesthetic administration.
Patients under anesthesia are being increasingly monitored using the BIS monitor (Medtronic – Covidien division, Minneapolis, MN, USA), a well-known device for guiding anesthesia administration. BIS technology calculates the Bispectral Index^™^, a dimensionless scaled index, referred to as the “BIS index” throughout the manuscript. The BIS index was initially validated with propofol,^1,2^ but, volatile anesthetics have shown similar effects on EEG spectral density.^3^ As these GABAergic agents are almost universally employed for anesthetic maintenance, monitoring with the BIS index is commonly thought to be robust during most anesthetics. However, changes in BIS index associated with other commonly administered drugs are less well quantified, particularly outside controlled experiments. Understanding how medications may influence the BIS index is important, as significant perturbations could influence decision-making. If these induced BIS index changes are discordant with actual depth of anesthesia, an inappropriate dose of anesthesia could be delivered unintentionally. Thus, quantification of BIS index changes following the administration of common intravenous (IV) drugs, as used in routine anesthesia practice, is of importance for optimizing delivery of personalized anesthesia care.
Here, we introduce a novel screening procedure, as an alternative to larger and more costly prospective controlled studies across many anesthetic agents. Our primary aim was to find isolated bolus medication administrations that were associated with meaningful BIS index changes during volatile anesthesia, and quantify the changes seen in routine practice. Changes in BIS index were assessed concurrently with inhaled anesthetic concentration and blood pressure while other drugs are not actively initiated or discontinued. These findings may inform clinicians to better contextualize BIS index changes while also motivating further study of these relationships in future prospective investigations.
All relevant portions of the STROBE guidelines for reporting observational studies were followed.^4^ The Biomedical Institutional Review Board of the University of Pittsburgh approved this retrospective study as minimal risk and waived the informed consent requirement (PRO15060130, approved 9/24/2015).
Patients aged 18 years or older undergoing noncardiac surgery with general anesthesia at University of Pittsburgh Medical Center Presbyterian and Montefiore hospitals between June 2011 and March 2018 were eligible for study. Surgeries were included in analysis if 1) BIS monitoring was used, and 2) intravenous propofol, midazolam, ketamine, dexmedetomidine, fentanyl, hydromorphone, phenylephrine, ephedrine, glycopyrrolate, labetalol, metoprolol, esmolol, rocuronium, or calcium chloride were administered at least once in isolation from other medications and fluids. These medications were selected because they are common intraoperative agents with potential to influence the BIS index, alter hemodynamics, or a combination of these. BIS monitoring was ubiquitously available in all anesthetizing locations that may employ volatile anesthesia and often employed as part of the intraoperative monitoring strategy for general anesthetics. However, the decision to use the BIS monitor was made on an individual basis, making case selection pragmatic. Patients not receiving a volatile-based anesthetic or having cardiac surgery were excluded.
The medical record number, patient and procedure characteristics, intraoperative medications, fluids, and measurements of BIS index, mean arterial blood pressure (MAP), and end-tidal inhaled anesthetic concentrations recorded at a maximum frequency of one measurement per minute were extracted for each study surgery from the electronic health record. Notably, the time of medication administration was manually entered by providers, by clicking on a graphical timeline computer interface that is part of our electronic anesthesia record. Minimum alveolar concentration (MAC) was calculated using standard isoflurane = 1.17%, sevoflurane = 1.8%, and desflurane = 6.6%.^5^ Nitrous oxide was not used in any of the cases. MAC values were not age-adjusted in the primary analysis, because they were not compared between patients. Instead changes in MAC within single patients, over a short time interval, were calculated. Missing data and those likely to be artifacts were removed prior to analysis. Both non-invasive and invasive MAP measurements were used, if both were available at the same time, the invasive measurement was selected. Non-invasive blood pressures would be windowed by the measurement interval used; the default value on our monitors is 2.5 minutes. Values <10 and >250 mmHg were considered extreme measurements and excluded as artifacts; these represented <0.01% of total MAPs.
The primary outcome was change in BIS index following medication administration for isolated, bolus doses of 14 drugs, which were the exposure variables of propofol, midazolam, ketamine, dexmedetomidine, fentanyl, hydromorphone, phenylephrine, ephedrine, glycopyrrolate, labetalol, metoprolol, esmolol, rocuronium, and calcium chloride. For each dose, the difference between two median BIS index values was calculated, one for measurements captured in the ten minutes before IV medication administration and another for the ten minutes after. Note that the medication administrations identified by this screening method come from different patients and surgeries, but the primary comparison is between before and after a drug dose on a within-patient basis. Changes in MAC and MAP were similarly calculated within the windows surrounding drug administration.
Isolated doses of each study medication were identified in the dataset as instances where no other medication or fluid (including blood products) was initiated or discontinued in the ten minutes before and after the dose. Only drug doses with surrounding BIS index, MAC, and MAP data required to calculate variable changes were included. The agents selected are some of the most common therapies provided during noncardiac surgery, however, isolated doses are far less frequent. To maximize the number available, administrations concurrent with IV infusions that remained unchanged throughout the twenty-minute window were included in the analysis. In a minority of cases, more than one qualifying medication administrations were found in the same surgery (thus in the same patient), and these were removed from the primary analysis to reduce confounding.
The R statistical software (R Foundation for Statistical Computing, version 3.5.3, Austria) was used for all described analyses. Data was plotted using the ggplot2 package. As a Bonferroni adjustment (accounting for 14 drugs investigated) 99.7% two-sided confidence intervals were calculated for the primary analysis, using the CI function in the Rmisc package. All other confidence intervals were similarly adjusted for the number of comparisons being made (e.g. 10 groups being compared in the analysis in Figure 3, so 99.5% confidence limits were calculated, for p < 0.05/10). The confidence limits used in each analysis are described in the figure legends. Descriptive statistics were generated for patient and procedure characteristics. First, second (median), and third quartiles were calculated for continuous variables patient age, BIS index, MAC, MAP, and dose amount. Proportions were calculated for categorical ASA physical status, emergency surgery, gender, and surgical specialty. The mean change in BIS index, MAC, and MAP following medication administration was calculated, with a confidence interval, for all identified doses of each medication. To screen for clinically meaningful changes, conservative thresholds were operationally defined and BIS index increase or decrease of >1, MAC increase or decrease of >0.05, and MAP increase or decrease of >1 mmHg. When confidence limits of the parameter change were larger than these thresholds, a clinically meaningful difference was determined to have occurred. To assess the potential for data skewing to affect the interpretation of the results, the primary analysis was also repeated calculating mean variable values, instead of median, before and after drug doses.
Findings from our primary analysis were tested across several sensitivity analyses. We suspected that BIS index changes resulting from a medication may be less likely to occur during deeper periods of volatile anesthesia. Thus, BIS index, MAC, and MAP changes were compared between drug administrations at 0.9 MAC or greater, versus those at less than 0.9 MAC. MAC values in this analysis were age-adjusted^6^ to allow comparison between anesthetic depths. The 0.9 MAC threshold was selected to generate a sufficient number of dose windows in both groups, while distinguishing those given at anesthetic depth definitely consistent with general anesthesia from doses provided at lighter depths. Dose amount, duration of BIS change, and influence of co-administered (or discontinued) medications and fluids were evaluated for ketamine and midazolam, the agents with the most pronounced BIS index changes. Larger doses of these drugs were hypothesized to result in larger BIS index changes compared to smaller ones. We considered whether these drug-associated changes may persist beyond ten minutes. Thus, the duration of BIS index change was also evaluated with a twenty-minute post-administration interval. Like the primary analysis, the intervals before and after the dose were free from administrations of other medications or fluids and contained requisite BIS index, MAC, MAP data. We also performed a sensitivity analysis for other drug-associated changes, in which this isolated administration constraint was removed. Windows were identified and analyzed in which at least one medication or fluid was co-administered (or an infusion was initiated or discontinued), within −10 and +10 minutes of the documented time the primary drug was administered. Sensitivity analyses required subsetting doses from the overall pool identified, and the resulting sample sizes are reported for each. Phenylephrine, ephedrine and rocuronium served as reference agents (for which no BIS change would be expected). Finally, we performed an analysis where a pair of administrations, one of rocuronium, the other from another study drug, were identified for the same patient, within the same surgery. Variable changes for each were calculated as in the primary analysis. This paired analysis allows direct comparison between differences seen with rocuronium and other study drugs that may be associated with BIS, or MAP changes.
The study cohort meeting inclusion criteria was compromised of 20,170 distinct patient surgical cases, 54.7% of patients were male, with a median age of 55, most frequently assigned an ASA physical status of 3. General, non-emergency surgery was most common. The study population demographic and surgical characteristics, as well as overall median BIS index, MAC and MAP values appear in Table 1. Of nearly 1 million study doses screened, 35,498 isolated administrations with requisite BIS index, MAP, and MAC data were identified. These medications, overall, represented 73.2% of all drug boluses administered in the study population (Table 2). The fraction of doses that were isolated was less than 10% across medications, however each generated more than 100 isolated doses for analysis. Glycopyrrolate yielded the fewest (112), and rocuronium the most (15,422). The right-most column of Table 2 documents the number of isolated doses in unique patient surgical cases. Rocuronium had the most repeated doses (33%), while glycopyrrolate had none. Unique doses were those used in our primary analysis.
Figure 1 shows BIS index, MAC, and MAP changes following medication administration in the primary analysis. Midazolam at a median dose of 2 mg (first and third quartiles of distribution = 1 and 2 mg) was associated with decrease in median BIS index of 3.0 (CI= 1.5 – 4.5), while ketamine at a median dose of 20 mg (first and third quartiles = 10 and 30 mg) was associated with a median increase of 3.8 (CI= 2.5 – 5.0), both without concomitant changes in MAC. Changes associated with fentanyl (50 mcg median dose, first and third quartiles = 50 and 100 mcg), hydromorphone (0.5 mg median dose, first and third quartiles = 0.4 and 1.0 mg), and propofol (30 mg median dose, first and third quartiles = 30 and 50 mg) were not outside of pre-determined, clinically meaningful boundaries for any variable (see Methods section). BIS index increased while MAC decreased following dexmedetomidine (12 mcg median dose), confounding the association of administrations of this drug with BIS index changes. Besides dexmedetomidine, no other drug was associated with MAC changes greater than ±0.05 in the analysis window. Anticipated MAP changes occurred with vasoactive medications without changes in BIS index. Median MAP increased 2.1, 3.7, and 2.0 mmHg following phenylephrine, ephedrine, and calcium chloride, respectively, and decreased 6.2 mmHg after labetalol. Esmolol and metoprolol were not associated with changes. MAP increased after glycopyrrolate, but failed to exceed clinically meaningful difference boundaries. No changes in BIS index, MAC, or MAP were observed following rocuronium administrations. The number of isolated administrations for each medication, as well as pre-administration BIS index, MAC and MAP values are listed in Supplemental Table 1. The median BIS value at which the various medications were given ranged from 38 to 47, with a volatile anesthetic dose between 0.7 and 1.0 MAC, and a MAP between 72 and 102 mmHg. Supplementary Figure 1 shows results for the primary performed using mean variable values (which may be more sensitive to extremes) in the pre-administration and post-administration windows; this yields similar results.
Figure 2 compares BIS index, MAC, and MAP changes between medications administered at less than 0.9 MAC (top) or 0.9 and greater (bottom). Findings are consistent with primary analysis and between MAC levels for midazolam, ketamine, phenylephrine and ephedrine. MAP did not change with labetalol given at lower MAC, a finding unrelated to changing MAC (Figure 2, middle panels) or dose quantity (Supplemental Table 2). Larger variable changes occurred when midazolam (median dose 1 vs. 2mg), ketamine (10 vs. 30 mg), phenylephrine (80 vs. 160 mcg), and ephedrine (5 vs. 10 mg) were administered at higher doses (Figure 3, Supplemental Table 3), although only the BIS index difference (2.7 vs. 6.6) between ketamine groups was statistically significant. Variable changes that occurred with these medications in the 10-minute post-administration interval persisted up to 20 minutes (Figure 4, Supplemental Table 4). Rocuronium was not associated with BIS index, MAC, and MAP changes that exceeded clinical boundaries in any of the sensitivity analyses performed (Figures 2, 3, 4).
Supplemental Figure 2 and Supplement Table 5 show the results of the analysis in which at least one medication or fluid was co-administered, or discontinued, within the ± 10-minute window surrounding the time of documented dose of the primary drug. Note that with the constraint of isolated administration removed, many more drug administration windows were available for analysis. Similar to the findings in the primary analysis, ketamine was associated with an increase and midazolam was associated with a decrease in BIS index.
Supplementary Figure 3 and Supplementary Table 6 shows the results for the paired analysis between rocuronium and the other primary drugs within the same surgeries. Observed BIS index decreases following midazolam and increases following ketamine are similar to the primary analysis. Importantly, no significant differences in BIS index or MAP were seen following matched rocuronium doses for any of the other drugs of interest.
Opioids,^7^ benzodiazepines,^8^ and alpha-2 agonists,^9^ common to a balanced anesthetic, have well-known MAC-sparing effects, while other drugs have been reported to increase MAC (ephedrine^10^). It is recognized that some of these effects may be independent of hypnosis (e.g. acting in the spinal cord), and medications can also impact the BIS index independent of anesthetic state.^11,12^ Correct interpretation of BIS index values requires clinicians to recognize such drug-mediated effects, but quantifying them across all conditions reflected in real-world practice may be impractical with prospective studies. We screened more than 950,000 drug administrations to find associated BIS index changes. We also included an analysis of vasoactive agents whose potential effect on BIS index has not been previously evaluated, despite being used clinically to maintain perfusion to the brain.
We present several important findings. First, our screen yielded more than 35,000 isolated drug boluses for analysis, and more than 100 for each study medication. This supports the feasibility of our approach, including all the constraints, when leveraged against large datasets. Second, midazolam and ketamine were associated with modest BIS index changes, less than 5 on average. MAP changed less than 7 mmHg, on average, following phenylephrine, ephedrine, labetalol, and calcium chloride administrations. Third, BIS index and MAP changes observed with midazolam, ketamine, phenylephrine, and ephedrine were consistent across administrations given at low and high concentrations of volatile anesthetic, as well as low and high medication dose. These changes persisted up to 20 minutes for each medication.
Propofol is well-known to decrease BIS index values, yet not observed in our findings. Propofol procedural sedation can have a wide range of BIS index values^14^, and effects have been mostly studied using high dose infusions for general anesthesia.^15,16^ Our methodology excluded induction doses of propofol, as these doses are nearly always co-administered with other medications, thus included smaller (median 30 mg) boluses administered mid-case. In our secondary analysis, propofol was associated with increasing BIS during lighter planes of anesthesia compared to deeper planes, where BIS decreased after administration, suggesting clinicians used these doses to supplement a patient’s anesthetic reflected by increasing BIS. Neither change was statistically significant, but lends possible explanation to our findings particular to propofol within the framework of our approach.
Ketamine’s effect on the BIS index has been studied under a variety of experimental conditions, notably all outside the context of surgery. Administering ketamine without any other sedative-hypnotic agents seems to have minimal impact on the BIS index.^17^ After induction with propofol and titration of the BIS index to 40, a bolus dose of ketamine 0.5 mg/kg caused an increase of the BIS index to 63.5 in one study.^18^ Interestingly, in the same study, a bolus dose of 0.2 mg/kg ketamine showed no significant increase in BIS index,^18^ suggesting a dose-dependent effect. On a background of 2% end-tidal sevoflurane anesthesia, a ketamine bolus dose of 0.5 mg/kg caused an increase in BIS index from 33 to 46.^19^ However, at lower sevoflurane concentrations (0.5–1.5% end-tidal), the same 0.5 mg/kg dose can be associated with a BIS index up to 79.^20^ Assimilating these previous findings, the magnitude of ketamine’s effect on BIS index seems to depend on both the dose administered and the background of other drugs being used for maintenance of anesthesia. Our results fit well within this framework where we demonstrate dose-dependency, with larger increases in BIS index were associated with higher doses of ketamine. The average BIS index change, overall, was 3.8 associated with a median dose of 20 mg, providing reassurance that low-dose boluses are unlikely to completely invalidate the BIS index for guiding overall depth of a volatile-based anesthetic. Similarly, an extremely elevated BIS index after ketamine administration should not be attributed solely to the medication.
Dexmedetomidine administration was associated with decreasing MAC over the same time period. This seems to indicate that anesthesia providers tend to bolus doses of dexmedetomidine at times when volatile anesthetic concentration is also being intentionally decreased. However, any relationship between drug and BIS index change is obscured by concomitantly decreasing volatile agent.
We detected no significant changes in BIS index for the two common opioids investigated, fentanyl and hydromorphone. Opioids have long been known to cause EEG slowing and spectral changes,^21,22^ despite not being considered hypnotic agents. In previous work, administration of high-dose fentanyl (10 μg/kg) after premedication with intramuscular morphine showed widely variable effects on BIS index and behavior, with BIS index values ranging from 43–98 before loss of responsiveness, and values persistently as high as 98 when unresponsive.^23^ When fentanyl was administered to an effect-site concentration of 1.5 ng/ml and propofol then titrated to loss of responsiveness, this change occurred at a BIS index of 69.5 with fentanyl, compared to 60.5 with propofol alone.^24^ This seems to demonstrate an enhanced hypotonic effect of propofol in the context of fentanyl that is not reflected by a lower BIS index. Investigations with other opioids have demonstrated similar effects.^24^ Thus, our data is consistent with the concept that the calculated BIS index is not directly affected by opioid administration, despite the addition to overall anesthetic depth that inclusion of opioids would be expected to provide.
Though the isolated administration of neuromuscular blocking agents has been shown to cause profound decreases in BIS index under unusual conditions,^11^ this effect is not commonly observed in clinical practice. In our study, isolated rocuronium boluses were very commonly found. This provided an important control for our analysis. Across more than 10,000 administrations, rocuronium dosing was not associated with meaningful changes in BIS index, nor MAC or MAP and highlights changes observed for ketamine and midazolam, as a direct comparator, when administered in the same surgeries
The primary strengths of our study include demonstration of a method to screen for BIS index, MAP, and MAC changes across many common medications and over more than thousands of individual administrations. Our findings reflect changes observed in real-world practice, while adjusting for several confounders within the limitations of retrospective data. Changes were tested under several distinct clinical scenarios to verify results, but also assess the impact of potential confounding as selected doses, and their associated patient, procedure characteristics, as well as the precise manner they were administered, varied between analyses. All seven experiments yielded similar results. Finally, this methodology could be modified to study changes associated with medication administration on other recorded monitor values, for example, blood pressure following the treatment of intraoperative hypotension.
Our study is not without limitations, many of which are a result of initial assumptions in designing the analysis. First, we have averaged across different volatile anesthetics, effectively treating them as having identical effects on the EEG. We chose to focus on isolated administrations occurring at the center of an otherwise pristine time window. Though this is a relatively broad timeframe (and thus stringent requirement), this was to account for the inherent imprecisions that may exist between the time of actual dose delivery and the documentation of that dose. An important consequence of our isolation windowing approach is that our primary analysis does not capture all administrations of the investigated drugs, but rather a minority. Identified changes in BIS index, MAC, and MAP could be influenced by drugs administered before the beginning of the window (more than 10 minutes prior), undocumented drug doses, changes in surgical stimulation, differences in patient and procedure factors between selected drug administrations, etc. As it is infeasible to reasonably quantify these effects, they represent an unavoidable confound. However, consistent results across multiple sensitivity analyses, suggest that the magnitude of these confounders is not so large that they negate our general findings.
We introduce a screening method for evaluating BIS index and MAP changes following isolated intraoperative medication administrations. Our data suggest that only midazolam and ketamine were associated with changes in BIS index, at commonly-used doses, during general anesthesia administered for adult, non-cardiac surgery. This supports the utility of the BIS index for monitoring during a volatile-based general anesthetic, even if bolus doses of multimodal anesthetic adjuncts are administered. Further, the magnitude of change should be modest, but may persist for up to 20 minutes.