Secondary Logo

Journal Logo

Anesthetic Pharmacology: Research Report

Associations Between Age and Dosing of Volatile Anesthetics in 2 Academic Hospitals

Van Cleve, William C. MD, MPH; Nair, Bala G. PhD; Rooke, G. Alec MD, PhD

Author Information
doi: 10.1213/ANE.0000000000000819

The roughly linear relationship between increasing age and the “depth” of anesthesia achieved at a given concentration of potent inhaled volatile anesthetic is an accepted and widely taught precept guiding the titration of these drugs in patients of advanced age.1–4 The implications of this relationship for the clinician monitoring a patient’s depth of anesthesia are clear: when depth of anesthesia is defined by the fractional minimum alveolar concentration (MAC) of exhaled volatile anesthetic, MAC fraction should be adjusted for age, and older patients will generally require lower absolute doses to obtain equivalent anesthetic effects.

Knowledge of, and behavior guided by, this relationship will grow in importance as the prevalence of old age increases in the United States over the coming decades.5 Superimposed on a clinical practice driven by an aging population is an increasing appreciation for the potential for inhaled anesthetics to exert effects that persist in the postoperative period. In animal models, inhaled anesthetics have been implicated in brain cell apoptosis, deposition of amyloid protein, and memory impairment.6–9 In elderly humans, emerging, yet controversial, evidence suggests that increased depth of anesthesia may be associated with postoperative cognitive dysfunction, delirium, and increased mortality.10–13

To our knowledge, the question of whether anesthesia providers actively and accurately consider the age–MAC relationship in administering and titrating volatile anesthetics has not been explored. To address this question, we conducted an analysis of anesthetics provided to adult patients at 2 hospitals within our quaternary care academic medical system.

A priori, we hypothesized that anesthetic providers would provide decreased doses of inhaled anesthetics (expressed as mean delivered MAC during maintenance of anesthesia) to older patients but that this decrease would fall short of the adjustment suggested by available scientific evidence (6.7% decrease per increased decade of age). We based this hypothesis on a clinical observation that providers in our institution generally feel it necessary to maintain a MAC fraction no less than 0.7 during maintenance of anesthesia, although our anesthesia workstations do not adjust this calculation for age.


This study was approved by the University of Washington IRB. The requirement for informed consent was waived. As a retrospective review, this study was not registered before data collection.

Clinical Setting

This study was a retrospective cross-sectional review of anesthetic records collected from 2 hospitals in our academic health care system. One hospital (hospital A) serves as a referral center for complex cardiac, transplant, and hematology/oncology services; the other (hospital B) is the only level I adult trauma and burn center for several states. Notably, during the period under study, at hospital A, monitors displayed drug concentration alone but could be configured to display MAC fraction (unadjusted for age), whereas at hospital B, both MAC fraction (unadjusted for age) and drug concentration were displayed by default. At hospital B, monitors were technically capable of displaying age-adjusted MAC (aaMAC) after a long series of configuration steps, but an informal survey of providers demonstrated that this capability was not widely appreciated and virtually never used. Data regarding the configuration of the monitor regarding display of MAC fraction were not recorded in the anesthetic information management system (AIMS).

Data Sources

Both hospitals use Merge AIMS (Merge Healthcare, Hartland, WI). Approximately 100% of anesthetics at each center are recorded in Merge AIMS. Merge AIMS records demographic, clinical, and administrative data for each anesthetic provided. Notably, the AIMS did not record information regarding medications taken in the perioperative period.

Inclusion/Exclusion Criteria

We included all general anesthetics provided to patients aged ≥18 years with a minimum surgical time of 1 hour for which a single, inhaled potent volatile anesthetic (isoflurane, sevoflurane, or desflurane) was administered. An upgrade in recording capabilities at hospital B allowed accurate data acquisition beginning May 1, 2012. Our IRB application for retrospective review of records was submitted in November 2012. As a result, our sample included anesthetics performed at both hospitals during a 7-month period (May to November) of 2012. We then excluded all obstetric, cardiac, thoracic, transplant, and ASA physical status V patients as possibly unrepresentative of usual general anesthetic practices. Finally, we excluded cases in which nitrous oxide or continuously infused IV anesthetics (e.g., propofol, benzodiazepines) were administered.

Clinical Data

The data extracted from Merge AIMS included patient age, gender, ASA physical classification and emergency statuses,14 procedural duration (as recorded in the anesthesia record from “surgery start” to “surgery end”), whether a preinduction dose of midazolam was administered, total dose of fentanyl administered during the first two-thirds of surgical time, whether doses of neuromuscular-blocking drugs were administered after tracheal intubation, whether bispectral index (BIS) monitoring was performed, use of vasopressor agents, which inhaled drug was used, and whether regional or neuraxial analgesia was used as an adjunct to general anesthesia.

Outcome: Mean Delivered MAC Fraction

For the purposes of this investigation, we summarized anesthetic depth in a single value. We did so by dividing the surgical time into 3 equal periods. We defined the middle third of surgical time as “maintenance” of anesthesia and queried Merge AIMS for minute-to-minute end-tidal agent concentrations (ETACs) during this period. These values were first filtered through a local-median algorithm with a 5-minute moving “window” to reduce signal contamination from circuit disconnects, suctioning, and similar noise. Filtered ETAC values were then transformed to MAC using the values provided by the U.S. Food and Drug Administration-approved product inserts included with each drug (each manufacturer provides MAC at several “reference ages,” with the MAC closest to age 40 years traditionally cited as that drug’s singular MAC: MAC sevoflurane = 2.1%, MAC desflurane = 6.0%, MAC isoflurane = 1.15%, at ages 40, 45, and 44 years, respectively).15–17 After this transformation, the mean-filtered delivered MAC fraction (hereafter referred to as “mean delivered MAC fraction”) during maintenance of anesthesia was calculated and used to summarize anesthetic depth for that patient. For a final set of analyses, filtered mean delivered ETAC values were transformed to aaMAC using an equation derived from Mapleson1:

Statistical Analysis

Statistical analysis was performed in R version 3.0.1 (R Foundation for Statistical Computing, Vienna, Austria). Univariate statistics were analyzed using counts and percentages. In multivariate generalized linear models (GLMs), statistical significance for covariates was predefined as a P value <0.01.

In our first phase of analysis, GLMs examined the adjusted association between age and MAC for each hospital. In this linear model, the SE was expected to be different at different ages, if for no other reason than the number of cases in the data at each decade. By using a heteroscedasticity-consistent (i.e., robust) model, SEs were permitted to vary as a function of age. Age was parameterized in decades to make coefficients more easily interpretable and modeled as either a linear term or a piecewise linear spline with a knot imposed at age 65 years. When age is modeled using a piecewise spline, the slope (i.e., the magnitude of observed association between predictor and outcome) is allowed to change at one point in the model, in this case at age 65 years. In other words, the slope of the relationship between age and mean delivered MAC is permitted to be different at age <65 and >65 years, but both lines must have the same value at age 65 years (the knot).

In subsequent modeling, all cases were analyzed in a single GLM with care at a specific hospital modeled through an indicator variable (i.e., varying intercept, constant slope). Covariates with nonoverlapping 95% confidence intervals (CIs) from phase 1 were modeled as hospital interactions (i.e., varying slope). Model coefficients, denoted as β in our results, can be interpreted as the estimated difference in MAC fraction associated with a 1-unit (or 1 category) change in the predictor. Hypothesis tests examining the assumption that the magnitude of the association between age and mean delivered MAC was a 6.7% decrease per decade, as described in the previous studies, were conducted using Wald tests.18 To test for possible differences in anesthetic management of the elderly between the 2 hospitals, an interaction between hospital and age was examined. Finally, model-based predictions were derived from creation of a “standard patient” with fixed values for clinical covariates and variable values for age and data set membership.


At the 2 hospitals during the time period covered by this study, 12,107 general anesthetics were administered for adult patients. After further restricting the data set by case type, minimum duration of maintenance, and eliminating cases with propofol infusions and nitrous oxide, 7878 cases remained for analysis. Table 1 displays univariate statistics for the clinical features under examination at each hospital. As a percentage of total, fewer geriatric cases (age ≥65 years) were performed at hospital B than at hospital A. Reflecting its designation as a trauma center, patients at hospital B were more often male, hospitalized at the time of surgery, and classified as emergency cases. Use of BIS monitoring was far more common at hospital A, whereas isoflurane use was more common at hospital B. To explore the relationship between age and MAC fraction, Figure 1 was plotted, depicting the unadjusted relationship between the 2 variables, suggesting that the relationship might be nonlinear with the slope becoming increasingly negative at higher ages.

Table 1
Table 1:
Patient and Case Descriptors
Figure 1
Figure 1:
Plot of patient age versus mean delivered MAC fraction during maintenance of anesthesia at the 2 hospitals under study. A locally smoothed mean regression curve is superimposed on raw data. MAC = minimum alveolar concentration.

Next, we sought to determine whether relationships between MAC fraction and covariates were similar in both hospitals. A separate GLM was constructed for data from each hospital, and model coefficients and CIs are displayed in Figure 2 and the Supplemental Digital Content (Table, Noting that the relationships between MAC and covariates were generally similar across hospitals, we then created a single GLM for all data and modeled covariates with nonoverlapping CIs from the first set of models as interactions between the covariate and a hospital indicator (Table 2).

Table 2
Table 2:
Model Coefficients from Single Generalized Linear Model (Hospital Modeled as Fixed Effect) Examining Association Between Covariates and MAC
Figure 2
Figure 2:
Regression coefficients from generalized linear models for each data set (separate analyses). Error bars reflect 95% confidence intervals. Covariates with confidence intervals to the right of the vertical dashed line are positively associated with mean delivered MAC, whereas those with confidence intervals to the left are negatively associated with mean delivered MAC. Reference group for ASA is ASA physical status I/II. Reference group for volatile anesthetic agent is sevoflurane. For specific coefficients, see Supplemental Digital Content, Table ( MAC = minimum alveolar concentration; NMBD = neuromuscular-blocking drug.

In this final model, a statistically significant association between increasing age for patients <65 years and MAC fraction was observed with MAC fraction decreasing 1.8% per decade (95% CI, 1.5%–2.2%, P < 0.0001). For patients aged >65 years, MAC seemed to decrease more quickly: 3.8% per decade (95% CI, 2.9%–4.7%, P < 0.0001). In addition, the hospital-specific intercept for hospital B was significantly higher (6.9%, 95% CI, 5.6%–8.3%, P < 0.0001) than that at hospital A, indicating a propensity toward higher absolute doses of volatile anesthetics at hospital B. Statistically significant positive associations relating MAC fraction with both male gender and increasing doses of fentanyl were observed at both hospitals, whereas negative associations were observed at both hospitals for use of BIS monitoring, increasing ASA and ASA emergency status, and vasopressor administration.

Figure 3
Figure 3:
Predicted mean delivered MAC (left) and aaMAC (right) as a function of age for a standardized patient anesthetized in each data set using a linear model with a piecewise linear spline for age and hospital modeled as a random effect. The dashed line represents an “equipotent” volatile anesthetic defined by the Mapleson equation centered on the mean delivered MAC/aaMAC for a 40-year-old patient. aaMAC = age-adjusted minimum alveolar concentration; CI = confidence interval; MAC = minimum alveolar concentration.

Finally, to visualize the implications of these findings on absolute and aaMAC fraction, model-based predictions were created (Fig. 3). The left panel depicts the predicted decrease in absolute ETAC at each hospital and suggests that this age-dependent decrease accelerates after age 65 years. Viewed in terms of aaMAC fraction, however, the right panel demonstrates that patients actually receive higher age-adjusted doses as they age, and this decreases after age 65 years.

Sensitivity Analyses

We conducted 3 sensitivity analyses to strengthen our confidence in the results we obtained (Supplemental Digital Content, First, given the data loss required by summarizing a case’s anesthetic depth in a single measure, we repeated our analyses using median MAC fraction, rather than mean, and found no substantive changes in our results. Second, given our decision to model opioid administration solely using fentanyl administration, we restricted our analysis only to cases in which fentanyl was administered: again, the resulting model coefficients were similar in magnitude and direction. Finally, we considered the possibility that variation in anesthetic practice at the level of the individual provider (modeled as the attending physician starting the case) could influence our outcome. By using a hierarchical linear model with attending provider modeled as a random effect, we found that the association between patient age and mean delivered MAC fraction remained similar in magnitude and direction.


In this retrospective analysis of data obtained from 2 hospitals, we observed a consistent negative association between increasing patient age and the dose of volatile anesthetics delivered to patients. By using linear modeling, we estimate that providers generally delivered 1.8% lower MAC fraction of volatile anesthetic with each additional decade of age to patients younger than 65 years and 3.8% less per decade for elderly patients. These observed decreases are less than those predicted by Mapleson1 and Eger,2 who predicted that for each decade of increasing age, approximately 6.7% less volatile anesthetic was required to obtain equivalent effects. It can be speculated that providers are sensitive to the age–aaMAC relationship but fail to recognize its magnitude, doubt its reliability, or choose to titrate volatile anesthetics to a different—as yet unspecified—endpoint. While acknowledging both the lack of certainty regarding persistent toxicity of volatile anesthetics and the difficulty of calculating the “correct” dose of volatile anesthetic for a given patient, it seems reasonable to conclude that providers did not optimally consider the established relationship between age and anesthetic requirement.

As patients reach age >65 years, provider behavior appears to change with a greater propensity to “dial back” the concentration of volatile anesthetic. This finding, in the absence of marked differences in the types of surgeries or differences in patients’ requirements for anesthesia, suggests that providers do become sensitive to patients’ age when they become “old.” However, after adjustment for age, the dose of volatile anesthetics delivered to older patients was still considerably higher than that delivered to younger patients.

If providers appreciate the relationship between increasing age and decreased anesthetic requirement but remain concerned that using it will increase the risk of intraoperative awareness, what should be their approach? One possible strategy would be to titrate volatile anesthetics with the assistance of depth of anesthesia monitoring and then use means other than increased volatile anesthesia to treat intraoperative hypertension. It has been demonstrated that elderly patients require lower absolute doses of volatile anesthetics to maintain a BIS <50.19 In our study, the use of BIS was frequent only at hospital A, but its use was associated with only a trivial decrease in volatile anesthetic dose. Even if providers were successful at decreasing anesthetic doses as a result of BIS monitoring—a supposition challenged by the complexity of a BIS–ETAC relationship20—the evidence that doing so would alter long-term patient outcomes is conflicting. Although a recent large study was unable to detect improvements in postanesthesia care unit readiness, intensive care unit stay, postoperative nausea/vomiting, or pain among patients whose anesthetics were guided by BIS,21 a different study showed decreased rates of postoperative delirium associated with BIS use.22 In that latter study, however, BIS use was not observed to alter the incidence of postoperative cognitive dysfunction. Therefore, an anesthetic delivery strategy capable of reliably improving the clinical outcomes of elderly patients is a tantalizing but as yet unrealized prospect.

Given the synergy between volatile anesthetics and opioids, appropriately modeling the impact of opiate administration on anesthetic dosing is fundamental to understanding providers’ dosing strategies for volatile anesthetics. We chose to model the interaction between opioid dosing and MAC fraction in a relatively simplistic fashion, a decision that deserves further comment. First, we observed that in our data sets, fentanyl was, by far, the most common short-acting opioid titrated during the early and middle phases of general anesthesia. Second, the basic science literature surrounding alterations in MAC achieved by opioids has historically examined fentanyl and its chemical relatives.23,24 Third, although there are formulae purporting to convert enteral and parenteral opioids from different classes to equivalent doses, these formulae ignore half-life effects and individual patient tolerances that would be critical to accurate adjustment during general anesthesia. Although conceding that our fentanyl-only method of adjustment is, at best, a rough approximation, we find it very interesting that increased doses of fentanyl were associated with increased rather than decreased doses of volatile anesthetics. It is possible that providers elect to uptitrate both opioids and inhaled anesthetics when managing a patient who exhibits hemodynamic or motor evidence of light anesthesia rather than selecting one drug and waiting to determine whether their selection was successful. Although neither class of drugs is ideal as an intraoperative treatment for hypertension, the high prevalence of hypertension among the elderly likely contributed to increased doses of both opioids and volatile anesthetics for older patients.

In designing this analysis, we attempted to gather broadly available clinical data to answer the question of whether any observed associations between age and MAC fraction truly represented generalizable features of anesthetic practice. By initially creating separate models for each hospital, we minimized the imposition of interhospital similarity on our observed associations. Despite this approach, many clinically relevant variables (age, gender, ASA physical status, the impact of fentanyl dosing, and the use of vasopressors) displayed consistent direction and strength of association across hospitals. This suggests to us that our models do accurately describe certain aspects of anesthetic delivery, at least in our healthcare system. Where between-hospital variability was observed, little is known and much remains of interest concerning the delivery of inhaled anesthesia in realistic care settings.

This study is strengthened by its relatively large sample size, inclusion of data from multiple hospitals, and its lack of reliance on a single model to explain all variability in the data. As an observational analysis, we caution the reader that we can make no statement regarding the causality of our observed associations and note several other potential weaknesses in our analysis.

Our data describe the care delivered in a single healthcare system with residents, nurse anesthetists, and attending anesthesiologists who share training, practice patterns, monitors, and anesthetic delivery devices. We can only speculate on the practice of other systems and how the potential for provider training, real-time AIMS feedback, or differently configured monitoring systems could affect anesthetic delivery. The development of multicenter repositories of AIMS data could help to answer these questions.

Because of limitations in our AIMS, we are unable to adjust our analysis for medications taken in the perioperative period. Given the likely association between age and the probability of taking opioids, benzodiazepines, and other classes of medications known to affect anesthetic requirements, this gap in our modeling does represent a potential unmeasured confounder. However, presuming that increasing age is associated with increasing likelihood of using these drugs, we think it is reasonable to conclude that adjustment for their use would increase the association between age and lower mean delivered MAC fraction. Hopefully, further research with richer data sets will explore this important question.

Finally, and importantly, our study is complicated by the challenges of selecting a single parameter to describe the complex distribution of anesthetic drug delivery during the course of an anesthetic and further by the simplification of individual anesthetic depth to a population parameter such as MAC. In addition, the transformation of MAC to aaMAC, although mathematically simple, reduces the complexity of physical age to calendar age and does so using a model that describes only a central tendency. Although other researchers have used the aaMAC concept in recent studies,25 we caution readers to interpret our statistics carefully and to treat our research as preliminary and deserving of critique and confirmation.


Anesthesia providers deliver lower age-adjusted doses of potent volatile anesthetics as age increases, and this relationship strengthens as patients enter the geriatric age range. However, age-adjusted dose seems to increase with age. Given the potential for unnecessarily deep anesthesia to exert negative physiologic and cognitive effects on patients, and assuming our study results are not unique, more effective adjustment of anesthetic delivery for relevant clinical factors, including age, may need to be implemented.


Name: William C. Van Cleve, MD, MPH.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: William C. Van Cleve has seen the original study data, reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.

Name: Bala G. Nair, PhD.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Bala G. Nair has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: G. Alec Rooke, MD, PhD.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: G. Alec Rooke has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

This manuscript was handled by: Ken B. Johnson, MD.


1. Mapleson WW. Effect of age on MAC in humans: a meta- analysis. Br J Anaesth. 1996;76:179–85
2. Eger EI II. Age, minimum alveolar anesthetic concentration, and minimum alveolar anesthetic concentration-awake. Anesth Analg. 2001;93:947–53
3. Sieber FE, Pauldine R Miller’s Anesthesia. 20107th ed. Philadelphia Churchill Livingstone/Elsevier:2270
4. Barash PG Clinical Anesthesia. 2013 Philadelphia Wolters Kluwer Health/Lippincott Williams & Wilkins:458
5. U.S. Census Bureau DIS. 2012 National Population Projections. Available at: Accessed August 28, 2013
6. Xie Z, Dong Y, Maeda U, Alfille P, Culley DJ, Crosby G, Tanzi RE. The common inhalation anesthetic isoflurane induces apoptosis and increases amyloid beta protein levels. Anesthesiology. 2006;104:988–94
7. Zhen Y, Dong Y, Wu X, Xu Z, Lu Y, Zhang Y, Norton D, Tian M, Li S, Xie Z. Nitrous oxide plus isoflurane induces apoptosis and increases beta-amyloid protein levels. Anesthesiology. 2009;111:741–52
8. Culley DJ, Baxter M, Yukhananov R, Crosby G. The memory effects of general anesthesia persist for weeks in young and aged rats. Anesth Analg. 2003;96:1004–9
9. Culley DJ, Baxter MG, Yukhananov R, Crosby G. Long-term impairment of acquisition of a spatial memory task following isoflurane–nitrous oxide anesthesia in rats. Anesthesiology. 2004;100:309–14
10. Kertai MD, Pal N, Palanca BJ, Lin N, Searleman SA, Zhang L, Burnside BA, Finkel KJ, Avidan MSB-Unaware Study Group. B-Unaware Study Group. . Association of perioperative risk factors and cumulative duration of low bispectral index with intermediate-term mortality after cardiac surgery in the B-Unaware Trial. Anesthesiology. 2010;112:1116–27
11. Lindholm ML, Träff S, Granath F, Greenwald SD, Ekbom A, Lennmarken C, Sandin RH. Mortality within 2 years after surgery in relation to low intraoperative bispectral index values and preexisting malignant disease. Anesth Analg. 2009;108:508–12
12. Monk TG, Saini V, Weldon BC, Sigl JC. Anesthetic management and one-year mortality after noncardiac surgery. Anesth Analg. 2005;100:4–10
13. Chan MTV, Cheng BCP, Lee TMC, Gin T. BIS-guided anesthesia decreases postoperative delirium and cognitive decline. J Neurosurg Anesthesiol. 2013;25:33–42
14. ASA Physical Status Classification System. Available at: Accessed June 5, 2015
15. Forane (Isoflurane, USP) Liquid for Inhalation. Available at: Accessed September 1, 2013
16. Sevoflurane, USP Volatile Liquid for Inhalation. Available at: Accessed September 1, 2013
17. Suprane (Desflurane, USP) Volatile Liquid for Inhalation. Available at: Accessed September 1, 2013
18. Fox J, Friendly M, Weisberg S. Hypothesis tests for multivariate linear models using the car package. R J. 2013;5:39–53
19. Matsuura T, Oda Y, Tanaka K, Mori T, Nishikawa K, Asada A. Advance of age decreases the minimum alveolar concentrations of isoflurane and sevoflurane for maintaining bispectral index below 50. Br J Anaesth. 2009;102:331–5
20. Whitlock EL, Villafranca AJ, Lin N, Palanca BJ, Jacobsohn E, Finkel KJ, Zhang L, Burnside BA, Kaiser HA, Evers AS, Avidan MS. Relationship between bispectral index values and volatile anesthetic concentrations during the maintenance phase of anesthesia in the B-Unaware trial. Anesthesiology. 2011;115:1209–18
21. Fritz BA, Rao P, Mashour GA, Abdallah AB, Burnside BA, Jacobsohn E, Zhang L, Avidan MS. Postoperative recovery with bispectral index versus anesthetic concentration-guided protocols. Anesthesiology. 2013;118:1113–22
22. Radtke FM, Franck M, Lendner J, Krüger S, Wernecke KD, Spies CD. Monitoring depth of anaesthesia in a randomized trial decreases the rate of postoperative delirium but not postoperative cognitive dysfunction. Br J Anaesth. 2013;110(suppl 1):i98–105
23. Glass PS, Gan TJ, Howell S, Ginsberg B. Drug interactions: volatile anesthetics and opioids. J Clin Anesth. 1997;9:18S–22S
24. Katoh T, Ikeda K. The effects of fentanyl on sevoflurane requirements for loss of consciousness and skin incision. Anesthesiology. 1998;88:18–24
25. Aranake A, Gradwohl S, Ben-Abdallah A, Lin N, Shanks A, Helsten DL, Glick DB, Jacobsohn E, Villafranca AJ, Evers AS, Avidan MS, Mashour GA. Increased risk of intraoperative awareness in patients with a history of awareness. Anesthesiology. 2013;119:1275–83
© 2015 International Anesthesia Research Society