Secondary Logo

Journal Logo

Economics, Education, and Health Systems Research: Research Report

Anesthetic Management and One-Year Mortality After Noncardiac Surgery

Monk, Terri G. MD, MS*; Saini, Vikas MD, FACC; Weldon, B Craig MD*; Sigl, Jeffrey C. PhD

Author Information
doi: 10.1213/01.ANE.0000147519.82841.5E

Predictors of perioperative morbidity and mortality generally occur in three broad categories: those related to associated comorbid conditions of the patient, those attributable to the surgery itself, and those associated with anesthesia management (1). In the current era, the risk of anesthesia in the immediate perioperative period appears to be quite small (2,3). However, little is known about the effect of anesthetic management on long-term outcomes. Although no consistent long-term benefit has been shown to be related to a specific anesthetic, a recent study has suggested that regional anesthesia may improve survival in some patients (4).

Monitoring of hypnotic depth of anesthesia is now possible using digital signal processing techniques applied to the electroencephalogram (EEG) (5). Although no technology, including pulse oximetry, has yet been definitively shown to reduce mortality, it has been suggested that “monitoring of depth of anesthesia should allow exact dosage of anesthetic drugs and therefore reduce cardiovascular side effects caused by overdosage” (6).

A recent publication by geriatric experts outlined a research agenda aimed at improving postoperative outcomes and recommended that observational cohort studies be performed to identify the most important preoperative predictors of outcome in surgical patients (7). We report the results of just such a prospective observational study designed to evaluate the association of various demographic, clinical, surgical, and intraoperative anesthetic factors, including hypnotic depth of anesthesia, with 1-yr outcome for all-cause mortality.

Methods

The study received ethics committee approval and each participant gave written informed consent. Eligible patients were those aged 18 yr or older presenting for major noncardiac surgery to Shands Hospital at the University of Florida. Exclusion criteria were as follows: preoperative Mini-Mental State Examination score of <24, preoperative disease of the central nervous system (including but not limited to infections, metabolic disorders, tumors, major head trauma, degenerative diseases, major depression or psychosis, and Parkinson’s disease), scheduled surgical procedures known to affect postoperative cognitive function (carotid surgery, neurosurgical and cardiopulmonary bypass procedures), current alcoholism or drug dependence, severe visual or auditory handicaps, inability to comprehend or follow directions, previous enrollment in the study, and refusal to give informed consent for the study.

Preoperative clinical information and medication use were recorded from the patient’s medical history. Comorbidity was quantified using the Charlson Comorbidity Score (8). Additional data were recorded, including ASA physical status and New York Heart Association Functional Class (9). A series of subjective psychological distress and cognitive function measures were determined preoperatively, including Beck Depression Inventory (10), State-Trait Anxiety Inventory (11) and Mini-Mental State Examination. Patients were also tested with a neuropsychological battery designed to assess cognitive function at 1 wk and 3 mo after surgery. The results for postoperative cognitive function will be reported separately to minimize the complexity and length of the manuscript. All data were immediately checked for compatibility and completeness on entry into a dedicated computer program; staff training ensured consistency and reliability of data collection.

All surgery was performed during general anesthesia. There were no protocol-based restrictions of anesthetic or surgical technique, and all clinical decisions were made by the primary anesthesia providers. Perioperative information, including surgery type, hemodynamic variables, anesthetic drugs used, other intraoperative medication administration, and postoperative analgesic management, was recorded on a standardized study data collection sheet.

Hypnotic depth of anesthesia was quantified using a Bispectral Index® (BIS®) monitor and a standardized EEG electrode montage (A1050 BIS Monitor and BIS Sensors; Aspect Medical Systems, Newton, MA). The BIS sensor was placed on the forehead in the standard frontal-temporal position as recommended by the manufacturer. BIS values were recorded throughout the procedure and anesthesia providers were blinded as to the BIS data. At the conclusion of the surgery, the BIS trend was printed and the BIS values later digitized at 5-min intervals. The entire BIS trend was used for analysis. A key study evaluating the efficacy of BIS monitoring compared patients whose anesthetic was titrated to maintain their BIS values within a specific range (45–60) with patients who were not titrated to any specific range and found those titrated to the specified range had faster emergence and discharge, less drug usage, and greater level of postoperative alertness (12). Cumulative deep hypnotic time was therefore quantified as the total amount of time (in hours) that BIS was less than a value of 45.

Hemodynamic variables were electronically recorded preoperatively and at 5-min intervals throughout surgery. Each sample of heart rate, mean arterial blood pressure, and systolic blood pressure was classified as low (below normal), normal, or high (above normal) according to the methodology of Reich et al. (13). The normal ranges were as follows: heart rate, 45–110 bpm, mean arterial blood pressure, 55–100 mm Hg, and systolic pressure, 80–160 mm Hg.

Patients were visited daily until discharge and medical records were reviewed for medical complications. Patients were contacted at 1 yr via a phone interview to determine survival status.

Cox proportional hazards modeling was used to evaluate categorical and continuous cofactors in univariate and multivariate models of time to death or survival to 1 yr. Variables were coded as indicated in Table 1. Categorical variables were grouped in ranges to have at least 2 occurrences of each possible outcome for each value of the variable. As a first step, univariate models of 1-yr mortality were evaluated to assess the predictive capacity of individual variables. Univariate models were computed for all collected variables thought by the investigators to have a potential effect on mortality; these are listed in Table 1. Statistical significance was defined as P ≤ 0.05. The resulting set of statistically significant univariate predictors (Table 2) constituted the starting set of covariates in a forward step-wise multivariate model, which allowed for entry and removal of covariates at each step. Covariate entry was based upon the significance of the score statistic (criteria for entry was P ≤ 0.05). Covariate removal was based on the conditional likelihood-ratio statistic (criteria for removal was P > 0.05). The inclusion/removal process continued until no more covariates could be entered or removed. The relative risk (odds ratio) attributable to each variable was estimated from the parameters of the final multivariate model calculated on the entire study population. To evaluate the stability of the multivariate model, we used the bootstrap method to evaluate the selection frequency of all the variables by estimating step-wise models on 500 random samples obtained using sampling with replacement of the study population (14). The 95% confidence intervals of the relative risks were estimated empirically from a second set of 500 bootstrapped models having the same variables as the final model obtained from the entire study population. The c-statistic was used to compare each patient’s predicted relative risk derived from the multivariate Cox regression model with the actual outcome. A c-statistic of 1.0 indicates perfect prediction, while a c-statistic of 0.5 indicates no predictive value. Analyses were performed with the use of SPSS software (version 11.5.2.1; SPSS Inc., Chicago, IL).

Table 1
Table 1:
Demographic and Clinical Characteristics
Table 2
Table 2:
Univariate Predictors of 1-yr Postoperative Mortality

Results

A total of 1064 patients were enrolled. The clinical and demographic profiles of the study population are shown in Table 1. The mortality rate was 0.7% (7 of 1064) at 30 days and 5.5% (58 of 1064) at 1 yr. The Kaplan-Meier survival curve (Fig. 1) demonstrates that the incidence of death is relatively constant during the year after surgery. The cause of death is shown in Table 3. In elderly patients (≥65 yr of age), the mortality rates were 1.8% at 30 days and 10.3% at 1 yr (5 of 271 and 28 of 271, respectively).

Figure 1
Figure 1:
Figure 1.
Table 3
Table 3:
Cause of Death

Table 2 shows significant univariate predictors of 1-yr mortality. Preoperative clinical indicators (Comorbidity Score ≥3, ASA physical status III–IV, age ≥65 yr, and history of hypertension, heart disease, previous myocardial infarction, or hepatic disease) increased a patient’s risk of mortality. Advanced educational level, larger values of body mass index, increased preoperative diastolic blood pressure and higher preoperative Mini-Mental State Examination score were protective factors. Perioperative factors, including longer surgical duration, intracavitary surgery, longer duration of intraoperative systolic hypotension and increased cumulative deep hypnotic time were associated with increased risk of mortality. The other variables listed in Table 1 were not significant univariate predictors of mortality.

In the multivariate model 184 patients had missing data; therefore, the analysis was performed on the remaining 880. The demographics and mortality rates were not significantly different between the patients with an incomplete data set and those who were included in the multivariate analysis. The multivariate analysis performed on the entire study population found three significant independent predictors of 1-yr mortality (Table 4). The most significant was the Charlson Comorbidity Score. Comorbidity scores of 3 or more increased the risk of mortality by 16.116 times compared with a score of 2 or less. In this group of high-risk patients, the 1-yr mortality rate was 21.7% (43 of 198). Cumulative deep hypnotic time had a relative risk of 1.244, or a 24.4% increased risk per hour of time BIS was <45. Intraoperative systolic hypotension increased mortality risk by 1.036 times per minute below 80 mm Hg. The variables that were selected in the multivariate analysis performed on the entire study population were also those which were most often selected when the same stepwise modeling procedure was applied to 500 randomly selected samples (Table 5), demonstrating that the variables selected by the modeling procedure are not an artifact of the sampling process. The c-statistic of the multivariate model is 0.847 (95% confidence interval, 0.788–0.906, P < 0.0001), indicating the multivariate model has strong ability to predict total patient risk.

Table 4
Table 4:
Multivariate Predictors of 1-yr Postoperative Mortality
Table 5
Table 5:
Selection Frequencies of the Variables in the Forward Step-Wise Cox Regression Model (when fitted to 500 bootstrap samples)

Discussion

We report a prospective observational study evaluating the influence of preoperative patient characteristics and intraoperative anesthetic management on 1-year mortality. Our results indicate that patient comorbidity is the most important predictor of death in the first year after surgery. This finding is in agreement with previous studies that have shown an association between comorbidity and postoperative mortality. The correlation between increasing ASA physical status and the risk of postoperative mortality was originally reported three decades ago (15). A prospective, longitudinal study of complications associated with anesthesia found that advancing age itself adds little risk in the absence of comorbid disease (16). Nevertheless, our finding that a Charlson comorbidity score of 3 or more confers a 16-fold increased risk of 1-year mortality is striking and underscores the need to identify and implement perioperative risk reduction strategies in these patients to determine if outcome can be improved. The fact that the Charlson comorbidity scoring system remained in the multivariate model while ASA physical status score did not implies that it may be a better predictor of long-term outcome than the ASA score.

Our results suggest that mortality in the first year after surgery may be influenced by the intraoperative management of the anesthetic itself, specifically the management of hypnotic depth and arterial blood pressure. The independent association of cumulative deep hypnotic time with 1-year mortality is an unexpected new finding of our study. It is well established that hemodynamic variables are not predictive of hypnotic depth (17). Monitoring hypnotic depth with techniques such as the BIS index yields clinically useful information because routine practice results in significant variability in anesthetic dosing and patient response (18). Elderly patients and patients with many comorbidities require less anesthetic than their healthier, younger counterparts, and BIS monitoring is able to detect this (19). It is possible that the failure of previous studies to detect a long-term effect of general anesthesia on outcome may be explained by the fact that they studied the type of anesthesia administered and not the amount of anesthesia or the anesthetic effect on the brain.

Our finding that every minute of hypotension in the operating room increased the risk of dying in the first year after surgery requires validation. Intraoperative hypotension has been reported as a perioperative risk factor for years; however, a randomized trial of deliberate hypotension (45–55 mm Hg mean arterial blood pressure) with epidural anesthesia found no significant additional morbidity risk at 4 months postoperatively (20). Perhaps adverse effects of intraoperative hypotension require longer follow-up or are related more to the cause of the hypotension (e.g., hypovolemia, myocardial dysfunction, sepsis, or anesthetic overdose) than to the arterial blood pressure per se.

Previous studies have found that the duration of surgery is one of the most significant factors affecting patient outcome (21,22). We found that although surgical duration is associated with increased risk on a univariate basis, it is not a significant multivariate predictor. Our multivariate model includes two other variables, cumulative deep hypnotic time and duration of hypotension, that confer increased risk with additional time under their respective thresholds. It is thus possible that, in this study, the risk ascribed to surgical duration on a univariate basis may be attributable to these two variables, implying that surgical duration per se contributes no additional risk if deep hypnotic time and hypotension are minimized.

How is it possible that anesthetic management could influence 1-year mortality? Inflammation has been linked to the pathogenesis of atherosclerosis, cancer, and Alzheimer’s disease and has been labeled the “secret killer” (23–25). As the majority of the deaths in our study were attributed to cancer (52%) or cardiovascular (17%) causes, it is possible that the anesthetic management altered the inflammatory response (26,27). The postoperative immune response is extremely complex and has both beneficial (improved wound healing and tissue repair) and detrimental (procoagulant and immunosuppressive) effects (26,28). Proinflammatory cytokines, especially tumor necrosis factor-α and interleukin-6, play a major role in the systemic inflammatory response syndrome and multiple organ dysfunction after trauma, ischemia-reperfusion injuries, surgery, and infection, especially sepsis (26,27). Genetic factors can also play a role in enhanced systemic inflammatory responses and increased postoperative morbidity (29). Because anesthetics and techniques per se have been shown to alter cytokine profiles significantly (30–32), our findings of an association of depth and duration of intraoperative hypnosis with 1-year mortality raise the reasonable possibility of a cytokine-mediated process; it is possible that prolonged deep anesthesia may alter the inflammatory response in high-risk patients and predispose them to worsened outcomes.

Because anesthetic drugs and dosing, hypnotic depth, surgical stimulation, and hemodynamic variables are strongly interacting elements, carefully controlled prospective studies will be required (33).

Our study has several limitations. As with any observational study, undefined factors in patient selection may result in random error. However, our use of a large sample diminishes the likelihood of an erroneous result by increasing the precision of the estimate (34). It is notable that since our work was first reported, a similar study by Lennmarken et al. (35) in a cohort of 5057 patients has found a 20% increase in risk for 1-year mortality per hour of BIS values <45. In our study, we are unable to establish the presence of a true dose-response effect because total anesthetic dose was not assessed. However, BIS values are highly correlated with calculated anesthetic concentrations in the brain during anesthesia, and the BIS correlates with hypnotic end-points such as level of sedation and the probability of recall or implicit memory formation (36–38). Because the BIS is so closely linked to the physiologic and pharmacologic effect of anesthetics on the brain, this study raises the interesting question of whether intraoperative cerebral state (i.e., cerebral hypoxia or increased cerebral susceptibility to the effects of anesthetics) is associated with adverse outcomes long after surgery. Another limitation of the study is that we did not evaluate the occurrence of perioperative myocardial damage, a factor that affects long-term morbidity and mortality (39). Despite the limitations of a prospective observational study, our findings are hypothesis generating and suggest that intraoperative anesthetic management may influence long-term outcomes.

In conclusion, in this study of all-cause mortality after noncardiac surgery, we confirm that comorbidity is the major predictor of mortality after major noncardiac surgery but find new associations between intraoperative hypotension, cumulative deep hypnotic time, and 1-year postoperative mortality. The type and duration of surgery, patient age, and other demographic variables did not explain these findings. These associations suggest that intraoperative anesthetic management may affect outcomes over longer time periods than previously appreciated. Clearly, large randomized trials are needed to confirm our results and to determine if changes in anesthetic management can improve long-term outcome in high-risk patients.

The authors would like to acknowledge Maria van der Aa, MS for her assistance with data collection and coordination of the research protocol, Jan van der Aa, Ph.D. for data management and statistical support, and Steffen E. Meiler, MD for his insights into the role of inflammation on long-term outcomes.

References

1. Fleisher LA, Anderson GF. Perioperative risk: How can we study the influence of provider characteristics? Anesthesiology 2002;96:1039–41.
2. Arbous MS, Grobbee DE, van Kleef JW, et al. Mortality associated with anaesthesia: A qualitative analysis to identify risk factors. Anaesthesia 2001;56:1141–53.
3. Sigurdsson GH, McAteer E. Morbidity and mortality associated with anesthesia. Acta Anaesthesiol Scand 1996;40:1057–63.
4. Rasmussen LS, Johnson T, Kuipers HM, et al. Does anaesthesia cause postoperative cognitive dysfunction? A randomized study of regional versus general anaesthesia in 438 elderly patients. Acta Anaesthesiol Scand 2003;47:260–6.
5. Glass PS, Bloom M, Kearse L, et al. Bispectral analysis measures sedation and memory effects of propofol, midazolam, isoflurane, and alfentanil in healthy volunteers. Anesthesiology 1997;86:836–47.
6. Buhre W, Rossaint R. Perioperative management and monitoring in anesthesia. Lancet 2003;362:1839–46.
7. Solomon DH, LoCicero J, Rosenthal RA. New frontiers in geriatric research. New York: American Geriatrics Society, 2004.
8. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies development and validation. J Chron Dis 1987;40:373–83.
9. Fisher JD. New York Heart Association classification. Arch Intern Med 1972;129:836.
10. Beck AT, Ward CH, Mendelson M, et al. An inventory for measuring depression. Arch Gen Psychiatry 1961;4:561–71.
11. Spielberger CD, Gorsuch RL, Lushene RE. The state-trait anxiety inventory: Preliminary test manual for Form X. Tallahassee: Florida State University, 1970.
12. Gan TJ, Glass PS, Windsor A, et al. Bispectral index monitoring allows faster emergence and improved recovery from propofol, alfentanil, and nitrous oxide anesthesia. Anesthesiology 1997;87:808–15.
13. Reich DL, Bennett-Guerrero E, Bodian CA, et al. Intraoperative tachycardia and hypertension are independently associated with adverse outcome in noncardiac surgery of long duration. Anesth Analg 2002;95:273–7.
14. Altman DG, Anderson PK. Bootstrap investigation of the stability of a Cox regression model. Stat Med 1989;8:771–83.
15. Vacanti C, Van Houten R, Hill R. A statistical analysis of the relationship of physical status to postoperative mortality in 68,388 cases. Anesth Analg 1970;49:564–6.
16. Tiret L, Desmonts JM, Hatton F, Vourc’h G. Complications associated with anaesthesia: A prospective study in France. Can Anaesth Soc J 1986;33:336–44.
17. Nakayama M, Hayashi M, Ichinose H, et al. Values of the bispectral index do not parallel the hemodynamic response to the rapid increase in isoflurane concentration. Can J Anaesth 2001;48:958–62.
18. Guignard B, Coste C, Menigaux C, Chauvin M. Reduced isoflurane consumption with bispectral index monitoring. Acta Anaesthesiol Scand 2001;45:308–14.
19. Katoh T, Bito H, Sato S. Influence of age on hypnotic requirement, bispectral index, and 95% spectral edge frequency associated with sedation induced by sevoflurane. Anesthesiology 2000;92:55–61.
20. Williams-Russo P, Sharrock NE, Mattis S, et al. Randomized trial of hypotensive epidural anesthesia in older patients. Anesthesiology 1999;91:926–35.
21. Kessler S, Kinkel S, Kafer W, et al. Influence of operation duration on perioperative morbidity in revision total hip arthroplasty. Acta Orthop Belg 2003;69:328–33.
22. Cook TM, Britton DC, Craft TM, et al. An audit of hospital mortality after urgent and emergency surgery in the elderly. Ann R Coll Surg Engl 1997;79:361–7.
23. Gorman C, Park A, Cray D. The fires within. Time 2004;163:38–46.
24. Libby P. Inflammation in atherosclerosis. Nature 2002;420:868–74.
25. Coussens LM, Werb Z. Inflammation and cancer. Nature 2002;420:860–7.
26. McBride WT, Armstrong MA, McBride SJ. Immunomodulation: an important concept in modern anaesthesia. Anaesthesia 1996;51:465–73.
27. Salo M. Effects of anaesthesia and surgery on the immune response. Acta Anaesthesiol Scand 1992;36:201–20.
28. Nortcliffe SA, Buggy DJ. Implications of anesthesia for infection and wound healing. Int Anesthesiol Clin 2003;41:31–64.
29. Tomasdottir H, Hjartarson H, Ricksten A, et al. Tumor necrosis factor gene polymorphism is associated with enhanced systemic inflammatory response and increased cardiopulmonary morbidity after cardiac surgery. Anesth Analg 2003;97:944–9.
30. Crozier TA, Muller JE, Quittkat D, et al. Effects of anaesthesia on the cytokine response to abdominal surgery. Br J Anaesth 1994;72:280–5.
31. Schneemilch CE, Bank U. Pro- and anti-inflammatory cytokine release in plasma of patients under different anesthetic techniques. Anaesthesiol Reanimat 2001;26:4–10.
32. Kudoh A, Katagai H, Takazawa T, Matsuki A. Plasma proinflammatory cytokine response to surgical stress in elderly patients. Cytokine 2001;15:270–3.
33. Ropcke H, Rehber B, Koenen-Bergmann M, et al. Surgical stimulation shifts EEG concentration-response relationship of desflurane. Anesthesiology 2001;94:390–9.
34. Hulley SB, Newman TB, Cummings SR. Getting started; the anatomy and physiology of clinical research. In: Hulley SB, Cummings SR, Browner WS, et al., eds. Designing clinical research. Philadelphia: Lippincott Williams & Wilkins, 2001:3–15.
35. Lennmarken C, Lindholm M-L, Greenwald SD, Sandin R. Confirmation that low intraoperative BIS levels predict increased risk of post-operative mortality [abstract]. Anesthesiology 2003;99(3, CD-ROM):A303.
36. Ludbrook GL, Visco E, Lam A. Propofol: Relation between brain concentrations, electroencephalogram, middle cerebral artery blood flow velocity, and cerebral oxygen extraction during induction of anesthesia. Anesthesiology 2002;97:1363–70.
37. Olofsen E, Dahan A. The dynamic relationship between end-tidal sevoflurane and isoflurane concentrations and bispectral index and spectral edge frequency of the electroencephalogram. Anesthesiology 1999;90:1345–53.
38. Lubke GH, Kerssens C, Phaf H, Sebel PS. Dependence of explicit and implicit memory on hypnotic state in trauma patients. Anesthesiology 1999;90:670–80.
39. Landesberg G, Shatz V, Akopnik I, et al. Association of cardiac troponin, ck-mb, and postoperative myocardial ischemia with long-term survival after major vascular surgery. J Am Coll Cardiol 2003;42:1547–54.
© 2005 International Anesthesia Research Society