Drugs that affect the renin-angiotensin system, such as angiotensin-converting enzyme inhibitors (ACEI) and angiotensin II receptor subtype 1 antagonists (ARA), are often used in the management of hypertension, congestive heart failure, and chronic renal failure (1). These drugs may interfere with the regulation of arterial blood pressure by several different mechanisms, including sympathetic blockade, reduction in responsiveness to α-adrenergic agonists, impaired degradation of bradykinin, which promotes vasodilation, and inhibition of the receptor binding of angiotensin II (1–4).
Current practice guidelines recommend the perioperative continuation of therapies that have potential for myocardial protection, such as β-adrenoceptor blocking drugs and α2 agonists (5–8), or drugs that can cause rebound hypertension, such as α2-adrenoceptor agonists (9,10). In contrast, preoperative discontinuation of ACEI/ARA therapy has been proposed (11) on the basis of reports of intraoperative hypotension (2,11–15). Several small, controlled, randomized studies found an increased frequency of hypotension after the induction of anesthesia when ACEI/ARA therapy was continued through the morning of surgery compared with discontinuation of therapy the night before surgery (3,13). Moreover, some authors reported a refractory nature of this hypotension (16). However, in contrast to these controlled study conditions, patients in actual clinical settings are treated with a variety of different strategies, including diverse anesthetic techniques and concurrent medications. The clinical significance of variations in the timing of ACEI/ARA administration before surgery has not been evaluated in a large general surgical population.
We conducted this study to determine the effect of timing of preoperative interruption of ACEI/ARA therapy on intraoperative arterial blood pressure in general surgical patients. We hypothesized that the frequency of postinduction hypotension is correlated with the duration of abstinence from ACEI/ARA therapy, even when other aspects of anesthetic management are not controlled. To determine the possible clinical significance of any hypotension, we also examined its timing and treatment.
The study was approved by the Mayo Foundation IRB, and informed consent was obtained from all patients. Patients undergoing surgery at the Mayo Clinic receive preoperative evaluation by a number of physicians who have different practices regarding perioperative treatment with ACEI/ARA: these include discontinuation the day before surgery or treatment on the day of surgery. We took advantage of this variation in practice, which provides a wide range of time intervals between the last administration of ACEI/ARA and the induction of anesthesia.
Patients were recruited over 3 mo, between July 2003 and September 2003, at Saint Mary’s Hospital, a tertiary-care facility that is part of the Mayo Clinic Rochester campus. Because of logistical constraints, not every eligible patient was approached for enrollment, such that the study population represents a convenience sampling of eligible patients. Eligible patients included those receiving ACEI/ARA therapy for at least 3 mo with admission preoperative arterial blood pressures of ≤150/90 mm Hg. Only outpatients and patients admitted to the hospital on the morning of their elective surgery were eligible. Exclusion criteria included patients undergoing cardiovascular surgery, patients whose pathology was characterized by systemic secretion of vasoactive substances (e.g., carcinoid or pheochromocytoma), and patients undergoing procedures in which the use of vasopressors was anticipated (e.g., carotid endarterectomy). Patients who required airway management techniques other than direct laryngoscopy and intubation—such as fiberoptic intubation, mask, or laryngeal mask airway—were not enrolled.
All patients receiving β-adrenergic blocker therapy continued this treatment perioperatively with dosing up to and including the morning of surgery, a practice enforced in our institution. All clinical management decisions were left to the discretion of the attending anesthesiologist, such that anesthetic management was not standardized by any study protocol. Moreover, the primary anesthesiologist was not explicitly informed that the patient signed consent for the study.
Patients were interviewed before surgery by a member of the research team to ensure that criteria for enrollment were met, to ascertain the time and dose of the last ACEI/ARA dose, and to query regarding other antihypertensive medications. All other data were gathered from postoperative review of the medical record. Intraoperative and postoperative recovery room data were collected by using the Picis (Wakefield, MA) electronic anesthesia record system and the Mayo Clinic electronic medical record. Information extracted included (a) demographic characteristics (age, sex, and ASA physical status); (b) comorbid medical conditions [coronary artery disease, stroke, transient ischemic attack, diabetes mellitus, renal functions (as assessed from preoperative serum creatinine concentration), and others as defined by our published criteria (17)]; (c) other relevant preoperative antihypertensive/cardiac medications (ACEIs, ARAs, β-adrenergic blockers, calcium channel antagonists, or diuretics); (d) the exact time (in hours) elapsed between the last dosing of ACEI/ARA and the induction of anesthesia; (e) the amount and type of induction and maintenance anesthetics used; (f) estimated blood loss; and (g) total intraoperative fluids administered.
Arterial blood pressures were measured either noninvasively by an oscillometric device (Philips CMS V24/26; Philips Medical Systems, Best, The Netherlands) or invasively by an arterial catheter. The method of monitoring was left to the discretion of the primary anesthesiologist. Arterial blood pressures were recorded during surgery at 2-min intervals as either the last oscillometric reading obtained before the end of the 2-min interval or the current value measured via arterial catheter at the 2-min mark. The number of episodes of hypotension (see criteria for hypotension below) and their cumulative length (in increments of 2 min) were used for all statistical calculations. In the postanesthesia care unit (PACU), arterial blood pressure was measured at 5-min intervals. Doses and numbers of perioperative treatments with ephedrine, phenylephrine, and vagolytics (atropine and glycopyrrolate) and continuous infusions of potent vasopressors (dopamine, epinephrine, vasopressin, and so on) were recorded.
Hemodynamic variables were retrieved from the Picis system. For each patient, the anesthesia record was printed out and carefully examined by three members of our research team (TC, MD, and DPW). To assess interobserver reliability, reviewing authors collected data independently from 20 charts and confirmed the uniformity of data extraction. Measurement artifacts were minimized in two ways. First, as part of standard practice, anesthesia personnel are instructed to recheck every unexpected extreme arterial blood pressure value before undertaking pharmacologic correction. If detected, these errors can be manually corrected in the automated data collection system. In addition, the oscillometric blood pressure device includes an algorithm for detecting possibly spurious values, and this causes the measurement to be automatically repeated. The medical records were reviewed for postoperative complications such as unplanned intensive care unit admissions, hemodynamic instability in the PACU (arterial blood pressure or heart rate in a range that triggered vasopressor or vagolytic treatment), acute renal impairment (postoperative creatinine increase of >0.5 mg/dL), transient ischemic attacks, stroke, myocardial ischemia/myocardial infarction, and death, as defined by our previously published criteria (17).
We defined moderate hypotension as systolic blood pressure (SBP) ≤85 mm Hg and severe hypotension as SBP ≤65 mm Hg. The threshold for moderate hypotension was based on identical (85 mm Hg) (18) or similar (90 mm Hg) (13,19) SBP values used by other authors. Severe hypotension was defined as SBP ≤65 mm Hg, as decided by an a priori consensus of anesthesiologists (the authors) involved in this study. The primary variables studied were the development of moderate or severe hypotension within 30 min and between 31 and 60 min after anesthetic induction. The rationale for choosing these two time periods was based on the assumption that greater volatility in SBP may be expected immediately after induction because of factors such as volume depletion (from preoperative fasting) and rapid changes in anesthetic depth. The later period (31–60 min) was chosen as a time period characterized by a more stable level of anesthesia, still largely independent from any influences on SBP of large intravascular fluid shifts or blood loss. The primary independent variable was the time elapsed (T) between the last dosing of ACEI/ARA and anesthetic induction (measured in hours).
All variables were analyzed separately within the two patient groups according to T <10 h and T ≥10 h. This threshold was chosen post hoc on the basis of the following analysis. A logistic regression model was fit with T modeled as a smoothing spline with 4 df and moderate hypotension as the binary outcome (Fig. 1). A sharp decline in the likelihood of moderate hypotension was observed at approximately 8–12 h since ACEI/ARA discontinuation. On the basis of this relationship, we chose to base further analyses on a dichotomized version of time since ACEI/ARA discontinuation. By using methods for determining “optimal” cutpoints (20), it was determined T = 10 h best discriminated patients at high and low risk of developing moderate hypotension. This value for T may also have physiologic sense, because it represents approximately one half-life for many of the ACEI/ARA drugs (see Discussion). We used a twofold cross-validation technique with logistic regression to calculate odds ratios (ORs) for low versus high values of time since ACEI/ARA discontinuation (21). Adjusted ORs were estimated from a multivariate logistic regression model that included variables with a known or assumed association with the development of moderate hypotension, namely, age, sex, ASA status, diabetes, use of antihypertensive medications (other than ACEI/ARA), initial SBP reading in the operating room, and type of anesthetic induction drug used (propofol and thiopental).
All study variables are reported separately for low (T <10 h) and high (T ≥10 h) values of time since ACEI/ARA discontinuation. Categorical variables are presented as counts and percentages; differences between groups were assessed for statistical significance by using χ2 or Fisher’s exact tests. Continuous variables are reported as medians and interquartile ranges; differences between groups were assessed for statistical significance by Wilcoxon’s ranked sum test. P values were considered significant at <0.05. All analyses were performed with SAS software, Release 6.12 (SAS Institute Inc., Cary, NC).
On the basis of a previous report (22), we assumed that 65% of the ACEI-treated patients would experience hypotension after the induction of anesthesia. We determined that a sample size of 100 patients per group would provide approximately 80% power to detect a 30% difference in frequency in hemodynamic instability between treatment groups (two-tailed α = 0.05). We also assumed a 2:1 ratio of ACEI-treated patients versus those who stopped using ACEI before surgery and thus targeted 300 patients to provide at least 100 patients in each group.
Over the study period, 308 patients were enrolled. Of this total, 41 were subsequently determined to be ineligible (e.g., they required fiberoptic intubation or received regional anesthesia or monitored anesthesia care), such that 267 patients are the subject of this report. Except for a history of cerebrovascular and peripheral vascular disease, which were more frequent in the T ≥10 h group, the two groups did not differ in the prevalence of comorbidities (Table 1), the choice of induction and maintenance anesthetics, or the use of intraoperative fluids (Table 2). The SBPs and heart rates before anesthetic induction were not different between groups (Table 2).
Patients who took their ACEI/ARA <10 h before anesthesia had an increased likelihood of developing moderate hypotension during the first 30 min after anesthetic induction (60.4% versus 46.3% for T <10 h and T ≥10 h, respectively; P = 0.02 in univariate analysis; (Table 3). This increased risk remained statistically significant even after adjusting for potential confounders (OR, 1.74; 95% confidence interval [CI], 1.03–2.93; P = 0.04) (for a list of confounders see Methods; (Table 3). During the same time period, there was no difference between groups in the frequency of severe hypotension, nor was there a difference in the use of vasopressors. The number of hypotensive episodes during the 31 to 60 min after induction was not different between T <10 h (33.3%) and T ≥10 h (28.5%) (P = 0.43). In univariate analysis, the concurrent use of other antihypertensive drugs, including β-adrenergic blockers, calcium channel blockers, and diuretics, either alone or in combination, did not significantly affect the incidence of hypotension when analyzed across both time periods (all combinations, P > 0.05; (Table 4). This lack of significant differences persisted after adjusting for potential confounding factors. Increased age was not related to the development of hypotension by univariate (OR per 10-yr increase in age, 0.98; 95% CI, 0.79–1.23; P = 0.88) or multivariate (adjusted OR per 10-yr increase in age, 0.88; 95% CI, 0.68–1.12; P = 0.29) analysis (data not shown in tables). The postoperative hemodynamic profile in the PACU did not differ between groups. The overall incidence of major postoperative complications was infrequent and similar between the T <10 h and T ≥10 h groups (Table 5).
In general, the decision to stop drugs with cardiovascular effects before surgery depends on the risk balance between the deleterious interaction of the drug with anesthetics versus morbidity resulting from hemodynamic effects that may occur in the absence of these drugs. For example, continuation of perioperative β-adrenergic blockers and α2 agonists because of their protective role against myocardial morbidity is widely accepted (5–8,23–25). Similarly, uninterrupted perioperative use of calcium channel blockers may reduce perioperative ischemia and dysrhythmias and also decrease death and myocardial infarction rates (26,27). In contrast, intraoperative hemodynamic instability has been observed in patients treated with preoperative ACEI (13,28) and ARA (11,16), and adequate perioperative management of antihypertensive therapy in patients receiving angiotensin system blockers has been debated (1,11,13,15,29,30). Because of the fear of postinduction hypotension (13), some groups have developed algorithms to routinely omit the ACEI/ARA therapy before surgery (31). Other authors have suggested that omitting ACEI before surgery does not have sufficient advantage to be routinely recommended (29).
Existing studies examining the perioperative use of ACEI/ARA randomized patients to receive their last dose either the night before or the morning of surgery. These studies were performed on smaller populations and higher-risk patients (3,13,15). They found no alteration in hemodynamic stability during cardiac surgery (3) but found an increased probability of hypotension at induction for vascular surgery (13,15). However, there is little information regarding how the preoperative management of these drugs may affect actual clinical practice, in which there are many other uncontrolled factors (such as the concomitant use of other antihypertensive drugs). This study addresses this issue. Our major finding is that omitting ACEI/ARA therapy at >10 hours before anesthesia significantly reduced the likelihood of developing hypotension. Of interest, this time period roughly corresponds to most of the half-lives of ACEI/ARA used in this study (Table 6). There were more hypotensive episodes during the first 30 minutes compared with 31–60 minutes after induction; this may be attributed to improved cardiovascular adaptation and increased volume loading, which led to stabilization of arterial blood pressure. Of note, the percentage of hypotensive episodes after induction in the T <10 h and T ≥10 h groups was 60.4% and 46.3%, respectively, which resembles the incidence of hypotension reported by Ryckwaert and Colson (22) (66% versus 44% in ACEI-treated and control patients, respectively). Therefore, it appears that the likelihood for having postinduction hypotension in patients who do not interrupt their ACEI/ARA therapy before surgery is roughly 20 percentage points more than in other populations undergoing anesthesia. Although we did observe that recent use of ACEI/ARA drugs was a significant independent risk factor for the development of moderate hypotension in the period immediately after induction, it should be noted that the lower limit of the 95% confidence interval for the OR approached 1, indicating that this was not a pronounced effect. In addition, most hypotensive episodes in our study prompted relatively simple interventions (ephedrine/phenylephrine/fluids), and only a few patients (6 of 267; 2.2%) required vasopressor infusions to maintain normotension (Table 3). This is consistent with the findings of other authors (28), although profound hypotension not responsive to standard measures has been reported (11,16,19). We could not confirm that the concurrent use of antihypertensive medications affected the incidence of hypotension after anesthetic induction (Table 4). In contrast, Colson et al. (28) suggested more pronounced hypotension in patients who were, besides ACEI, taking multiple antihypertensive drugs. We did not identify increased postoperative morbidity or mortality that could be directly associated with hemodynamic instability; however, the sample size in our study was underpowered to discover differences between the two treatment groups with regard to any major adverse events.
Despite reports of intraoperative hypotension, some authors have recommended continued perioperative ACEI/ARA therapy. Boldt et al. (32) suggested that their uninterrupted use could be associated with a reduction of ischemia-related myocardial cell damage in cardiac surgery. Furthermore, Colson et al. (33) found that during cardiopulmonary bypass, effective renal plasma flow and the glomerular filtration rate remained unaltered, whereas the urinary excretion of sodium was greater in patients receiving captopril compared with the placebo group. No studies examined the potential protective value of uninterrupted ACEI therapy in patients at higher risk for developing renal failure. Therefore, the exact management of ACEI/ARA therapy in the perioperative period still needs to be studied in larger patient populations with different risk factors to assess the risk-benefit profile of these therapies.
In humans, most ACEI undergo renal elimination via glomerular filtration or tubular secretion. Renal insufficiency may have a significant effect on the half-life of certain ACEI, and the altered pharmacokinetics of ACEI in chronic renal failure are considered a potential hazard; thus, a reduced-dose regimen is recommended (34). However, these effects do not apply to all ACEI, and some ACEI (e.g., benazeprilat) have both renal and hepatic elimination; i.e., this drug is significantly eliminated even in patients with end-stage renal failure (34). In our study, there was no significant difference in median creatinine concentrations across groups. In addition, we had only nine patients in the entire study with creatinine values >2.0 mg/dL. Therefore, in this study, pharmacokinetic complexity, which may arise from slower elimination of ACEI/ARA, did not appear to be an issue.
This study had several limitations. By design, anesthetic management was nonstandardized, and this could have induced variability between groups. Both known and unknown confounding variables may not have been evenly distributed among patients and thus may not have been accounted for in the analyses. However, there was no difference among anesthetics (types and dosages) used in the two groups. Furthermore, because the threshold for treatment of hypotension was not standardized by protocol, we may have underestimated the incidence of hypotensive episodes in this report because some anesthesiologists use vasopressors or fluid boluses before hypotension, as defined in this study, occurs. We did not study patients who were not receiving ACEI/ARA therapy. Thus, we cannot determine what proportion of hypotensive episodes can be attributed to this therapy, especially in patients who discontinued therapy well before surgery. The review of arterial blood pressures was performed retrospectively from electronic numerical records, and there is always the probability that some arterial blood pressure values represent artifacts. Finally, a strategy for finding an “optimal” cutpoint was used to dichotomize time since ACEI/ARA discontinuation. This strategy may be problematic both for assessing the statistical significance of a prognostic variable (35,36) and for estimating its effect size (21,37). To circumvent these potential problems, we used a twofold cross-validation technique (21) for estimating the appropriate P values and ORs for low versus high values of time since ACEI/ARA discontinuation.
In conclusion, in patients receiving chronic ACEI/ARA therapy who receive general anesthesia, the administration of these drugs <10 hours before anesthesia is a significant independent risk factor for developing moderate hypotension within 30 minutes after induction. However, this hypotension responded to conventional therapy and thus seemed to be of little clinical consequence in the surgical population studied. These results do not provide a strong rationale for strict guidelines regarding preoperative management of these drugs. Nonetheless, preoperative withholding of ACEI/ARA should be considered for patients who may be especially prone to hypotension-induced complications (e.g., patients with severe aortic stenosis or critical cerebrovascular disease).
We thank Heidi Woxland for assistance with data acquisition.
1. Behnia R, Molteni A, Igic R. Angiotensin-converting enzyme inhibitors: mechanisms of action and implications in anesthesia practice. Curr Pharm Des 2003;9:763–76.
2. Kellow NH. The renin-angiotensin system and angiotensin converting enzyme (ACE) inhibitors. Anaesthesia 1994;49:613–22.
3. Licker M, Neidhart P, Lustenberger S, et al. Long-term angiotensin-converting enzyme inhibitor treatment attenuates adrenergic responsiveness without altering hemodynamic control in patients undergoing cardiac surgery. Anesthesiology 1996;84:789–800.
4. Licker M, Schweizer A, Hohn L, et al. Cardiovascular responses to anesthetic induction in patients chronically treated with angiotensin-converting enzyme inhibitors. Can J Anaesth 2000;47:433–40.
5. Mangano DT, Layug EL, Wallace A, Tateo I. Effect of atenolol on mortality and cardiovascular morbidity after noncardiac surgery: Multicenter Study of Perioperative Ischemia Research Group. N Engl J Med 1996;335:1713–20.
6. Wallace A, Mangano DT. Use of beta-blockade to prevent death after noncardiac surgery. West J Med 1997;166:203–4.
7. Wallace A, Layug B, Tateo I, et al. Prophylactic atenolol reduces postoperative myocardial ischemia: McSPI Research Group. Anesthesiology 1998;88:7–17.
8. Stevens RD, Burri H, Tramer MR. Pharmacologic myocardial protection in patients undergoing noncardiac surgery: a quantitative systematic review. Anesth Analg 2003;97:623–33.
9. Neusy AJ, Lowenstein J. Blood pressure and blood pressure variability following withdrawal of propranolol and clonidine. J Clin Pharmacol 1989;29:18–24.
10. Schmidt GR, Schuna AA. Rebound hypertension after discontinuation of transdermal clonidine. Clin Pharm 1988;7:772–4.
11. Brabant SM, Bertrand M, Eyraud D, et al. The hemodynamic effects of anesthetic induction in vascular surgical patients chronically treated with angiotensin II receptor antagonists. Anesth Analg 1999;89:1388–92.
12. Kamp-Jensen M, Olsen NV. ACE inhibitors and general anesthesia. Ugeskr Laeger 1994;156:5539.
13. Coriat P, Richer C, Douraki T, et al. Influence of chronic angiotensin-converting enzyme inhibition on anesthetic induction. Anesthesiology 1994;81:299–307.
14. Williams NE. Profound bradycardia and hypotension following spinal anaesthesia in a patient receiving an ACE inhibitor: an important ‘drug’ interaction? Eur J Anaesthesiol 1999;16:796–8.
15. Bertrand M, Godet G, Meersschaert K, et al. Should the angiotensin II antagonists be discontinued before surgery? Anesth Analg 2001;92:26–30.
16. Brabant SM, Eyraud D, Bertrand M, Coriat P. Refractory hypotension after induction of anesthesia in a patient chronically treated with angiotensin receptor antagonists. Anesth Analg 1999;89:887–8.
17. Hosking MP, Warner MA, Lobdell CM, et al. Outcomes of surgery in patients 90 years of age and older. JAMA 1989;261:1909–15.
18. Tuman KJ, McCarthy RJ, O’Connor CJ, et al. Angiotensin-converting enzyme inhibitors increase vasoconstrictor requirements after cardiopulmonary bypass. Anesth Analg 1995;80:473–9.
19. Boccara G, Ouattara A, Godet G, et al. Terlipressin versus norepinephrine to correct refractory arterial hypotension after general anesthesia in patients chronically treated with renin-angiotensin system inhibitors. Anesthesiology 2003;98:1338–44.
20. Mazumdar M, Glassman JR. Categorizing a prognostic variable: review of methods, code for easy implementation and applications to decision-making about cancer treatments. Stat Med 2000;19:113–32.
21. Faraggi D, Simon R. A simulation study of cross-validation for selecting an optimal cutpoint in univariate survival analysis. Stat Med 1996;15:2203–13.
22. Ryckwaert F, Colson P. Hemodynamic effects of anesthesia in patients with ischemic heart failure chronically treated with angiotensin-converting enzyme inhibitors. Anesth Analg 1997;84:945–9.
23. Auerbach AD, Goldman L. Beta-blockers and reduction of cardiac events in noncardiac surgery: clinical applications. JAMA 2002;287:1445–7.
24. Nishina K, Mikawa K, Uesugi T, et al. Efficacy of clonidine for prevention of perioperative myocardial ischemia: a critical appraisal and meta-analysis of the literature. Anesthesiology 2002;96:323–9.
25. Poldermans D, Boersma E, Bax JJ, et al. Bisoprolol reduces cardiac death and myocardial infarction in high-risk patients as long as 2 years after successful major vascular surgery. Eur Heart J 2001;22:1353–8.
26. Wijeysundera DN, Beattie WS. Calcium channel blockers for reducing cardiac morbidity after noncardiac surgery: a meta-analysis. Anesth Analg 2003;97:634–41.
27. Butterworth J, Furberg CD. Improving cardiac outcomes after noncardiac surgery. Anesth Analg 2003;97:613–5.
28. Colson P, Saussine M, Seguin JR, et al. Hemodynamic effects of anesthesia in patients chronically treated with angiotensin-converting enzyme inhibitors. Anesth Analg 1992;74:805–8.
29. Pigott DW, Nagle C, Allman K, et al. Effect of omitting regular ACE inhibitor medication before cardiac surgery on haemodynamic variables and vasoactive drug requirements. Br J Anaesth 1999;83:715–20.
30. Samain E, Marty J. How to handle cardiovascular treatments during general anesthesia? Ann Cardiol Angeiol (Paris) 1999;48:624–9.
31. Whalley DG, Maurer WG. Hemodynamic effects of anesthesia in patients with ischemic heart failure chronically treated with angiotensin-converting enzyme inhibitors. Anesth Analg 1997;84:945–9.
32. Boldt J, Rothe G, Schindler E, et al. Can clonidine, enoximone, and enalaprilat help to protect the myocardium against ischaemia in cardiac surgery? Heart 1996;76:207–13.
33. Colson P, Ribstein J, Mimran A, et al. Effect of angiotensin converting enzyme inhibition on blood pressure and renal function during open heart surgery. Anesthesiology 1990;72:23–7.
34. Toutain PL, Lefebvre HP, Laroute V. New insights on effect of kidney insufficiency on disposition of angiotensin-converting enzyme inhibitors: case of enalapril and benazepril in dogs. J Pharmacol Exp Ther 2000;292:1094–103.
35. Altman DG, Lausen B, Sauerbrei W, Schumacher M. Dangers of using “optimal” cutpoints in the evaluation of prognostic factors. J Natl Cancer Inst 1994;86:829–35.
36. Hilsenbeck SG, Clark GM. Practical p-value adjustment for optimally selected cutpoints. Stat Med 1996;15:103–12.
© 2005 International Anesthesia Research Society
37. Lausen B, Schumacher M. Evaluating the effect of optimized cutoff values in the assessment of prognostic factors. Comput Stat Data Anal 1996;21:307–26.