Current Issue Previous Issues Published Ahead-of-Print CME Subjects Timely Topics Translations Podcasts For Authors Journal Info
Skip Navigation LinksHome > July 2011 - Volume 113 - Issue 1 > Does Perioperative Systolic Blood Pressure Variability Predi...
Anesthesia & Analgesia:
doi: 10.1213/ANE.0b013e31820f9231
Cardiovascular Anesthesiology: Research Reports

Does Perioperative Systolic Blood Pressure Variability Predict Mortality After Cardiac Surgery? An Exploratory Analysis of the ECLIPSE trials

Aronson, Solomon MD*; Dyke, Cornelius M. MD; Levy, Jerrold H. MD; Cheung, Albert T. MD§; Lumb, Philip D. MB, BS; Avery, Edwin G. MD; Hu, Ming-yi PhD#; Newman, Mark F. MD*

Free Access
Supplemental Author Material
Continued Medical Education
Chinese Language Editions
Continuing Medical Education
Article Outline
Collapse Box

Author Information

From the *Department of Anesthesiology, Duke University Medical Center, Durham, North Carolina; SouthEast Texas Cardiovascular Surgery Associates, Houston, Texas; Department of Anesthesiology, Emory University School of Medicine, Atlanta, Georgia; §Department of Anesthesiology and Critical Care, University of Pennsylvania, Philadelphia, Pennsylvania; Department of Anesthesiology, Keck School of Medicine, Los Angeles, California; Department of Anesthesiology and Perioperative Medicine, University Hospitals Case Medical Center, Cleveland, Ohio; and #The Medicines Company, Parsippany, New Jersey.

Supported by a grant from The Medicines Company, 8 Sylvan Way, Parsippany, NJ.

Conflict of Interest: See Disclosures at the end of the article.

Reprints will not be available from the authors.

Address correspondence to Solomon Aronson, MD, Duke University Medical Center, Duke South, Baker House, Room 102DUMC, Box 3094, Durham, NC 27710. Address e-mail to arons002@mc.duke.edu.

Accepted December 28, 2010

Published ahead of print February 23, 2011

Collapse Box

Abstract

BACKGROUND: Few studies describe an association of perioperative blood pressure stability with postoperative outcome. We tested the hypothesis that systolic blood pressure (SBP) variability in patients undergoing cardiac surgery is associated with 30-day mortality.

METHODS: Perioperative blood pressure variability was evaluated in the 1512 patients who were randomized and had perioperative hypertension in the ECLIPSE trials. Blood pressure variability was assessed as the product of magnitude × duration of SBP excursions outside defined SBP ranges (area under the curve). SBP ranges were analyzed from 65 to 135 mm Hg intraoperatively and 75 to 145 mm Hg pre- or postoperatively, up to 105 to 135 mm Hg intraoperatively and 115 to 145 mm Hg pre- or postoperatively, with the narrower ranges defined by progressively increasing the lower SBP limit by 10 mm Hg increments. Multiple logistic regression was used to assess the association of blood pressure variability with 30-day mortality obtained from the primary ECLIPSE trial results.

RESULTS: Increased SBP variability outside a range of 75 to 135 mm Hg intraoperatively and 85 to 145 mm Hg pre- and postoperatively is significantly associated with 30-day mortality. The odds ratio was 1.16 (95% confidence interval, 1.04–1.30) for 30-day mortality risk per incremental SBP excursion of 60 mm Hg × min/h. The predicted probability of 30-day mortality increased for low-risk patients from 0.2% to 0.5%, and for high-risk patients from 42.4% to 60.7% if the area under the curve increased from 0 to 300 mm Hg × min/h.

CONCLUSIONS: Perioperative blood pressure variability is associated with 30-day mortality in cardiac surgical patients, proportionate to the extent of SBP excursions outside the range of 75 to 135 mm Hg intraoperatively and 85 to 145 mm Hg pre- and postoperatively. Predicted mortality was greater for high-risk patients than for low-risk patients.

Preexisting hypertension is present in more than two-thirds of all patients undergoing cardiac surgery and has been identified as an important risk factor for adverse perioperative outcome.110 However, a target range for arterial blood pressure during the acute perioperative period is not clearly defined.1,913 Furthermore, optimal perioperative blood pressure management strategy in high-risk patients is even less well understood. In preexisting hypertension, the autoregulatory capacity of the brain14,15 and kidney16 is impaired, potentially influencing end-organ tolerance of high or low blood pressures. As a result, the therapeutic window of acceptable blood pressure during surgery may be narrowed and shifted toward higher pressures in these patients.3,4,11 Little is understood regarding the association of acute perioperative blood pressure changes with outcomes in these or other vulnerable patients.

The Evaluation of CLevidipine In the Perioperative Treatment of Hypertension Assessing Safety Events (ECLIPSE) trials compared clevidipine (CLV) (Cleviprex®; The Medicines Company, Parsippany, NJ) with nitroglycerin (NTG), sodium nitroprusside (SNP), and nicardipine (NIC) for the treatment of perioperative hypertension in patients undergoing cardiac surgery, approximately half of whom had documented preexisting hypertension.17 Systolic blood pressure (SBP) control was a secondary end point of the ECLIPSE trials.

In the present analysis, we evaluated the pooled ECLIPSE dataset, combining data from the 3 trials and all treatment arms to test our hypothesis that SBP variability in patients undergoing cardiac surgery is associated with the incidence of mortality postoperatively. Our objectives were, first, to investigate an association between SBP control and 30-day mortality and, second, to further characterize risk within the pooled study population.

Back to Top | Article Outline

METHODS

The ECLIPSE trials,17 consisting of 3 parallel, prospective, randomized (1:1), open-label trials performed at 61 United States study sites, were conducted in compliance with the International Conference on Harmonization Good Clinical Practice Guidelines and the Declaration of Helsinki.18,19 The studies were approved by the IRB at each participating institution. Written informed consent was obtained from all patients before enrollment.

The ECLIPSE trials evaluated CLV versus an active comparator (NTG, SNP, or NIC) in patients 18 years or older and scheduled for elective coronary artery bypass graft (CABG) surgery and/or cardiac valve repair or replacement. Patients were excluded for recent cerebrovascular accident, permanent ventricular pacing, childbearing potential, inability to tolerate study treatment, any condition or disease putting the patient at risk if enrolled in the trial per investigator assessment, or participation in another study within 30 days of study start. Eligible patients were randomized to receive antihypertensive treatment with CLV, NTG, SNP, or NIC. To receive study treatment, patients had to have perioperative hypertension as a postrandomization criterion assessed by the site investigators. Patients were treated to achieve and maintain the blood pressure levels considered clinically appropriate by investigators, and were evaluated for the incidences of prespecified safety outcomes as the primary end point of the trials.

Blood pressure control, a prespecified secondary efficacy end point, was assessed based on SBP. SBP measurements were obtained by study site investigators via direct intraarterial monitoring per their respective institutional standard of care, recorded at specified time intervals (see below), and entered into electronic case report forms. For analysis purposes, blood pressure control was defined as the summation of cumulative SBP excursions outside a defined SBP range (Fig. 1). This parameter, referred to as area under the curve (AUC),a captured the products of magnitude × duration of SBP variability as excursions outside the ranges. The prespecified SBP range was 65 to 135 mm Hg during the intraoperative period (from chest incision through chest closure) and 75 to 145 mm Hg during the pre- and postoperative time periods.

Figure 1
Figure 1
Image Tools

AUC was calculated from time-stamped, recorded blood pressure measurements in the case report form. The trapezoidal rule20 for approximating the area between a curve and an axis, by obtaining the areas of trapezoids, was used to estimate the area of the blood pressure–time curve outside the predefined range. Blood pressure was recorded from an intraarterial catheter placed according to usual practice from preinduction through the time of its removal or at 24 hours, whichever occurred first. Measurements were obtained every 15 minutes preoperatively, every 5 minutes intraoperatively, and every 15 minutes postoperatively for 4 to 6 hours, then once every hour through 24 hours.

A large AUC indicated a lesser degree of blood pressure control and greater blood pressure variability. Blood pressure control was characterized by calculating the total area of the SBP–time curve outside (either above or below) the prespecified SBP range, normalized per hour, in units of mm Hg × min/h (Fig. 1).

To evaluate the association between blood pressure variability as assessed by AUC and 30-day mortality, blood pressure and mortality data from the 3 ECLIPSE trials were pooled for analysis. A total of 1964 patients were randomized in the ECLIPSE trials. Statistical analysis was performed using the 1512 patients who were randomized and met the postrandomization inclusion criterion of having perioperative hypertension. This population was defined as modified intent-to-treat patients (n = 1512). Patients who were randomized but failed to meet the postrandomization inclusion criterion (n = 452), who were thus untreated and in whom outcome data were not captured, were excluded from the analyses. Thirty-day mortality data were available for all treated patients. Multiple logistic regression analysis was performed to identify from the variables collected in the ECLIPSE trials, which variables, including SBP variability as assessed by AUC, were associated with an increased risk of 30-day mortality. Candidate risk factors, such as demographics, baseline characteristics (e.g., preoperative hemodynamic and clinical laboratory results such as creatinine clearance calculated by the Cockcroft-Gault equation), medical history, treatment group, and procedural characteristics (e.g., surgery duration [defined as the time interval from chest incision to chest closure] and additional procedures during index surgical procedures) had been collected in the study case report forms and were considered in the model selection processes (Appendix). Associations between each of the candidate risk factors and 30-day mortality were investigated using a χ2 test or Fisher exact test, as appropriate, for categorical variables and a t test or Wilcoxon rank sum test for continuous variables. All risk factors with a nominal 2-sided P value ≤0.15 along with interactions between AUC and each risk factor were then entered into independent stepwise, backward, and forward model selection processes. A P value <0.05 was required for inclusion in the final logistic regression model. The results from these processes were compared to construct the final logistic regression model. As a final step, all risk factors and interactions had to have a P value <0.05 in the final model.

In addition to analysis of AUC based on the ECLIPSE-specified SBP range (65–135 mm Hg intraoperatively and 75–145 mm Hg pre- and postoperatively), post hoc analysis of the same dataset for association of 30-day mortality to AUC at progressively narrower perioperative SBP ranges was performed to explore SBP range sensitivity. These SBP ranges were progressively narrowed by raising the lower limit of the range by 10 mm Hg increments up to 105 to 135 mm Hg intraoperatively and 115 to 145 mm Hg pre- and postoperatively.

An odds ratio was derived from the final logistic model based on an increment of 60 mm Hg × min/h in AUC. This increment is equivalent to an average SBP excursion of 1 mm Hg (above or below the range) over 60 minutes, or an SBP of 60 mm Hg for 1 minute, among other examples.

The predicted risk of 30-day mortality with AUC was also evaluated in low- and high-risk patient profiles. These patient profiles were derived from the logistic regression model after a final model was obtained through model selection processes. First, it was noted which variables from the final multiple logistic model had odds ratios of >1 (higher risk of mortality with higher values for the variable) or <1 (higher risk of mortality with lower values for the variable). The low-risk patient profile was defined for continuous variables as <58 years of age (first 25% of ECLIPSE patients for age), <2.62 hours of surgery duration (first 25% of ECLIPSE patients for surgery duration), and preoperative hemoglobin >15 g/dL (upper 25% of ECLIPSE patients for preoperative hemoglobin); and for binary variables as preoperative serum creatinine <1.2 mg/dL, preoperative SBP ≤160 and diastolic blood pressure (DBP) ≤105 mm Hg, no history of chronic obstructive pulmonary disease (COPD), and no other procedures besides the index surgery. The high-risk patient profile was defined for continuous variables as >73 years of age (upper 25% of ECLIPSE patients for age), >4.12 hours of surgery duration (upper 25% of ECLIPSE patients for surgery duration), preoperative hemoglobin <12.50 g/dL (first 25% of ECLIPSE patients for preoperative hemoglobin); and for binary variables as preoperative serum creatinine ≥1.2 mg/dL, preoperative SBP >160 or DBP >105 mm Hg, with history of COPD, and with an additional procedure(s) besides the index surgery.

In addition to the multiple logistic regression analysis, Kaplan-Meier survival analysis was used to generate 30-day survival curves, which were separated by 4 patient groups categorized by the quartiles of ECLIPSE patients based on AUC values. The log-rank test was used to compare the overall differences in the 30-day survival distributions among the 4 AUC quartile groups.

All statistical analyses were performed using the SAS® system (version 8.2; SAS Institute, Cary, NC). For continuous demographic or baseline variables, descriptive statistics are presented as number of patients (n), and as mean and standard deviation for approximately normal data distributions. For categorical variables, the data are presented as frequencies and percentages.

Back to Top | Article Outline

RESULTS

Patient Characteristics and Outcomes

Demographics, medical history and other characteristics for the pooled ECLIPSE patient population are summarized in Table 1. Slightly more than half of the patients were aged 65 years or older and most were male and Caucasian. Diabetes, renal impairment, and previous stage 1 hypertension were prevalent. Approximately 11% of patients were taking β-blockers as antihypertensive therapy preoperatively on the day of surgery (Table 1), and approximately half of these patients (87 of 174, 47%) had β-blockers withdrawn before surgery. Most patients underwent CABG surgery with cardiopulmonary bypass as the index procedure, with the remainder undergoing valve repair and/or replacement, off-pump CABG, or a combination of procedures (Table 2). The mean procedure duration was 3.5 hours. The mean duration “on pump” during cardiopulmonary bypass was 1.6 hours (n = 1304). The median duration of SBP measurements was approximately 23.5 hours. Patients remained within the protocol-prespecified SBP ranges (65–135 mm Hg intraoperatively and 75–145 mm Hg pre- and postoperatively) most of the time throughout the duration of the ECLIPSE trials (median percentage 97.8% [range, 4.0%–100%]).

Table 1
Table 1
Image Tools
Table 2
Table 2
Image Tools

The overall pooled incidence of 30-day all-cause mortality for patients enrolled in the ECLIPSE trials was 3.2% (48 of 1512). The causes of death reported by investigators were consistent with complications of cardiac surgery, or with the natural history of advanced end-organ dysfunction, and followed an expected distribution with cardiac deaths most common initially (over the first 10 days postoperatively) and deaths from respiratory and infectious disease more common subsequently. Reported causes of death were cardiac or cardiovascular for 21 patients (cardiac arrest, n = 10; pulmonary embolism, n = 4; coronary artery disease, n = 3; cardiac failure, n = 1; right ventricular failure, n = 1; right ventricular rupture, n = 1; aortic rupture, n = 1), respiratory for 11 patients (pulmonary distress/failure, n = 9; hemorrhagic pulmonary edema, n = 1; tension pneumothorax, n = 1), multiorgan failure for 7 patients, infectious disease (pneumonia/sepsis) for 3 patients, stroke for 2 patients, surgical complications for 2 patients, bowel infarction for 1 patient, and unknown for 1 patient in a long-term care facility and lost to follow-up. Patients with multiorgan failure as cause of death all had total AUC values (based on SBP ranges of 75–135 mm Hg intraoperatively and 85–145 mm Hg pre- and postoperatively) at or above the third quartile.

The mortality rates were 2.9% for patients who underwent CABG only and 2.1% for patients who underwent valve surgery only as the index procedure. Patients who underwent a combination of CABG and valve surgery had a 7.8% mortality rate.

The distribution of total AUC (based on the SBP range of 75–135 mm Hg intraoperatively and 85–145 mm Hg pre- and postoperatively) in the pooled ECLIPSE patient population was evaluated. Total AUC equals the summation of AUC above the SBP range and AUC below the SBP range. Patients with total AUC >0 mm Hg × min/h were divided into 9 approximately equal-sized groups ranked by total AUC; an additional group had AUC = 0 mm Hg × min/h. Patient groups with higher total AUC values showed numerically higher AUC above the SBP range and higher AUC below the SBP range (Fig. 2A), and the 5 groups with the highest total AUC had the highest incidences of mortality per group (Fig. 2B).

Figure 2
Figure 2
Image Tools

Six of the 48 total deaths occurred in patients with no recorded SBP variability (“AUC above” = 0 mm Hg × min/h and “AUC below” = 0 mm Hg × min/h). Seven deaths occurred in patients with SBP excursions below the range only (AUC above = 0 mm Hg and AUC below >0 mm Hg × min/h), 14 deaths in patients with SBP excursions above the range only (AUC above >0 mm Hg × min/h and AUC below = 0 mm Hg × min/h), and 21 deaths (44% of total deaths) in patients who had SBP excursions both above and below the range (AUC above >0 mm Hg × min/h and AUC below >0 mm Hg × min/h). Because both SBP excursions above and below were associated with mortality, it was considered appropriate to combine them as total AUC in an overall analysis of association with mortality.

Back to Top | Article Outline
Association of Blood Pressure Variability and 30-Day Mortality

Risk variables that were investigated for association with mortality but did not form part of the final model included treatment regimen, combined CABG and valve surgery as procedure type, medical history of atrial fibrillation/flutter, and atrial fibrillation at the time of surgery (Appendix). Interaction between AUC and each of the risk factors was also considered and was not in the final model.

The association of blood pressure variability as assessed by AUC with 30-day mortality, derived from all SBP ranges evaluated based on logistic regression analysis, is presented in Table 3. AUC was found to be significantly associated with mortality at 30 days after cardiac surgery beginning with the SBP ranges of 75 to 135 mm Hg intraoperatively and 85 to 145 mm Hg pre- and postoperatively, up to the narrowest SBP ranges of 105 to 135 mm Hg intraoperatively and 115 to 145 mm Hg pre- and postoperatively. A significant association between AUC and 30-day mortality was not found for the ECLIPSE-prespecified SBP range of 65 to 135 mm Hg intraoperatively and 75 to 145 mm Hg pre- and postoperatively.

Table 3
Table 3
Image Tools

The logistic regression model shown in Table 4 summarizes the association between postoperative 30-day mortality and AUC based on the SBP range of 75 to 135 mm Hg intraoperatively and 85 to 145 mm Hg pre- and postoperatively, controlling for other potential risk factors. The c-index measuring the concordance for the model was 0.832. The Hosmer-Lemeshow goodness-of-fit test was also performed on the model and showed a P value of 0.7334, indicating that the logistic regression model in Table 4 provides a good fit for the data. The association between postoperative 30-day mortality and AUC is characterized by an odds ratio of 1.16 (95% confidence interval, 1.04–1.30) per incremental AUC value of 60 mm Hg × min/h; for example, equal to an SBP excursion of 1 mm Hg either above or below the range every minute for 60 minutes. The relationship between risk of 30-day mortality and given magnitudes of AUC based on this model is shown in Figure 3, which presents odds ratios and 95% confidence intervals estimated using the model for 5 different AUC values ranging from 60 to 300 mm Hg × min/h.

Table 4
Table 4
Image Tools
Figure 3
Figure 3
Image Tools

Kaplan-Meier survival curves for estimated mortality as time to event (Fig. 4) support the association between blood pressure variability as assessed by AUC (SBP range: 75–135 mm Hg intraoperatively and 85–145 mm Hg pre- or postoperatively) and risk of 30-day mortality demonstrated by the logistic regression model in Table 4. The AUC quartiles in the Kaplan-Meier analysis were based on distribution of AUC values in the pooled ECLIPSE population and were defined as follows: first quartile group = AUC <0.85 mm Hg × min/h, second quartile group = AUC from ≥0.86 to <9.44 mm Hg × min/h, third quartile group = AUC from ≥9.44 to <38.41 mm Hg × min/h, and fourth quartile group = AUC ≥38.41 mm Hg × min/h. Higher AUC quartiles were associated with higher probabilities of mortality. The log-rank test was performed and demonstrated a difference among all 4 quartile groups (P = 0.0187).

Figure 4
Figure 4
Image Tools
Back to Top | Article Outline
High-Risk and Low-Risk Patient Profiles and Predicted Mortality

The logistic regression model included the following risk variables: increasing age, increasing preoperative serum creatinine ≥1.2 mg/dL, history of COPD, decreasing preoperative hemoglobin, preoperative SBP >160 mm Hg or DBP >105 mm Hg, additional procedures during the index surgical procedure, and prolonging surgical duration. These variables were used to define high-risk and low-risk patient profiles (described in the previous section). The absolute amount of increase in the predicted risk of mortality due to blood pressure variability was more substantial for patients with the high-risk profile. Figure 5 shows the difference in the predicted probability of 30-day mortality between the high- and low-risk patients for the same level of SBP variability based on the logistic regression model. As expected, the high-risk patient profile demonstrates much higher predicted mortality. The predicted probability of 30-day mortality for high-risk patients increases from 42.4% to 60.7%, whereas for low-risk patients, a small increase was predicted from 0.2% to 0.5% if AUC increased from 0 to 300 mm Hg × min/h.

Figure 5
Figure 5
Image Tools
Back to Top | Article Outline

DISCUSSION

The present data demonstrate an association between perioperative blood pressure variability (characterized as excursions outside an SBP range) and 30-day mortality after cardiac surgery.

Recent studies of ambulatory blood pressure indicate that blood pressure variability2123 and treatment effects on blood pressure variability24,25 are associated with the risk of stroke and other cardiovascular outcomes. These studies support the use of both blood pressure magnitude and variability in patient assessment.26,27 The findings of variability in ambulatory blood pressure management may also apply to the perioperative setting. Blood pressure variability is common in the perioperative time period,6,11,28 and it is unclear to what extent the actual magnitude or the duration of blood pressure abnormalities may contribute separately to postoperative risk. Our analysis captured both aspects for SBP outside the range, both above and below the defined ranges.

Early studies showed associations of intraoperative hemodynamic abnormalities and increased morbidity and mortality,2931 but were often limited by small numbers of patients, heterogeneous patient populations and surgical procedures, varying and increasingly outdated definitions of hypertension, retrospective study design, and no predefined criteria for outcomes of subgroup or retrospective analysis. More recent studies of cardiac and noncardiac surgery have shown associations between poor outcomes and systolic hypotension, low mean arterial blood pressure (MAP), large fluctuations in MAP, systolic hypertension, history of hypertension, or pulmonary hypertension.3239

The challenge these findings present to the clinician is the wide range of perioperative risk variables, patients and surgical procedures evaluated, the lack of consistently replicated results among studies, and the lack of definitive evidence that specific therapeutic targets or ranges minimize patient risk. Preventing excessive hypotension, which may lead to hypoperfusion and ischemia, is clearly important. Acute perioperative hypertension is characterized by excessive catecholamine release, peripheral vasoconstriction, and reduced baroreceptor sensitivity7 with consequent increase in myocardial oxygen consumption, left ventricular end-diastolic pressure, and increased surgical bleeding from anastomotic sites.6,40,41 Perioperative hypertension has also been reported to worsen reperfusion injury, humoral and cellular inflammatory response, and platelet activation, which may compromise microvascular blood flow as well as contribute to subendocardial hypoperfusion and myocardial ischemia in vulnerable patients.6,7,11,41

The findings of the present study are consistent with results we have recently reported42 from an independent, risk-equivalent dataset using the Duke Clinical Database. The Duke study demonstrated an association between SBP variability and 30-day mortality in 5038 consecutive patients undergoing primary, nonemergent CABG surgery, with the mean duration of systolic excursion (outside a range of 105–130 mm Hg) found to be the most predictive of 30-day mortality (odds ratio of 1.03 per minute; 95% confidence interval, 1.02–1.39; P < 0.0001). The Duke study used electronic blood pressure data recorded every 30 seconds intraoperatively after arterial line placement in the operating room, and analyzed SBP excursions outside the range by number, duration, and magnitude of episodes as well as by magnitude × duration of episode (AUC).

The present analysis is based on a pooled patient population from the ECLIPSE comparative trials of antihypertensive safety. The lower limits of the ECLIPSE protocol-specified SBP ranges (65 mm Hg intraoperatively and 75 mm Hg pre- and postoperatively) were primarily selected for purposes of demonstrating patient safety in the 3 ECLIPSE trials. It should be noted that these ranges were not used as therapeutic targets during the trials, but were prespecified for analysis purposes for each individual trial. In the present pooled analysis, the protocol SBP ranges may not have been sensitive enough to capture SBP variability, probably because they failed to capture some of the lower SBP levels as excursions outside the range. Narrowing the ranges by raising the lower SBP limits by 10 mm Hg increments revealed a significant association between blood pressure variability and 30-day mortality. The Kaplan-Meier estimated survival curves provide this context visually, demonstrating increasing mortality with increasing AUC; the purpose of this analysis was not to demonstrate comparative differences between any 2 adjacent quartiles.

In this analysis, we chose to evaluate perioperative blood pressure variability by measuring SBP, among other hemodynamic variables, and by performing multiple logistic regression analysis of AUC. SBP represents a more sensitive index for adverse outcomes than DBP or MAP in the ambulatory setting43,44 and the perioperative period overall.45 Further studies could utilize other indices in addition to SBP, such as pulse pressure and MAP.

Total AUC characterized variability by incorporating duration of SBP excursions as well as magnitude, and hypertensive as well as hypotensive excursions outside the range. The use of this parameter captured multiple aspects of SBP variation that may make a simultaneous contribution to patient risk and was consistent with the surgical practice of maintaining blood pressure within a therapeutically acceptable range. The use of combined hypertensive and hypotensive SBP excursions in a single risk variable may be considered a limitation of the present study, because it is assumed that their effects on postoperative mortality are always in the same direction. Nevertheless, the distribution of hypotensive and hypertensive excursions outside the range and postoperative mortality among study patients suggests that a combined influence from both types of excursions may be at work.

The original ECLIPSE trials are among the largest comparative, prospective studies evaluating the safety of antihypertensive treatments in the clinical setting of cardiac surgery. The present analysis is probably not powered sufficiently to demonstrate predictive risk by treatment groups for risk of postoperative 30-day mortality. Similarly, although the subgroup of patients undergoing combined CABG and valve surgery had a higher incidence of mortality in our patient sample as would be expected, the risk variable of combined CABG and valve surgery was not found to be significant in the final logistic regression model in our analysis.

Limitations of this study include the lack of a control group (e.g., hypertensive patients who were not treated with IV antihypertensive therapy, and/or nonhypertensive patients). However, ethical considerations precluded the former type of control, and the latter type was not included because of the design of the original ECLIPSE trials as a safety assessment program for active treatment. A control group of nonhypertensive patients presumably would have shown the lowest total AUC values and lowest risk of 30-day mortality in our analysis, but that assumes a lack of other comorbidities and may not consider the potential volatility of blood pressure in the perioperative period. The comparison of low- and high-risk patients in the present analysis, and of patients by AUC quartile, may be helpful in the absence of a control group. Another limitation is the low frequency at which blood pressure was obtained, specifically measurements of once per hour from postoperative hour 6 through 24. Although a narrower time interval would have been desirable and possibly more illuminating, the value of these data in demonstrating an association between blood pressure variability and postoperative mortality is evident.

The exact site for intraarterial monitoring was not prespecified in the study protocols because this was considered impractical for cardiac surgery, but may represent a study limitation. For example, radial artery pressure measurements may vary compared with other arterial sites. It is unclear to what extent this would have a clinically meaningful impact for the surgical patient, especially given limited data in the literature.46 Additional potential limitations include the analysis of combined data from patients undergoing CABG and/or valve surgery and the potential bias inherent in transcribing digital blood pressure data. Our analysis did not make use of risk prediction models in cardiac surgery such as the Society of Thoracic Surgeons mortality risk score or the European System for Cardiac Operative Risk Evaluation. The ECLIPSE trials, which focused on safety, were not designed to collect patient characteristics data to the extent that would have been required for these models.

The statistical preference for only one-tenth the number of predictors in a logistic regression model as outcomes, or a minimum of 10 events per predictor variable, is based on a 1996 simulation study.47 However, a more recent study,48 also based on simulation, concluded that the rule may be too conservative and estimated that 5 to 9 events per predictor variable would be acceptable. Our analysis is within this range at slightly >5 events per predictor variable, a favorable result in terms of the more recent simulation.

In summary, our data demonstrate an association between SBP variability as assessed by excursions outside perioperative SBP ranges and 30-day mortality. The validity of this post hoc analysis must be confirmed by prospective study. However, these findings indicate that further studies are warranted, which may in the future address mechanistic relationships and evaluate the implications of blood pressure variability outside of cardiac surgery.

Back to Top | Article Outline

RECUSE NOTE

Jerrold H. Levy is the Cardiovascular Section Editor of Hemostasis and Transfusion Medicine for the Journal. This manuscript was handled by Martin J. London, Section Editor of Perioperative Echocardiography and Cardiovascular Education, and Dr. Levy was not involved in any way with the editorial process or decision.

Back to Top | Article Outline

DISCLOSURES

Name: Solomon Aronson, MD.

Contribution: This author conceived and designed the study, gathered and analyzed the ECLIPSE data, wrote the manuscript, and shared in the decision of all authors to publish the paper.

Attestation: Solomon Aronson 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.

Conflicts of Interest: Solomon Aronson received honoraria from The Medicines Company and consulted for The Medicines Company consultant fees and/or honoraria related to advisory activities from The Medicines Company.

Name: Cornelius M. Dyke, MD.

Contribution: This author gathered the ECLIPSE data and shared in the decision of all authors to publish the paper.

Attestation: Cornelius M. Dyke has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Conflicts of Interest: Cornelius M. Dyke received honoraria from The Medicines Company and consulted for The Medicines Company consultant fees and/or honoraria related to advisory activities from The Medicines Company.

Name: Jerrold H. Levy, MD.

Contribution: This author gathered the ECLIPSE data and shared in the decision of all authors to publish the paper.

Attestation: Jerrold H. Levy has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Conflicts of Interest: Jerrold H. Levy received honoraria from The Medicines Company, consulted for The Medicines Company, and received research support consultant fees and/or honoraria related to advisory activities and grant support related to research activities from The Medicines Company.

Name: Albert T. Cheung, MD.

Contribution: This author gathered the ECLIPSE data and shared in the decision of all authors to publish the paper.

Attestation: Albert T. Cheung has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Conflicts of Interest: Albert T. Cheung received research support from The Medicines Company and grant support from the University of Pennsylvania related to research activities from The Medicines Company.

Name: Philip D. Lumb, MB, BS.

Contribution: This author gathered the ECLIPSE data and shared in the decision of all authors to publish the paper.

Attestation: Philip D. Lumb has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Conflicts of Interest: Philip D. Lumb has no conflicts of interest.

Name: Edwin G. Avery, MD.

Contribution: This author gathered the ECLIPSE data and shared in the decision of all authors to publish the paper.

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

Conflicts of Interest: Edwin G. Avery received honoraria from The Medicines Company and consulted for The Medicines Company consultant fees and/or honoraria related to advisory activities from The Medicines Company.

Name: Ming-yi Hu, PhD.

Contribution: This author analyzed the ECLIPSE data and shared in the decision of all authors to publish the paper.

Attestation: Ming-yi Hu 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.

Conflicts of Interest: Ming-yi Hu is employed by The Medicines Company.

Name: Mark F. Newman, MD.

Contribution: This author gathered the ECLIPSE data and shared in the decision of all authors to publish the paper.

Attestation: Mark F. Newman has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Conflicts of Interest: Mark F. Newman received research support from The Medicines Company and financial support from DCRI for work contracted by The Medicines Company.

Back to Top | Article Outline

ACKNOWLEDGMENTS

The authors recognize the work of the GPRO study group/ ECLIPSE investigators in the conduct of the primary ECLIPSE trials.

a The term AUC is used for convenience although the areas under consideration are not always under the curve. However, the method for computing area is the same whether above or below the curve. In the present study, AUC is the total area bounded by the SBP–time curve, the upper or lower SBP limit, and the starting and ending SBP measurement times, computed by summation of cumulative SBP excursions outside a defined SBP range and normalized by hour. See Figure 1. Cited Here...

Back to Top | Article Outline
APPENDIX
Continuous and Categorical Risk Variables Investigated for Association with 30-Day Mortality Before Logistic Regression Model Selection Processes
Back to Top | Article Outline
Treatment Group

1. Randomized treatment: clevidipine, comparators

2. Randomized treatment, 4 groups: clevidipine, nicardipine, nitroglycerin, sodium nitroprusside

Back to Top | Article Outline
Demographics

3. Age ≥65 years: no, yes

4. Age (years): continuous

5. Sex: female, male

6. Race: African American, other, Caucasian

7. African American: no, yes

8. Caucasian: no, yes

9. Weight (kg): continuous

10. Height (cm): continuous

11. Body mass index ≥25 kg/m2?: no, yes

12. Body mass index: continuous

Back to Top | Article Outline
Preoperative Hemodynamic Results

13. Quartile of screening systolic blood pressure: first quartile, second quartile, third quartile, fourth quartile

14. Screening systolic blood pressure: continuous

15. Quartile of screening diastolic blood pressure: first quartile, second quartile, third quartile, fourth quartile

16. Screening diastolic blood pressure: continuous

17. Screening systolic blood pressure >160 mm Hg or diastolic blood pressure >105 mm Hg: no, yes

Back to Top | Article Outline
Clinical Laboratory Results

18. Preoperative creatinine clearance group: >0–30, >30–50, >50–80, >80, unknown

19. Preoperative creatinine clearance: continuous

20. Preoperative creatinine clearance ≤50 mL/min: no, yes

21. Preoperative serum creatinine (mg/dL): continuous

22. Preoperative serum creatinine ≥1.2 mg/dL: no, yes

23. Quartile of preoperative hemoglobin: first quartile, second quartile, third quartile, fourth quartile

24. Preoperative hemoglobin (g/dL): continuous

Back to Top | Article Outline
Medical History

25. Medical history—prior coronary artery bypass graft (CABG): no, yes

26. Medical history—prior percutaneous coronary intervention: no, yes

27. Medical history—family history of coronary artery disease: no, yes

28. Medical history—hypertension: no, yes

29. Medical history—congestive heart failure: no, yes

30. Medical history—dyslipidemia: no, yes

31. Medical history—diabetes: unknown, yes

32. Medical history—prior atrial fibrillation (AF)/flutter: no, yes

33. Medical history—transient ischemic attack: no, yes

34. Medical history—recent myocardial infarction (<6 months): no, yes

35. Medical history—angina pectoris: no, yes

36. Medical history—peripheral vascular disease: no, yes

37. Medical history—current smoker: no, yes

38. Medical history—chronic obstructive pulmonary disease: no, yes

39. Previous CABG: yes

40. Previous valve repair: yes

Back to Top | Article Outline
Procedural Characteristics

41. Procedure type: 1. primary off-pump CABG, 2. primary on-pump CABG, 3. primary valve (single), 4. primary CABG + valve (single), 5. redo CABG, 6. primary valve (double), 7. redo valve (single), 8. redo CABG + valve (single), 9. primary other

42. Primary versus redo: primary procedure, redo procedure

43. CABG versus non-CABG: CABG, non-CABG

44. Primary CABG versus others: other procedure, primary CABG

45. Redo CABG + valve: no, yes

46. Any redo valve: no, yes

47. Double valve or any redo valve: no, yes

48. Double valve or any redo: no, yes

49. Two or more procedures previously or currently: no, yes

50. Any valve or redo CABG: no, yes

51. On-pump CABG or any valve: no, yes

52. CABG in this procedure: yes

53. Off-pump CABG in this procedure: yes

54. Valve replacement or repair: no, yes

55. Experienced AF at the time of surgery?: no, yes

56. Additional procedures during surgery: no, yes

57. Quartile of surgery duration: first quartile, second quartile, third quartile, fourth quartile

58. Duration of surgery (hours): continuous

59. Quartile of on-pump duration: first quartile, second quartile, third quartile, fourth quartile

60. Duration of on pump (hours): continuous

61. On pump?: no, yes

Notes: Based on the results of χ2 test or Fisher exact test, as appropriate, for categorical variables, and a t test or Wilcoxon rank sum test for continuous variables, a nominal 2-sided P value ≤0.15 was required for entry of variables into independent stepwise, backward, and forward model selection processes as part of logistic regression analysis. Because the focus of the analysis was association of area under the curve (AUC) with mortality and it was a required part of logistic regression analysis, AUC is not listed as a candidate variable for entry into model selection processes.

Back to Top | Article Outline
Continuous and Categorical Risk Variables Entered into Logistic Regression Model Selection Processes (AUC Based on Systolic Blood Pressure Range of 65 to 135 mm Hg Intraoperatively and 75 to 145 Preoperatively or Postoperatively)
Back to Top | Article Outline
Demographics

1. Age (years): continuous

2. Body mass index: continuous

Back to Top | Article Outline
Preoperative Hemodynamic Results

3. Screening systolic blood pressure: continuous

4. Screening systolic blood pressure >160 mm Hg or diastolic blood pressure >105 mm Hg: no, yes

Back to Top | Article Outline
Clinical Laboratory Results

5. Preoperative creatinine clearance: continuous

6. Preoperative creatinine clearance ≤50 mL/min: no, yes

7. Preoperative serum creatinine ≥1.2 mg/dL: no, yes

8. Preoperative hemoglobin (g/dL): continuous

Back to Top | Article Outline
Medical History

9. Medical history—prior percutaneous coronary intervention: no, yes

10. Medical history—hypertension: no, yes

11. Medical history—congestive heart failure: no, yes

12. Medical history—prior AF/flutter: no, yes

13. Medical history—recent myocardial infarction (<6 months): no, yes

14. Medical history—peripheral vascular disease: no, yes

15. Medical history—chronic obstructive pulmonary disease: no, yes

Back to Top | Article Outline
Procedural Characteristics

16. Two or more procedures previously or currently: no, yes

17. Any valve or redo CABG: no, yes

18. Additional procedures during surgery: no, yes

19. Valve replacement or repair: no, yes

20. Experienced AF at the time of surgery?: no, yes

21. Additional procedures during surgery: no, yes

22. Duration of surgery (hours): continuous

Note: A P value <0.05 was required for inclusion of variables in the final logistic regression model. Because the focus of the analysis was association of AUC with mortality and it was a required part of logistic regression analysis, AUC is not included in this list of risk variables. Cited Here...

Back to Top | Article Outline

REFERENCES

1. Aronson S, Fontes ML, Miao Y, Mangano DT; Investigators of the Multicenter Study of Perioperative Ischemia Research Group, Ischemia Research and Education Foundation. Risk index for perioperative renal dysfunction/failure: critical dependence on pulse pressure hypertension. Circulation 2007;11:733–42

2. Lawes CM, Bennett DA, Feigin VL, Rodgers A. Blood pressure and stroke: an overview of published reviews. Stroke 2004;35:1024–33

3. Roach GW, Kanchuger M, Mangano CM, Newman M, Nussmeier N, Wolman R, Aggarwal A, Marschall K, Graham SH, Ley C. Adverse cerebral outcomes after coronary bypass surgery. Multicenter Study of Perioperative Ischemia Research Group and the Ischemia Research and Education Foundation Investigators. N Engl J Med 1996;335:1857–63

4. Charlson ME, MacKenzie CR, Gold JP, Ales KL, Topkins M, Shires GT. Intraoperative blood pressure: what patterns identify patients at risk for postoperative complications? Ann Surg 1990;212:567–80

5. McCloskey G, Barash PG. Hypertension and coronary artery bypass surgery. Anesthesiol Clin North America 1999;17:593–607

6. Haas CE, LeBlanc JM. Acute postoperative hypertension: a review of therapeutic options. Am J Health Syst Pharm 2004;61:1661–73

7. Leslie JB. Incidence and aetiology of perioperative hypertension. Acta Anaesthesiol Scand Suppl 1993;99:5–9

8. Howell SJ, Sear JW, Foëx P. Hypertension, hypertensive heart disease and perioperative cardiac risk. Br J Anaesth 2004;92:570–83

9. Rosamond W, Flegal K, Furie K, Go A, Greenlund K, Haase N, Hailpern SM, Ho M, Howard V, Kissela B, Kittner S, Lloyd-Jones D, McDermott M, Meigs J, Moy C, Nichol G, O'Donnell C, Roger V, Sorlie P, Steinberger J, Thom T, Wilson M, Hong Y; American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statistics: 2008 update—a report from the American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Circulation 2008;117:e25–146

10. Hajjar I, Kotchen TA. Trends in prevalence, awareness, treatment, and control of hypertension in the United States, 1988–2000. JAMA 2003;290:199–206

11. Cheung AT. Exploring an optimum intra/postoperative management strategy for acute hypertension in the cardiac surgery patient. J Card Surg 2006;21:S8–14

12. Feneck R. Drugs for the perioperative control of hypertension: current issues and future directions. Drugs 2007;67:2023–44

13. Fontes ML, Aronson S, Mathew JP, Miao Y, Drenger B, Barash PG, Mangano DT; Multicenter Study of Perioperative Ischemia (McSPI) Research Group; Ischemia Research and Education Foundation (IREF) Investigators. Pulse pressure and risk adverse outcome in coronary bypass surgery. Anesth Analg 2008;107:1122–9

14. Immink RV, van den Born BJ, van Montfrans GA, Koopmans RP, Karemaker JM, van Lieshout JJ. Impaired cerebral autoregulation in patients with malignant hypertension. Circulation 2004;110:2241–5

15. Toyoda K, Fujii K, Ibayashi S, Kitazono T, Nagao T, Takaba H, Fujishima M. Attenuation and recovery of brain stem autoregulation in spontaneously hypertensive rats. J Cereb Blood Flow Metab 1998;18:305–10

16. Palmer BF. Renal dysfunction complicating the treatment of hypertension. N Engl J Med 2002;347:1256–61

17. Aronson S, Dyke CM, Stierer KA, Levy JH, Cheung AT, Lumb PD, Kereiakes DJ, Newman MF. The ECLIPSE trials: comparative studies of clevidipine to nitroglycerin, sodium nitroprusside, and nicardipine for acute hypertension treatment in cardiac surgery patients. Anesth Analg 2008;107:1110–21

18. Center for Drug Evaluation and Research (CDER), Food and Drug Administration (FDA), US Department of Health and Human Services. Guidance for Industry. E6 Good Clinical Practice: Consolidated Guidance. Federal Register , 1997, 62FR 25692

19. World Medical Association General Assembly. World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. J Int Bioethique 2004;15:124–9

20. Epperson JF. An Introduction to Numerical Methods and Analysis. Rev. ed. Hoboken, NJ: John Wiley & Sons, 2007

21. Stolarz-Skrzypek K, Thijs L, Richart T, Li Y, Hansen TW, Boggia J, Kuznetsova T, Kikuya M, Kawecka-Jaszcz K, Staessen JA. Blood pressure variability in relation to outcome in the International Database of Ambulatory Blood Pressure in Relation to Cardiovascular Outcome. Hypertens Res 2010;33:757–66

22. Rothwell PM, Howard SC, Dolan E, O'Brien E, Dobson JE, Dahlöf B, Sever PS, Poulter NR. Prognostic significance of visit-to-visit variability, maximum systolic blood pressure, and episodic hypertension. Lancet 2010;375:895–905

23. Brickman AM, Reitz C, Luchsinger JA, Manly JJ, Schupf N, Muraskin J, DeCarli C, Brown TR, Mayeux R. Long-term blood pressure fluctuation and cerebrovascular disease in an elderly cohort. Arch Neurol 2010;67:564–9

24. Rothwell PM, Howard SC, Dolan E, O'Brien E, Dobson JE, Dahlöf B, Poulter NR, Sever PS, on behalf of the ASCOT-BPLA and MRC Trial Investigators. Effects of β blockers and calcium-channel blockers on within-individual variability in blood pressure and risk of stroke. Lancet Neurol 2010;9:469–80

25. Webb AJS, Fischer U, Mehta Z, Rothwell RM. Effects of antihypertensive-drug class on interindividual variation in blood pressure and risk of stroke: a systematic review and meta-analysis. Lancet 2010;375:906–15

26. Rothwell PM. Limitations of the usual blood-pressure hypothesis and importance of variability, instability, and episodic hypertension. Lancet 2010;375:938–48

27. Carlberg B, Lindholm LH. Stroke and blood-pressure variation: new permutations on an old theme. Lancet 2010;375:867–9

28. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL Jr, Jones DW, Materson BJ, Oparil S, Wright JT Jr, Roccella EJ; Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. National Heart, Lung, and Blood Institute; National High Blood Pressure Education Program Coordinating Committee. Seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. Hypertension 2003;42:1206–52

29. Howell SJ, Sear JW, Foëx P. Hypertension, hypertensive heart disease and perioperative cardiac risk. Br J Anaesth 2004;92:570–83

30. Gal TJ, Cooperman LH. Hypertension in the immediate postoperative period. Br J Anaesth 1975;47:70–4

31. Seltzer JL, Gerson JI, Grogono AW. Hypertension in perioperative period. NY State J Med 1980;80:29–31

32. Jain U, Laflamme CJ, Aggarwal A, Ramsay JG, Comunale ME, Ghoshal S, Ngo L, Ziola K, Hollenberg M, Mangano DT. Electrocardiographic and hemodynamic changes and their association with myocardial infarction during coronary artery bypass surgery: a multicenter study. Multicenter Study of Perioperative Ischemia (McSPI) Research Group. Anesthesiology 1997;86:576–91

33. Reich DL, Bodian CA, Krol M, Kuroda M, Osinski T, Thys DM. Intraoperative hemodynamic predictors of mortality, stroke, and myocardial infarction after coronary artery bypass surgery. Anesth Analg 1999;89:814–22

34. Welsby IJ, Newman MF, Phillips-Bute B, Messier RH, Kakkis ED, Stafford-Smith M. Hemodynamic changes after protamine administration: association with mortality after coronary artery bypass surgery. Anesthesiology 2005;102:308–14

35. Reich DL, Wood RK Jr, Emre S, Bodian CA, Hossain S, Krol M, Feierman D. Association of intraoperative hypotension and pulmonary hypertension with adverse outcomes after orthotopic liver transplantation. J Cardiothorac Vasc Anesth 2003;17:699–702

36. Reich DL, Hossain S, Krol M, Baez B, Patel P, Bernstein A, Bodian CA. Predictors of hypotension after induction of general anesthesia. Anesth Analg 2005;101:622–8

37. Monk TG, Saini V, Weldon BC, Sigl JC. Anesthetic management and one-year mortality after noncardiac surgery. Anesth Analg 2005;100:4–10

38. Reich DL, Bennett-Guerrero E, Bodian CA, Hossain S, Winfree W, Krol M. Intraoperative tachycardia and hypertension are independently associated with adverse outcome in noncardiac surgery of long duration. Anesth Analg 2002;95:273–7

39. Ganushchak YM, Fransen EJ, Visser C, De Jong DS, Maessen JG. Neurological complications after coronary artery bypass grafting related to the performance of cardiopulmonary bypass. Chest 2004;125:2196–205

40. Weiss SJ, Longnecker DE. Perioperative hypertension: an overview. Coron Artery Dis 1993;4:401–6

41. Erstad BL, Barletta JF. Treatment of hypertension in the perioperative patient. Ann Pharmacother 2000;34:66–79

42. Aronson S, Stafford-Smith M, Phillips-Bute B, Shaw A, Gaca J, Newman M. Intraoperative systolic blood pressure variability predicts 30-day mortality in aortocoronary bypass surgery patients. Anesthesiology 2010;113:305–12

43. Staessen JA, Gasowski J, Wang JG, Thijs L, Hond ED, Boissel J-P, Coope J, Ekbom T, Gueyffier F, Liu L, Kerlikowske K, Pocock S, Fagard RH. Risks of untreated and treated isolated systolic hypertension in the elderly: meta-analysis of outcome trials. Lancet 2000;355:865–72

44. Kengne AP, Czernichow S, Huxley R, Grobbee D, Woodward M, Neal B, Zoungas S, Cooper M, Glasziou P, Hamet P, Harrap SB, Mancia G, Poulter N, Williams B, Chalmers J, on behalf of the ADVANCE Collaborative Group. Blood pressure variables and cardiovascular risk: new findings from ADVANCE. Hypertension 2009;54:399–404

45. Aronson S, Boisvert D, Lapp W. Isolated systolic hypertension is associated with adverse outcomes from coronary artery bypass grafting surgery. Anesth Analg 2002;94:1079–84

46. Howell MD, Curley FJ, Smyrnios NA. Routine monitoring of critically ill patients. In: Irwin RS, Rippe JM, Lisbon A, Heard SO, eds. Procedures, Techniques, and Minimally Invasive Monitoring in Intensive Care Medicine. 4th ed. Philadelphia: Lippincott Williams & Wilkins, 2008:269–89

47. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996;49:1373–9

48. Vittinghoff E, McCulloch CE. Relaxing the rule of ten events per variable in logistic and Cox regression. Am J Epidemiol 2007;165:710–8

© 2011 International Anesthesia Research Society

Login

Become a Society Member