Free
Perioperative Medicine  |   August 2010
Intraoperative Systolic Blood Pressure Variability Predicts 30-day Mortality in Aortocoronary Bypass Surgery Patients
Author Affiliations & Notes
  • Solomon Aronson, M.D.
    *
  • Mark Stafford-Smith, M.D.
    *
  • Barbara Phillips-Bute, Ph.D.
  • Andrew Shaw, M.D.
  • Jeffrey Gaca, M.D.
    §
  • Mark Newman, M.D.
    *
  • * Professor, † Assistant Professor, ‡ Associate Professor, Department of Anesthesiology, § Assistant Professor, Department of Surgery, Cardiovascular Division, Duke University Medical Center, Durham, North Carolina. ∥ See 1for members of Cardiothoracic Anesthesiology Research Endeavors.
Article Information
Perioperative Medicine / Cardiovascular Anesthesia
Perioperative Medicine   |   August 2010
Intraoperative Systolic Blood Pressure Variability Predicts 30-day Mortality in Aortocoronary Bypass Surgery Patients
Anesthesiology 8 2010, Vol.113, 305-312. doi:10.1097/ALN.0b013e3181e07ee9
Anesthesiology 8 2010, Vol.113, 305-312. doi:10.1097/ALN.0b013e3181e07ee9
What We Already Know about This Topic
  • ❖ Blood pressure variability during surgery has been associated with postoperative adverse events
  • ❖ These studies included a small number of patients and intermittent manual recording of blood pressure
What This Article Tells Us That Is New
  • ❖ In electronically captured blood pressure recording in more than 7,500 consecutive cardiac surgery patients, blood pressure variability was independently associated with an increased 30-day mortality, and although the absolute increased risk was small (3% for systolic blood pressure outside 105–130 mmHg), intraoperative systolic blood pressure variability may represent an important modifiable risk factor for 30-day mortality
FOR nearly a century, arterial blood pressure has been adopted as the cornerstone for clinical evaluation of cardiovascular function and risk. Poorly controlled blood pressure represents a major risk factor for cardiac, renal, cerebral, and metabolic dysfunction outcomes.1–3 
Hypertension, as a preexisting condition, exists in more than two thirds of the elderly surgical population in the United States, with upward of 80% prevalence in those requiring cardiac surgery,1,2,4–6 and remains a significant contributor to postoperative risk.4–10 
Acute intraoperative changes in blood pressure, although a common consequence of preexisting hypertension, may also be a consequence of excessive release of catecholamines, rapid intravascular volume shifts, peripheral vasoconstriction, reduced baroreceptor sensitivity, renin-angiotensin activation, altered cardiac reflexes, inadequate anesthesia, reperfusion injury, aortic occlusive clamps, as well as neural, humoral, and cellular responses.11 These factors, separately or in combination, could trigger hyperinflammatory and procoagulation conditions,12 including platelet activation,13 which may compromise microvascular blood flow. In addition, perioperative hypertension increases myocardial oxygen consumption, increases left ventricular end-diastolic pressure, and increases the risk of surgical bleeding from anastamotic sites,11 whereas perioperative hypotension contributes to subendocardial hypoperfusion and myocardial ischemia.
Despite what has been learned about characterization of blood pressure subtypes, associated risk of hypertension or hypotension, autoregulation shifts, and efficacy of therapy,4,9,10,14–16 few data have been presented regarding optimal intraoperative blood pressure, with treatment guidelines poorly defined and nearly untested in this setting.
An association between perioperative systolic blood pressure bandwidth (in other words, systolic blood pressure excursions outside a target systolic blood pressure range; e.g.  , systolic blood pressure higher than 135 or less than 65 mmHg) and 30-day adverse events has been found.1Data demonstrating increased therapeutic differentiation (i.e.  , drug differences, based on class and or pharmacokinetics, to control blood pressure within a target range) as blood pressure bandwidth narrowed (e.g.  , higher than 135 or less than 95 mmHg) have also been reported.17 However, these data were limited because of a relatively small sample size, reliance on manual recording of blood pressure, infrequent blood pressure capture frequency, and post hoc  analysis. Here we test a large cohort of similar cardiac surgical patients, using nearly continuous electronic data that capture the association between intraoperative systolic blood pressure variability as a surrogate of blood pressure control and 30-day postoperative mortality.
Materials and Methods
Study Population
With Institutional Review Board approval (Duke University Medical Center, Durham, North Carolina), we retrospectively studied a cohort of 7,808 consecutive patients undergoing primary nonemergent aorto-coronary bypass surgery procedures performed at Duke University Medical Center between September 1, 1996, and December 30, 2005. Intraoperative blood pressure was captured electronically after arterial line placement every 30 s in all patients. In addition, perioperative data, including 30-day mortality data, were accessed from the prospectively collected Duke Databank for Cardiovascular Diseases. In-hospital complications were classified using the Society of Thoracic Surgeons criteria.2The Databank is prospectively compiled from contemporaneous medical records, custom datasheets, and records of laboratory results. Database quality assurance involves random chart review for data confirmation and assessment of data completeness.
Long-term follow-up data were obtained from the follow-up group of the Duke Clinical Research Institute, which collects annual follow-up mortality data and nonfatal endpoint information for the Duke Databank for Cardiovascular Diseases. The annual surveys collect data on general health, hospitalizations, myocardial infarction, stroke, cardiac procedures, and medication use. Patients are surveyed 6 months after an index visit and yearly thereafter with a mailed, self-administered survey or a phone-administered survey to nonresponders.
Follow-up is 95% complete for mortality, and patients who are lost to follow-up (2%) or who have asked to be withdrawn (3%) are submitted for an annual search of the National Death Index. Death information is collected from next-of-kin interviews, hospital discharge summaries, death certificates, or cause of death provided from the National Death Index. Cause of death is assigned after agreement from independent reviews by a death committee from the Duke Clinical Research Institute.
Study Design and Conduct
For analysis, the study sample was a priori  divided randomly into hypothesis development and validation cohorts using a 66 to 33% sample division. All endpoints were specified in advance of data gathering and analysis. The study was conceived and designed by S.A., the data gathered and analyzed by S.A., M.S.S., A.S., and B.P.B. The manuscript was reviewed by J.G. and M.N.
Blood Pressure Analysis
Intraarterial catheter blood pressure values were recorded in the operating room every 30 s using an automated anesthesia record-keeping system. For purposes of the study, baseline blood pressure was defined as the median of the first five values recorded while the patient was awake and mildly sedated before anesthesia induction.
Development Phase
Independent Variable—Systolic Blood Pressure Variability.
Systolic blood pressure values recorded less than 20 or greater than 299 mmHg were considered erroneous and excluded from analysis. Measurements made during cardiopulmonary bypass were excluded. Based on a priori  evidence,17 the systolic blood pressure variability index area under the curve was assessed. Area under the curve is a continuous variable that reflects an integral of degree and duration of pressure incursions described in units of millimeters of mercury per minute.
  • Three area under the curve variables were calculated for various systolic threshold criteria: area above a value (e.g.  , 135 mmHg), area below a value (e.g.  , 95 mmHg), or the combined areas beyond a range (e.g.  , higher than 135 or less than 95 mmHg).

  • In addition, the isolated systolic blood pressure variables of, magnitude of excursion above or below a criteria blood pressure range, duration of episodes above or below a criteria blood pressure range, and number of episodes above or below a criteria blood pressure range were described separately and tested as outlined below to better understand which blood pressure index most likely predicts adverse postoperative outcome.

  • Magnitude of excursion was characterized as an absolute value by the difference between the criteria threshold (e.g.  , higher than 135 or less than 95 mmHg) and peak (for excursions above) or nadir (for excursions below) systolic blood pressure values. Peak or nadir values were averaged for a subject to determine the mean excursion value (millimeters of mercury). For values beyond a range (e.g.  , higher than 135 or less than 95 mmHg), a mean maximum excursion value was also determined.

  • Duration of excursion variables were described and calculated for various systolic threshold criteria: above a value (e.g.  , 135 mmHg), below a value (e.g.  , 95 mmHg), or the combined duration of excursion beyond a range (e.g.  , higher than 135 or less than 95 mmHg).

  • Number of excursions above (e.g.  , 135 mmHg), below (e.g.  , 95 mmHg), and above or below range criteria (e.g.  , higher than 135 or less than 95 mmHg) was also tested.

Dependent Variable—Postoperative Mortality.
Mortality risk factors evaluated were those identified by Parsonnet et al.  18 in their assessment of preoperative risk factors specific to coronary bypass surgery and coronary bypass surgery/valve surgery. In addition, the parameter of surgery duration was included for analysis as a marker of procedure complexity and risk, because it is directly correlated with factors such as number of coronary grafts performed and is significantly associated with in-hospital mortality.19 Finally, we further assessed the association of systolic blood pressure variability during surgery with outcome by including an investigation of possible confounding effects of aprotonin use.20 
Statistical Analysis
Associations between measures of systolic blood pressure variability and 30-day mortality were investigated using separate multivariable logistic regression models, adjusted for surgery length and Parsonnet score. Because previous research1has suggested that cumulative perioperative blood pressure excursion (area under the curve) beyond a systolic blood pressure range (e.g.  , higher than 135 or less than 95 mmHg) predict adverse outcome after cardiac surgery, our initial analyses sought to validate this association.
In addition, we sought to determine whether other systolic blood pressure index or thresholds also demonstrated a predictive association with 30-day mortality, thus we then explored other characteristics of “systolic blood pressure variability,” as outlined earlier, for their association with mortality (30 day and long term). Candidate criteria for systolic blood pressure variability were those surrounding the confirmed values from the previously published criteria (i.e.  , > 135 or < 95 mmHg).1For the upper bound, thresholds considered were 120, 125, 130, 135, 140, 145, 150, and 180 mmHg. For the lower bound, thresholds considered were 110, 105, 100, 95, 90, 85, 80, and 60 mmHg. Each value was investigated separately in a multivariable model including case length and Parsonnet score. The three most predictive upper and lower blood pressure values were then combined to create nine predictive combined ranges. The final validation analysis was conducted using the best combined predictive systolic blood pressure range. The other systolic blood pressure ranges were compared with the most predictive blood pressure range to determine their relative predictive value for 30-day mortality. In addition, to address the question of whether systolic blood pressure above or below the range is a more powerful predictor of 30-day mortality, we independently evaluated from multivariable logistic regression models mean duration both above and below threshold.
Because of the number of analyses conducted, all findings in the development dataset analysis were considered preliminary only, and no adjustments were made for multiple comparisons. After the single primary hypothesis was tested in the validation dataset, receiver operating characteristic curves from the development and validation samples were plotted for comparison. Randomization of samples was achieved with the use of a random number function to generate a uniform distribution (SAS statistical software version 9.1; SAS Institute, Inc., Cary, NC).
In addition, the relationship of the validation blood pressure criteria with long-term survival was also evaluated using an adjusted Cox proportional hazards analysis. Parsonnet score was included in the survival model. All analyses were performed using SAS statistical software (version 9.1). A value of P  less than 0.01 was considered significant, with a more conservative significance level chosen because of the size of the dataset.
Results
Seven thousand eight hundred eight patients were enrolled and underwent surgery during the period of the study. Three hundred four were excluded for having insufficient pressure data. This sample (n = 7,504) was divided into a development sample of 5,038 and a validation sample of 2,466. The average age was 65 yr; 68% were men; 86% white; and 70% had a history of hypertension, 8% underwent coronary bypass surgery without cardiopulmonary bypass, and 6% underwent combined coronary bypass surgery/valve surgery (table 1). The 30-day mortality was 2.1% (n = 99) in the development sample and 2.6% (n = 62) in the validation sample.
Table 1.  Patient Characteristics
Image not available
Table 1.  Patient Characteristics
×
Our initial hypothesis, that cumulative perioperative blood pressure excursion (area under the curve) beyond a systolic blood pressure range more than 135 or less than 95 mmHg predicted adverse outcome after cardiac surgery, was confirmed (odds ratio [OR]= 1.03 per mmHg min, 95% CI 1.007–1.044, P  < 0. 006; table 2). Additional characteristics of systolic blood pressure variability were calculated for threshold values of 135 and 95 mmHg for each patient and evaluated separately with logistic regression for its ability to contribute to a model in predicting 30-day mortality, after adjusting for length of surgery and Parsonnet score. The characteristic most highly associated with 30-day mortality (identified by P  value and odds ratio) was mean duration of systolic blood pressure excursions beyond threshold values (OR 1.03, 95% CI 1.02–1.39, P  < 0.0001). P  values and odds ratios for the other variables are presented in table 2. The best upper and lower thresholds, as tested using the association of criteria to describe episodes of systolic blood pressure variability with 30-day mortality, were 130, 135,140, and 95, 100, 105 mmHg, respectively. The nine combinations using these thresholds were all evaluated as ranges for their association with 30-day mortality. The upper and lower thresholds were also evaluated independently for their respective associations with 30-day mortality.
Table 2.  Characterizations of Blood Pressure Variability and Association with 30-day Mortality in the Development Sample
Image not available
Table 2.  Characterizations of Blood Pressure Variability and Association with 30-day Mortality in the Development Sample
×
All range combinations were highly associated with 30-day mortality. The systolic blood pressure upper threshold of 130–140 mmHg and the systolic blood pressure lower threshold of 100–105 mmHg demonstrated the strongest odds ratio, P  values, and 95% confidence interval (for the range greater than 130 and less than 105 mmHg, P  < 0.0001, OR 1.028, CI 1.016–1.040 and for the range greater than 140 through less than 105 mmHg, P  < 0.0001, OR 1.028, CI 1.015–1.041). Blood pressure excursions greater than 130 mmHg (P  < 0.005, OR 1.03, 95% CI 1.01–1.05) and less than 95 mmHg (P  < 0.0001, OR 1.03, 95% CI 1.01–1.04) were tested in separate multivariable logistic regression models and each was associated with 30-day mortality (tables 3 and 4).
Table 3.  Multivariable Logistic Regression Model for Mean Duration < 95 mmHg
Image not available
Table 3.  Multivariable Logistic Regression Model for Mean Duration < 95 mmHg
×
Table 4.  Multivariable Logistic Regression Model for Mean Duration > 130 mmHg
Image not available
Table 4.  Multivariable Logistic Regression Model for Mean Duration > 130 mmHg
×
The developed primary hypothesis was that the mean duration of excursions beyond the threshold criteria of the strongest systolic blood pressure range signal, 130/105 mmHg, is associated with 30-day mortality, after adjustment for case length and Parsonnet score. Aprotinin use (10% of patients) showed no significant association with mortality (P  = 0.58), whereas mean duration of excursions outside 130/105 continued to be associated (P  < 0.0001), hence aprotinin use was not included in the final model.
The logistic regression model primary hypothesis test confirmed a strong association (P  = 0.0008, OR 1.03, 95% CI 1.01–1.05). The mean duration of an independent excursion more than 130 mmHg was 5.12 ± 4.6 min and that less than 105 mmHg was 11.06 ± 9.6 min. The mean duration of a combination excursion more than 130 mmHg and less than 105 mmHg was 8.03 ± 5.4 min. There was a direct and a progressive predicted probability of mortality with mean duration of excursion such that for a mean duration of 8 min, the predicted probability of mortality was 1.9%; for mean duration of 10 min, predicted probability of mortality was 2.1%; and for mean duration of 20 min, predicted probability of mortality was 3.5%. The exact magnitude of this effect, per minute increase, can be seen in figure 1. Odds ratios represent an effect for a per-unit change. For example, for the duration variables, which are characterized by minutes outside a range, the odds ratio is per-minute effect. The interpretation is that for every minute outside the range, the odds ratio for 30-day mortality increases by the amount of the odds ratio.
Fig. 1.  Predicted association of mean duration per incursion (minutes) outside of threshold (105–130 mmHg) and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values.
Fig. 1. 
	Predicted association of mean duration per incursion (minutes) outside of threshold (105–130 mmHg) and 30-day mortality in the combined sample (n = 7504). Shaded area 
	represents the 95% confidence intervals for the predicted values.
Fig. 1.  Predicted association of mean duration per incursion (minutes) outside of threshold (105–130 mmHg) and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values.
×
Figure 2portrays the receiver operating characteristic curves generated from the logistic regression models in the development and validation samples. Results from the final validated model are graphically presented using the complete dataset (development and validation together). The c index for the logistic regression model in the development sample (equivalent to the area under the receiver operating characteristic curve) is 0.703. The c-index for the validation sample is 0.79.
Fig. 2.  Receiver operating characteristic curves showing association between duration of incursions and 30-day mortality in the development and validation samples. Results are graphically presented using the complete dataset divided into two parts, development and validation. The c index for the logistic regression model in the development sample (equivalent to the area under the receiver operating characteristic curve) is 0.703. The c index for the validation sample is 0.79. A reference line has been included to illustrate a c index of 0.5, indicating no association.
Fig. 2. 
	Receiver operating characteristic curves showing association between duration of incursions and 30-day mortality in the development and validation samples. Results are graphically presented using the complete dataset divided into two parts, development and validation. The c index for the logistic regression model in the development sample (equivalent to the area under the receiver operating characteristic curve) is 0.703. The c index for the validation sample is 0.79. A reference line has been included to illustrate a c index of 0.5, indicating no association.
Fig. 2.  Receiver operating characteristic curves showing association between duration of incursions and 30-day mortality in the development and validation samples. Results are graphically presented using the complete dataset divided into two parts, development and validation. The c index for the logistic regression model in the development sample (equivalent to the area under the receiver operating characteristic curve) is 0.703. The c index for the validation sample is 0.79. A reference line has been included to illustrate a c index of 0.5, indicating no association.
×
The secondary long-term survival analysis was performed using the whole dataset (combined development and validation). Median follow-up time for survivors was 74 months. One thousand six hundred fifty-three patients died during follow-up time; the remaining patients were censored at the date of last follow-up. Risk-adjusted survival estimates are shown in figs. 3A–Cand show a highly significant association of the mean duration of intraoperative systolic blood pressure variability episodes with short-term and long-term survival (P  < 0.0001; hazard ratio per minute 1.009 [95% CI 1.005–1.012]). An assessment, excluding patients who died within 30 days of surgery, noted similar findings (P  < 0.0001; hazard ratio = 1.009, 95% CI 1.004–1.013).
Fig. 3.  (A  ) Close-up view of the first 30 days of the survival curves, showing risk adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). (B  ) Risk-adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). P  value for the continuous variable is less than 0.0001. (C  ) Risk-adjusted survival estimates using the Cox proportional-hazard method for three mean incursion values outside the range of 105–130 mmHg: 9, 16, and 19 min (25th, 50th, and 75th percentiles). P  value for the continuous variable is less than 0.0001.
Fig. 3. 
	(A 
	) Close-up view of the first 30 days of the survival curves, showing risk adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). (B 
	) Risk-adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). P 
	value for the continuous variable is less than 0.0001. (C 
	) Risk-adjusted survival estimates using the Cox proportional-hazard method for three mean incursion values outside the range of 105–130 mmHg: 9, 16, and 19 min (25th, 50th, and 75th percentiles). P 
	value for the continuous variable is less than 0.0001.
Fig. 3.  (A  ) Close-up view of the first 30 days of the survival curves, showing risk adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). (B  ) Risk-adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). P  value for the continuous variable is less than 0.0001. (C  ) Risk-adjusted survival estimates using the Cox proportional-hazard method for three mean incursion values outside the range of 105–130 mmHg: 9, 16, and 19 min (25th, 50th, and 75th percentiles). P  value for the continuous variable is less than 0.0001.
×
Figures 4A and Bdepict the magnitude of the effect, per minute increase, of blood pressure excursions less than 95 mmHg and more than 130 mmHg, respectively. Figures 5A–Cdepict unadjusted data of 30-day mortality within each tertile of the dataset, based on blood pressure variability (minutes per episode spent outside the range).
Fig. 4.  (A  ) Predicted association of mean duration per incursion (minutes) below the threshold 95 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values. (B  ) Predicted association of mean duration per incursion (minutes) above the threshold 130 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values.
Fig. 4. 
	(A 
	) Predicted association of mean duration per incursion (minutes) below the threshold 95 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area 
	represents the 95% confidence intervals for the predicted values. (B 
	) Predicted association of mean duration per incursion (minutes) above the threshold 130 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area 
	represents the 95% confidence intervals for the predicted values.
Fig. 4.  (A  ) Predicted association of mean duration per incursion (minutes) below the threshold 95 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values. (B  ) Predicted association of mean duration per incursion (minutes) above the threshold 130 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values.
×
Fig. 5.  (A  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent outside the range). (B  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent below the range). (C  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent above the range).
Fig. 5. 
	(A 
	) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent outside the range). (B 
	) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent below the range). (C 
	) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent above the range).
Fig. 5.  (A  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent outside the range). (B  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent below the range). (C  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent above the range).
×
Discussion
This study reports that a strong association exists between intraoperative blood pressure variability and postoperative 30-day mortality in a nearly continuous electronically captured evaluation of blood pressure from a large database among a homogenous subset of patients undergoing cardiovascular revascularization.
Manipulation of blood pressure during cardiac surgery is common because of safety concerns related to ischemia modulation, the need for aortovascular stress-strain modulation (e.g.  , clamping, unclamping), maintaining adequate perfusion conditions during cardiopulmonary bypass, and balancing these pressure–perfusion requirements with surgical bleeding concerns throughout surgery and especially during the postoperative period when requirements for weaning from mechanical ventilation and analgesia are additional stresses for poor blood pressure control.
Although blood pressure abnormalities have been reported to be associated with death, stroke, cognitive and renal dysfunction, perioperative myocardial infarction, and increased mortality,6,7,14,16 many of these studies used a small sample size and lacked statistical power, used a discontinuous data analysis, did not include patient- or surgery-specific characterizations, and were not analyzed to determine specific target blood pressure thresholds and therefore provide limited information on the impact of blood pressure control to postoperative outcome.
In the elderly population with a high prevalence of systolic and pulse pressure hypertension, it has been demonstrated that acute pulsatile stress in conduit vessels cause the elastic elements in the vessel wall to break down, producing vessel dilation and stiffening,21,22 which may contribute to plaque rupture by a mechanical fatiguing effect.21,23 Blood pressure stress and increase in pulsatile load seen in patients with preexisting hypertension may further contribute to endothelial dysfunction.24,25 These mechanical stress induced changes in arterial pathology are uniquely vulnerable to the acute physiologic perturbations seen during surgery and anesthesia. In addition, blood pressure target depends on vascular properties, which include a shift in the autoregulatory range in those persons who have pre-existing disease, which thereby further increases the risk for organ hypoperfusion.
This study presents data on an association of intraoperative blood pressure variability to 30-day mortality in a relatively homogeneous population of patients undergoing coronary revascularization surgery. Although an odds ratio of 1.03 represents a modest effect, it is important (1) to realize that this effect size is per-minute effect, such that for every minute outside the 105–130 mmHg range, the odds ratio for 30-day mortality increases by 0.03 and (2) intraoperative blood pressure is a modifiable risk factor.
We understand that a target systolic blood pressure bandwidth or a “sweet spot” may be different in this population compared with other populations with different conditions or situations (e.g.  , non-cardiac surgery, acute stroke, or other hypertensive emergencies); nevertheless, we believe that these data strongly suggest a need to explore that such a “sweet spot” exists and should be the basis for further investigation.
References
Lewington S, Clarke R, Qizilbash N, Peto R, Collins R; Prospective Studies Collaboration: Age-specific relevance of usual blood pressure to vascular mortality: A meta-analysis of individual data for one million adults in 61 prospective studies. Lancet 2002; 14:1903–13Lewington, S Clarke, R Qizilbash, N Peto, R Collins, R Prospective Studies Collaboration,
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; 42:1206–52Chobanian, AV Bakris, GL Black, HR Cushman, WC Green, LA Izzo, JL Jones, DW Materson, BJ Oparil, S Wright, JT Roccella, EJ Joint National Committee on Prevention,
Ong KL, Cheung BM, Man YB, Lau CP, Lam KS: Prevalence, awareness, treatment, and control of hypertension among United States adults 1999–2004. Hypertension 2007; 49:69–75Ong, KL Cheung, BM Man, YB Lau, CP Lam, KS
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: 13:733–42Aronson, S Fontes, ML Miao, Y Mangano, DT Investigators of the Multicenter Study of Perioperative Ischemia Research Group, Ischemia Research and Education Foundation,
Aronson S, Boisvert D, Lapp W: Isolated systolic hypertension is associated with adverse outcomes from coronary artery bypass grafting surgery. Anesth Analg 2002; 94:1079–84Aronson, S Boisvert, D Lapp, W
Cheung AT: Exploring an optimum intra/postoperative management strategy for acute hypertension in the cardiac surgery patient. Card Surg 2006; 21(suppl 1):S8–14Cheung, AT
Howell SJ, Sear JW, Foëx P: Hypertension, hypertensive heart disease and perioperative cardiac risk. Br J Anaesth 2004; 92:570–83Howell, SJ Sear, JW Foëx, P
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; 95:273–7Reich, DL Bennett-Guerrero, E Bodian, CA Hossain, S Winfree, W Krol, M
Fontes ML, Aronson S, Mathew JP, Miao Y, Drenger B, Barash PG, Mangano DT: Pulse pressure and risk of adverse outcome in coronary bypass surgery. Anesth Analg 2008; 107:1122–9Fontes, ML Aronson, S Mathew, JP Miao, Y Drenger, B Barash, PG Mangano, DT
Benjo A, Thompson RE, Fine D, Hogue CW, Alejo DE, Kaw A, Gerstenblith G, Shah A, Berkowitz DE, Nyhan D: Pulse pressure is an age-independent predictor of stroke development after cardiac surgery. Hypertension 2007; 50:630–5Benjo, A Thompson, RE Fine, D Hogue, CW Alejo, DE Kaw, A Gerstenblith, G Shah, A Berkowitz, DE Nyhan, D
Weiss SJ, Longnecker DE: Perioperative hypertension: An overview. Coron Artery Dis 1993; 4:401–6Weiss, SJ Longnecker, DE
Hennein HA, Ebba H, Rodriguez JL, Merrick SH, Keith FM, Bronstein MH, Leung JM, Mangano DT, Greenfield LJ, Rankin JS. Relationship of the proinflammatory cytokines to myocardial ischemia and dysfunction after uncomplicated coronary revascularization. Thorac Cardiovasc Surg 1994; 108:626–35Hennein, HA Ebba, H Rodriguez, JL Merrick, SH Keith, FM Bronstein, MH Leung, JM Mangano, DT Greenfield, LJ Rankin, JS
Kobzar G, Mardla V, Rätsep I, Samel N: Platelet activity before and after coronary artery bypass grafting. Platelets 2006; 17:289–91Kobzar, G Mardla, V Rätsep, I Samel, N
Gottesman RF, Hillis AE, Grega MA, Borowicz LM, Selnes OA, Baumgartner WA, McKhann GM: Early postoperative cognitive dysfunction and blood pressure during coronary graft operation. Arch Neurol 2007; 64:E1–4Gottesman, RF Hillis, AE Grega, MA Borowicz, LM Selnes, OA Baumgartner, WA McKhann, GM
Verdecchia P, Angeli F: Natural history of hypertension subtypes. Circulation 2005; 111:1094–6Verdecchia, P Angeli, F
Cernaianu AC, Vassilidze TV, Flum DR, Maurer M, Cilley JH Jr, Grosso MA, DelRossi AJ. Predictors of stroke after cardiac surgery. Card Surg 1995; 10:334–9Cernaianu, AC Vassilidze, TV Flum, DR Maurer, M Cilley, JH Grosso, MA DelRossi, AJ
Aronson S, Dyke CM, Stierer KA, Levy JH, Cheung A, Lumb P, Kereiakis DJ, Newman MF: The ECLIPSE Trials: Comparative studies of clevidipine to nitroglycerin, sodium nitroprusside, and nicardipine for acute hypertensive treatment in cardiac surgery patients. Anesth Analg 2008; 107:1110–21Aronson, S Dyke, CM Stierer, KA Levy, JH Cheung, A Lumb, P Kereiakis, DJ Newman, MF The ECLIPSE Trials,
Parsonnet V, Dean D, Bernstein AD: A method of uniform stratification of risk for evaluating the results of surgery in acquired adult heart disease. Circulation 1989; 79 (suppl I):I-3–12Parsonnet, V Dean, D Bernstein, AD
Hill SE, van Wermeskerken GK, Lardenoye JW, Phillips-Bute B, Smith PK, Reves JG, Newman MF: Intraoperative physiologic variables and outcome in cardiac surgery: Part I. In-hospital mortality. Ann Thorac Surg 2000; 69:1070–5Hill, SE van Wermeskerken, GK Lardenoye, JW Phillips-Bute, B Smith, PK Reves, JG Newman, MF
Shaw A, Stafford-Smith M, White W, Phillips-Bute B, Swaminathan M, Milano C, Welsby I, Aronson S, Mathew J, Peterson E, Newman M: The effect of aprotinin on outcome after coronary-artery bypass grafting. N Engl J Med 2008; 21:784–93Shaw, A Stafford-Smith, M White, W Phillips-Bute, B Swaminathan, M Milano, C Welsby, I Aronson, S Mathew, J Peterson, E Newman, M
Dart AM, Kingwell BA. Pulse pressure—a review of mechanisms and clinical relevance. Am Coll Cardiol 2001; 15:975–84Dart, AM Kingwell, BA
Safar ME, Levy BI, Struijker-Boudier H: Current perspectives on arterial stiffness and pulse pressure in hypertension and cardiovascular diseases. Circulation 2003; 10:2864–9Safar, ME Levy, BI Struijker-Boudier, H
Imoto K, Hiro T, Fujii T: Longitudinal structural determinants of atherosclerotic plaque vulnerability: A computational analysis of stress distribution using vessel and three dimensional intravascular ultrasound imaging. Am Coll Cardiol 2005; 46:1507–15Imoto, K Hiro, T Fujii, T
Lee KW, Blann AD, Lip GY: High pulse pressure and nondipping circadian blood pressure in patients with coronary artery disease: Relationship to thrombogenesis and endothelial damage/dysfunction. Am J Hypertens 2004; 18:104–15Lee, KW Blann, AD Lip, GY
Küng CF, Lüscher TF: Different mechanisms of endothelial dysfunction with aging and hypertension in rat aorta. Hypertension 1995; 25:194–200Küng, CF Lüscher, TF
Appendix: Cardiothoracic Anesthesiology Research Endeavors, Duke University Medical Center
Anesthesiology:
Solomon Aronson, M.D., Katherine P. Grichnik, M.D., Steven Hill, M.D., G. Burkhard Mackensen, M.D., Ph.D., Joseph P. Mathew, M.D., Mark F. Newman, M.D., Barbara Phillips-Bute, Ph.D., Mihai V. Podgoreanu, M.D., Andrew D. Shaw, M.D., Mark Stafford-Smith, M.D., Madhav Swaminathan, M.D., Ian Welsby, M.D., William D. White, M.P.H., Lisa Anderson, Lauren Baker, B.S., Jerry Dove, R.N., Bonita L. Funk, R.N., Roger L. Hall, A.A.S., Gladwell Mbochi, A.A.S., Tiffany Bisanar, R.N., Prometheus T. Solon, M.D., Peter Waweru.
Perfusion Services:
Kevin Collins, B.S., C.C.P., Greg Smigla, B.S., C.C.P., Ian Shearer, B.S., C.C.P.
Surgery:
Thomas A. D'Amico, M.D., R. Duane Davis, M.D., Donald D. Glower, M.D., R. David Harpole, M.D., G. Chad Hughes, M.D., James Jaggers, M.D., Shu Lin, M.D., Andrew Lodge, M.D., James E. Lowe, M.D., Carmelo Milano, M.D., Peter K. Smith, M.D., Jeffrey Gaca, M.D., Mark Onatis, M.D.
Fig. 1.  Predicted association of mean duration per incursion (minutes) outside of threshold (105–130 mmHg) and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values.
Fig. 1. 
	Predicted association of mean duration per incursion (minutes) outside of threshold (105–130 mmHg) and 30-day mortality in the combined sample (n = 7504). Shaded area 
	represents the 95% confidence intervals for the predicted values.
Fig. 1.  Predicted association of mean duration per incursion (minutes) outside of threshold (105–130 mmHg) and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values.
×
Fig. 2.  Receiver operating characteristic curves showing association between duration of incursions and 30-day mortality in the development and validation samples. Results are graphically presented using the complete dataset divided into two parts, development and validation. The c index for the logistic regression model in the development sample (equivalent to the area under the receiver operating characteristic curve) is 0.703. The c index for the validation sample is 0.79. A reference line has been included to illustrate a c index of 0.5, indicating no association.
Fig. 2. 
	Receiver operating characteristic curves showing association between duration of incursions and 30-day mortality in the development and validation samples. Results are graphically presented using the complete dataset divided into two parts, development and validation. The c index for the logistic regression model in the development sample (equivalent to the area under the receiver operating characteristic curve) is 0.703. The c index for the validation sample is 0.79. A reference line has been included to illustrate a c index of 0.5, indicating no association.
Fig. 2.  Receiver operating characteristic curves showing association between duration of incursions and 30-day mortality in the development and validation samples. Results are graphically presented using the complete dataset divided into two parts, development and validation. The c index for the logistic regression model in the development sample (equivalent to the area under the receiver operating characteristic curve) is 0.703. The c index for the validation sample is 0.79. A reference line has been included to illustrate a c index of 0.5, indicating no association.
×
Fig. 3.  (A  ) Close-up view of the first 30 days of the survival curves, showing risk adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). (B  ) Risk-adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). P  value for the continuous variable is less than 0.0001. (C  ) Risk-adjusted survival estimates using the Cox proportional-hazard method for three mean incursion values outside the range of 105–130 mmHg: 9, 16, and 19 min (25th, 50th, and 75th percentiles). P  value for the continuous variable is less than 0.0001.
Fig. 3. 
	(A 
	) Close-up view of the first 30 days of the survival curves, showing risk adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). (B 
	) Risk-adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). P 
	value for the continuous variable is less than 0.0001. (C 
	) Risk-adjusted survival estimates using the Cox proportional-hazard method for three mean incursion values outside the range of 105–130 mmHg: 9, 16, and 19 min (25th, 50th, and 75th percentiles). P 
	value for the continuous variable is less than 0.0001.
Fig. 3.  (A  ) Close-up view of the first 30 days of the survival curves, showing risk adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). (B  ) Risk-adjusted survival estimates using the Cox proportional-hazard method for two mean incursion values outside the range of 105–130 mmHg: 9 and 19 min (25th and 75th percentiles). P  value for the continuous variable is less than 0.0001. (C  ) Risk-adjusted survival estimates using the Cox proportional-hazard method for three mean incursion values outside the range of 105–130 mmHg: 9, 16, and 19 min (25th, 50th, and 75th percentiles). P  value for the continuous variable is less than 0.0001.
×
Fig. 4.  (A  ) Predicted association of mean duration per incursion (minutes) below the threshold 95 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values. (B  ) Predicted association of mean duration per incursion (minutes) above the threshold 130 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values.
Fig. 4. 
	(A 
	) Predicted association of mean duration per incursion (minutes) below the threshold 95 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area 
	represents the 95% confidence intervals for the predicted values. (B 
	) Predicted association of mean duration per incursion (minutes) above the threshold 130 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area 
	represents the 95% confidence intervals for the predicted values.
Fig. 4.  (A  ) Predicted association of mean duration per incursion (minutes) below the threshold 95 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values. (B  ) Predicted association of mean duration per incursion (minutes) above the threshold 130 mmHg and 30-day mortality in the combined sample (n = 7504). Shaded area  represents the 95% confidence intervals for the predicted values.
×
Fig. 5.  (A  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent outside the range). (B  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent below the range). (C  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent above the range).
Fig. 5. 
	(A 
	) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent outside the range). (B 
	) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent below the range). (C 
	) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent above the range).
Fig. 5.  (A  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent outside the range). (B  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent below the range). (C  ) Unadjusted data of 30-day mortality within each tertiles of dataset, based on blood pressure variability (minutes per episode spent above the range).
×
Table 1.  Patient Characteristics
Image not available
Table 1.  Patient Characteristics
×
Table 2.  Characterizations of Blood Pressure Variability and Association with 30-day Mortality in the Development Sample
Image not available
Table 2.  Characterizations of Blood Pressure Variability and Association with 30-day Mortality in the Development Sample
×
Table 3.  Multivariable Logistic Regression Model for Mean Duration < 95 mmHg
Image not available
Table 3.  Multivariable Logistic Regression Model for Mean Duration < 95 mmHg
×
Table 4.  Multivariable Logistic Regression Model for Mean Duration > 130 mmHg
Image not available
Table 4.  Multivariable Logistic Regression Model for Mean Duration > 130 mmHg
×