Intraoperative Hypotension and Acute Kidney Injury, Stroke, and Mortality during and outside Cardiopulmonary Bypass: A Retrospective Observational Cohort Study

In this retrospective observational cohort study, we analyzed the data that were prospectively collected in patients having cardiac surgery from institutional electronic medical records, the Society of Thoracic Surgeons (Chicago, Illinois) adult cardiac surgery database, and Anesthesia Information Systems after institutional review board approval (Beth Israel Deaconess Medical Center, Boston, Massachusetts, protocol No. 2020P000074). The institutional review board waived informed consent, and the article adheres to the applicable Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) standards for observational studies.²¹  Data for adult patients (greater than 18 yr of age) who had cardiac surgery with CPB between January 2008 and June 2016 were included. Data for patients who underwent emergency cardiac surgery, aortic surgeries, and other procedures, such as atrial fibrillation ablation, pericardial window, maze procedure, left atrial appendage resection, septal myomectomy, and atrial septal defect closure, or invalid, incomplete, or unavailable demographic, baseline, hemodynamic, and outcome data were excluded (fig. 1).

The perioperative details of anesthesia management have been described previously.²²  In brief, preoperative medications were continued until the time of the surgery unless contraindicated by the patient’s condition. All patients received a preoperative carotid scan. Anyone with 80% or more carotid stenosis on either side or symptoms suggestive of carotid stenosis was consulted with vascular surgery for appropriate management before surgery. All patients underwent general anesthesia with endotracheal intubation. Standard IV induction was performed using IV propofol and fentanyl. Rocuronium was used for muscle relaxation and isoflurane (0.5 to 1%) for maintenance. Valvular surgeries were performed under cardiac arrest using a standard institutional cardioplegia solution (60 mEq potassium, 8 mEq magnesium, 2.5 g dextrose, 10 mEq Tromethamine, and 500 ml normal saline). The patients were maintained under mild hypothermia, and an α-stat pH strategy was used for blood gas management during CPB.

All patients received transoesophageal echocardiography (TEE) monitoring throughout the surgery unless contraindicated. The rate, rhythm, preload, afterload, and contractility were maintained with information obtained from arterial catheters, pulmonary artery or central venous catheters, and TEE. All cardiac anesthesiologists involved in patient care were TEE board–certified. After induction, IV phenylephrine or norepinephrine infusions were started to maintain a systolic blood pressure of 90 to 120 mmHg.

After CPB, in patients with low to moderately dysfunctional ventricular function (preoperative left ventricular ejection fraction less than 30% or with significant mitral or aortic regurgitation), IV epinephrine was started to maintain a cardiac index greater than 2 l · min^–1 · m^–2. Intravenous milrinone was added to this regimen to achieve the same goals at the clinician’s discretion. Other medications, such as dopamine and dobutamine, were rarely added to this strategy. Afterload was maintained with phenylephrine, norepinephrine, or both, either by themselves or along with the inotropes. The management was guided by direct visualization of the ventricles, echocardiographic assessment of cardiac function with or without pulmonary artery catheter measurements of mixed venous oxygen saturation, and cardiac index.

Vasopressor-inotrope dose was calculated by adding norepinephrine equivalents of total norepinephrine, epinephrine, phenylephrine, and vasopressin dose used during surgery, using the following formula: total vasopressor-inotrope dose = [norepinephrine (µg/min) × min] + [epinephrine (µg/min) × min] + [(phenylephrine (µg/min) × min) ÷ 10] + [vasopressin (U/h) × 8.33 × min].^23,24  The total dose of milrinone used was obtained from the database and analyzed separately.

Invasive blood pressure was recorded every 15 s, and artifacts were removed using the rules that were previously described.^5,25  Intraoperative hypotension was defined as MAP less than 65 mmHg per the noncardiac surgical literature.¹⁵  It was characterized by (1) total duration as cumulative minutes and (2) area under a MAP of 65 mmHg (area under the receiver operating characteristics curve [AUC]–MAP 65 mmHg) measured based on the trapezoidal rule.^10,26  Using this method, values below the threshold were subtracted from the threshold, multiplied by 0.25 (which corresponds to 15 s), and summed together. For example, each MAP value x less than 65 mmHg will be subtracted from 65, multiplied by 0.25 min, and all such values were summed together.

The duration of intraoperative hypotension was divided into three categories based on how much of the total intraoperative hypotension occurred during CPB (fraction of intraoperative hypotension during CPB): (1) more than 80%, (2) 80 to 60% (inclusive), and (3) less than 60%.

The primary outcome was a composite of three major adverse events: (1) stroke, defined as any confirmed postoperative neurologic deficit of abrupt onset caused by a disturbance in cerebral blood supply that did not resolve within 24 h; (2) AKI, defined as one or both of the following: threefold rise in serum creatinine from baseline or a creatinine level greater than 4.0 mg/dl with a minimum rise of 0.5 mg/dl and/or a new requirement for dialysis postoperatively; and (3) mortality, defined as all deaths, regardless of cause, occurring during the hospitalization in which the surgery was performed, even if after 30 days (including patients transferred to other acute care facilities) or all deaths, regardless of cause, occurring after discharge from the hospital, but before the end of postoperative day 30, based on Society of Thoracic Surgeons definition, version 2.73 (Supplemental Digital Content 1, appendix 1, https://links.lww.com/ALN/C817). Stroke or AKI occurring during the index hospitalization, even if more than 30 days, was included in the primary outcome.

Descriptive statistics were computed as mean values with SD for continuous variables and frequencies with percentages for categorical variables. The Kruskal–Wallis test was used for normality testing. Continuous (normal and nonnormal) and categorical measures were appropriately tested with t tests, Wilcoxon rank sum tests, and chi-square tests, respectively. Multivariable logistic regression models were used to assess the association between intraoperative hypotension per 10-min exposure to MAP less than 65 mmHg, AUC-MAP 65 mmHg, vasopressor-inotrope dose, total dose of milrinone, and the composite primary outcome (primary model). Confounders were predefined based on clinical plausibility, which included age in years, sex, surgery category, Society of Thoracic Surgeons risk scores in tertiles, preoperative left ventricular ejection fraction, change in hematocrit percentage, and aortic cross-clamp time. The same confounders were used for both intraoperative hypotension variables (total duration per 10-min exposure to MAP less than 65 mmHg and AUC-MAP 65 mmHg) analyses. We introduced the statistically (P < 0.10 in the univariable analysis) and clinically significant variables into the model. The total duration of intraoperative hypotension per 10-min exposure to MAP less than 65 mmHg and AUC-MAP 65 mmHg were considered as two separate multivariable analyses. In an attempt to determine the phase-specific associations of intraoperative hypotension during cardiac surgery, we constructed similar multivariable logistic regression models exploring the relationship between hypotension duration per 10-min exposure to MAP less than 65 mmHg during and outside CPB with the composite primary outcome. Similar models were used to explore the association between the fraction of hypotension duration during CPB (greater than 80%, 80 to 60%, and less than 60%) with the composite outcome and individual major adverse events.

The linearity assumption of the effects of intraoperative hypotension, vasopressor-inotrope dose, and total dose of milrinone on the log odds of the outcome variable was assessed using restricted cubic splines with three knots and generalized additive models. In the cases in which significant nonlinear terms were detected, the corresponding variables were categorized using tertiles/quartiles. The effect modification of vasopressor-inotrope dose and total dose of milrinone on the association between intraoperative hypotension and the composite outcome was evaluated by including an interaction term of vasopressor-inotrope dose and intraoperative hypotension in the logistic regression models. In the case of nonsignificant interaction, the final model only with main effects was reported. The Hosmer–Lemeshow goodness-of-fit test and AUC were used to assess the model fit. The precision-recall AUC was reported given the class imbalance in this data set. The variable inflation factor was used to assess the correlation between predictors. Intraoperative hypotension duration was introduced into the model as a restricted cubic spline for measuring the predicted probabilities of the composite primary outcome. Statistical analyses were performed in R version 3.5.1 and RStudio version 1.1.463 (RStudio, USA). All tests were two-sided, with 0.05 as the level of significance. Correction for multiple testing and models was not performed. No specific power calculation was performed for this retrospective observational cohort study. The data analysis and statistical plan were written after the data were accessed.²⁷ 

We performed two sensitivity analyses to further explore the robustness of the association between intraoperative hypotension and the composite primary outcome. The first sensitivity analysis was designed to examine the association between total duration and AUC-MAP 65 mmHg with the composite outcome wherein we used preoperative comorbid conditions as the variables in the multivariable logistic regression models instead of the Society of Thoracic Surgeons risk score. Although we used the Society of Thoracic Surgeons risk score, which itself includes all preoperative comorbidities, in addition to multiple other perioperative variables in our primary analyses, we wanted to examine the association when considering individual comorbidities. The second sensitivity analysis was designed to eliminate the potential impact of outlier cases with an extremely long duration of hypotension. Accordingly, we excluded the top 5% of patients from our final cohort and determined the association between total hypotension duration and AUC-MAP 65 mmHg with the composite outcome. Our primary model in this study evolved as a result of the peer review process, and we focused on specific models following the suggestions from reviewers and editors.

Figure 1 displays an overview of patient selection and analysis. Data describing 6,024 patients who had cardiac surgeries from 2008 to 2016 were collected. We excluded 269 patients who had emergent surgeries. From the remaining 5,755 patients, 597 patients were excluded due to inadequate data; 174 patients who underwent procedures other than coronary artery bypass graft, valve, or combined were excluded; and the final analysis included 4,984 patients. Table 1 summarizes the baseline characteristics of patients stratified by the postoperative composite outcome. The mean ± SD age of our study population was 67 ± 11 yr. There were 3,491 (70%) male patients, 1,165 (23%) patients had diabetes, 3,933 (79%) patients had hypertension, and 3,807 (76%) patients had dyslipidemia. Previous myocardial infarction was present in 1,565 (31%), congestive heart failure was present in 1,741 (35%), and 85 (2%) patients were on dialysis before surgery.

Postoperative major adverse events are displayed in Supplemental Digital Content 2 (table 1, https://links.lww.com/ALN/C818). The composite primary outcome occurred in 256 (5.1%). AKI was present in 125 (2.5%) patients, 66 (1.3%) patients had a stroke, and 109 (2.2%) patients died after surgery. Baseline characteristics of patients stratified by the composite primary outcome are presented in table 1. Patients having one of the major adverse events of the composite primary outcome were significantly older and more likely to be male and have diabetes (table 1). They also had a higher incidence of congestive heart failure, previous myocardial injury, chronic lung disease, and need for inotropic support and steroid medications as compared to those without the composite primary outcome. Society of Thoracic Surgeons risk scores were statistically significantly higher in patients with the composite primary outcome compared to those without (mean ± SD, 0.05 ± 0.06 vs. 0.02 ± 0.03; P < 0.001).

In addition, patients with the composite primary outcome had a statistically significant increased mean ± SD CPB time (121 ± 61 vs. 93 ± 36 min; P < 0.001), cross-clamp time (89 ± 42 vs. 73 ± 29 min; P < 0.001], duration of MAP less than 65 mmHg (143 ± 75 vs. 104 ± 48 min; P < 0.001) and AUC-MAP 65 mmHg (1,528 ± 1,134 vs. 1,070 ± 656 mmHg/min; P < 0.001) compared to those without the composite primary outcome. Similarly, median (interquartile range) vasopressor-inotrope dose (milligrams) was statistically significantly higher in patients with the composite primary outcome compared to those without the composite primary outcome (0.98 [0.53 to 2.24] vs. 0.65 [0.36 to 1.07] mg; P < 0.001; table 1).

Every 10-min increase in intraoperative hypotension duration of MAP less than 65 mmHg, AUC-MAP 65 mmHg, vasopressor-inotrope dose, and total dose of milrinone were categorized as continuous variables. The results of multivariable logistic regression analysis evaluating the association between the total duration of intraoperative hypotension throughout the surgery and the composite primary outcome are displayed in table 2. The association of total duration of intraoperative hypotension per 10-min exposure to MAP less than 65 mmHg with the composite primary outcome was statistically significant (adjusted odds ratio, 1.05; 95% CI, 1.02 to 1.08; P = 0.032). No statistically significant associations were observed for vasopressor-inotrope dose or the total dose of milrinone (adjusted odds ratio, 1.00; 95% CI, 1.00 to 1.00; P = 0.247; and adjusted odds ratio, 1.00; 95% CI, 0 to 1.00; P = 0.648).

Statistically significant associations were seen with intraoperative hypotension defined as AUC-MAP 65 mmHg throughout the surgery and the composite outcome (adjusted odds ratio, 1.00; 95% CI, 1.00 to 1.00; P < 0.001; Supplemental Digital Content 3, table 2, https://links.lww.com/ALN/C819). However, no statistically significant associations were seen with vasopressor-inotrope dose or total dose of milrinone (adjusted odds ratio, 1.00; 95% CI, 1.00 to 1.00; P = 0.475; and adjusted odds ratio, 1.00; 95% CI, 0 to 1.00; P = 0.756; Supplemental Digital Content 3, table 2, https://links.lww.com/ALN/C819).

The relationship between the total duration of intraoperative hypotension (quartiles), vasopressor-inotrope dose (quartiles), and the composite primary outcome rate is shown in figure 2. A higher total duration of hypotension and vasopressor-inotrope dose was associated with a higher composite primary outcome incidence (fig. 2). While assessing the model fit, the models demonstrated an average AUC of 0.761 and precision-recall AUC of 0.164.

Tables 3 and 4 illustrate the results of multivariable logistic regression models examining the association between duration of hypotension per 10-min exposure to MAP less than 65 mmHg during and outside the CPB phase with the composite primary outcome. Statistically significant associations were seen with hypotension per 10-min exposure to MAP less than 65 mmHg (adjusted odds ratio, 1.06; 95% CI, 1.03 to 1.10; P = 0.001) outside the CPB phase and the composite outcome. However, hypotension per 10-min exposure to MAP less than 65 mmHg during CPB phase (adjusted odds ratio, 1.04; 95% CI, 0.99 to 1.09; P = 0.118) was not associated with the primary outcome.

The association between varying fractions of hypotension during CPB and the composite primary outcome is presented in table 5. When compared with more than 80% hypotension duration occurring during CPB, exposure to 80 to 60% of hypotension during CPB was not statistically significantly associated with the primary composite outcome (adjusted odds ratio, 1.45; 95% CI, 0.97 to 2.23; P = 0.082), while less than 60% of hypotension occurring during CPB was statistically significantly associated with the primary composite outcome (odds ratio, 1.67; 95% CI, 1.10 to 2.60; P = 0.019). The associations between varying fractions of hypotension occurring during CPB with individual components of the composite outcome are presented in Supplemental Digital Content 4 (table 3, https://links.lww.com/ALN/C820).

The association between the total duration of hypotension per 10-min exposure to MAP less than 65 mmHg and AUC-MAP 65 mmHg with the composite primary outcome remained statistically significant in sensitivity analyses. In the model that included preoperative comorbidities instead of the Society of Thoracic Surgeons risk score (Supplemental Digital Content 5, table 4, https://links.lww.com/ALN/C821), both duration and AUC-MAP 65 mmHg were statistically significantly associated with the composite primary outcome (adjusted odds ratio, 1.00; 95% CI, 1.00 to 1.01; P = 0.002; and adjusted odds ratio, 1.00; 95% CI, 1.00 to 1.00; P = 0.017). Similarly, in the other sensitivity analysis that explored similar associations excluding the top 5% of patients who had an extremely long duration of hypotension (Supplemental Digital Content 6, table 5, https://links.lww.com/ALN/C822), the relationship between duration and AUC-MAP 65 mmHg was similar to our primary analysis (adjusted odds ratio, 1.01; 95% CI, 1.00 to 1.01; P < 0.001; and adjusted odds ratio, 1.00; 95% CI, 1.00 to 1.00; P = 0.002).

Supplemental Digital Content 7 (table 6, https://links.lww.com/ALN/C823) illustrates the results of demographic information comparing patients included in the final cohort and patients who were excluded due to missing data.

Supplemental Digital Content 8 (fig. 1, https://links.lww.com/ALN/C824) presents the predicted probabilities of the composite primary outcome plotted against the total duration of intraoperative hypotension. The top 5% of the cohort have extremely long durations of hypotension (approximately 200 min; Supplemental Digital Content 8, fig. 1A, https://links.lww.com/ALN/C824). Supplemental Digital Content 8 (fig. 1B, https://links.lww.com/ALN/C824) presents a similar relationship excluding this top 5% of patients from the final cohort. Supplemental Digital Content 9 (fig. 2, https://links.lww.com/ALN/C825) illustrates the predicted probabilities of the composite primary outcome with vasopressor-inotrope dose (Supplemental Digital Content 9, fig. 2A) and the total dose of milrinone (Supplemental Digital Content 9, fig. 2B).

Supplemental Digital Content 10 (fig. 3, https://links.lww.com/ALN/C826) demonstrates the distribution of the Society of Thoracic Surgeons risk score among the final cohort of patients. We could observe the nonlinearity of the Society of Thoracic Surgeons risk score among the cohort. Supplemental Digital Content 11 (fig. 4, https://links.lww.com/ALN/C827) illustrates the correlation between intraoperative hypotension total duration with vasopressor-inotrope dose, total dose of milrinone, Society of Thoracic Surgeons risk score, age, sex, surgery category, hematocrit change, ejection fraction, duration of surgery, and cross-clamp time. We provide open access to all our data extraction, filtering, data wrangling, modeling, figures, and table code and queries at https://github.com/theonesp/vasopressor_dose_mae.

In patients undergoing cardiac surgery with CPB, the total duration of intraoperative hypotension per 10-min exposure of MAP less than 65 mmHg throughout the surgery was statistically significantly associated with the composite primary outcome of stroke, AKI, or death. Similar associations were seen with the duration of intraoperative hypotension outside the CPB phase and composite primary outcome.

We observed a significant association between intraoperative hypotension during cardiac surgery and adverse outcomes. These results broadly support the work of previous research showing that hypotension during cardiac surgery was significantly associated with stroke,²⁰  renal injury,²⁸  and mortality.²⁹  Similar to this research, in studies exploring the CPB phase–specific hypotension in patients undergoing cardiac surgery, the risk of postoperative stroke²⁰  and renal replacement therapy²⁸  increased for every 10-min exposure to MAP less than 65 mmHg.

In addition to duration, we characterized the severity of intraoperative hypotension as AUC-MAP 65 mmHg. Both total intraoperative hypotension duration and AUC-MAP 65 mmHg showed similar results. Our results were consistent with previous studies,^5,25  in which both the duration and AUC for blood pressure thresholds were associated with adverse outcomes.

When exploring an intraoperative phase–specific relationship with adverse events, hypotension duration outside the CPB phase was significantly associated with the primary outcome. Our results mirror the findings of a similar study in which MAP less than 65 mmHg for 10 min or more during the post-CPB phase significantly increased the risk of adverse events.²⁸  Compared with the fraction of overall hypotension duration occurring during CPB of more than 80%, a fraction of hypotension duration occurring during CPB of less than 60% was statistically significantly associated with a higher risk of the composite outcome. This observation may be explained by the physiologic stress during the post-CPB phase. Various mechanisms such as catecholamine surges, mechanical trauma to red blood cells triggering inflammatory states, and decreased vasomotor reactivity after CPB in the presence of hypotension could worsen end-organ damage, resulting in adverse outcomes. It is possible, therefore, that hypotension outside the CPB phase is an important risk factor that needs to be taken into consideration, and more research is needed regarding the effects of hypotension during and outside CPB.^30,31 

Advanced hemodynamic monitoring and protocolized treatment strategies in the cardiac surgical setting can identify and treat hemodynamic changes earlier than in noncardiac surgery with noninvasive monitoring. In this study, we did not observe a statistically significant association between vasopressor-inotrope dose or total dose of milrinone and the composite outcome. Even though vasopressor-inotropes were used to treat hypotension using advanced hemodynamic monitoring, the duration of intraoperative hypotension was still associated with the composite outcome.

Transesophageal echocardiography, invasive arterial blood pressure monitoring, and central venous pressures were routinely used in our care setting to optimize fluids and maintain blood pressures throughout cardiac surgery. In this context, vasopressor use to optimize blood pressures was not associated with a higher or lower incidence of major adverse events in our study. Other studies have shown a similar lack of reduction in postoperative adverse outcomes with the use of vasopressors to achieve higher MAP thresholds.^22,32–34  It remains to be explored whether vasopressor treatment aiming to achieve a fixed intraoperative blood pressure target (e.g., MAP of 65 mmHg) could improve postoperative outcomes.

Weis et al.³⁵  observed that cardiac surgical patients needing more vasopressors had a greater degree of a systemic inflammatory response that has been associated with poor outcomes.³⁶  Moreover, significant associations between increased vasopressor support and renal replacement therapy, length of stay, and periods of ventilation after cardiac surgery were reported.³⁵ 

The relationship between organ perfusion and adverse outcomes remains complex and has recently gained increased scientific attention. Further work is needed to gain a better understanding regarding markers such as oxygen delivery as a measure for tissue perfusion.

This is a retrospective study, and finding a causative link between intraoperative hypotension or vasopressor-inotrope dose and postoperative outcomes is thus not possible. Moreover, the vasopressor-inotrope dose was calculated by adding norepinephrine equivalents of vasopressors and inotropes used. As each drug might have a different molecular target, there is a possibility that we might have missed the effect of an individual medication. The sample size may be relatively small to evaluate the true interaction between intraoperative hypotension, vasopressor-inotrope dose, and outcome. Our analyses did not include data reflecting postoperative hypotension and other factors related to tissue perfusion, such as mixed venous oxygen saturation or blood gas parameters that might have an association with the incidence of adverse outcomes. In addition, our analysis is limited by a possible lack of power for many of the nonsignificant observations and lack of adjustment for multiple testing.

To conclude, in patients having cardiac surgery with CPB, the total duration of intraoperative hypotension per 10-min exposure of MAP less than 65 mmHg was statistically significantly associated with the composite primary outcome of stroke, AKI, or mortality. Hypotension per 10-min exposure of MAP less than 65 mmHg outside the CPB phase was statistically significantly associated with the composite outcome. These results confirm findings from previous studies exploring the intraoperative phase–specific association between hypotension and adverse outcomes. Vasopressor-inotrope dose or total dose of milrinone was not statistically significantly associated with the composite primary outcome.

The authors thank the Center for Anesthesia Research Excellence within the Department of Anesthesia, Critical Care and Pain Medicine at Beth Israel Deaconess Medical Center (Boston, Massachusetts).

Supported by National Institutes of Health (Bethesda, Maryland) Research Project grant Nos. GM 098406 and R01AG065554 (to Dr. Subramaniam) and funds from Cardiomed Medical Consultants LLC (Edgewater, Maryland; to Dr. Novack).

Dr. Saugel has received honoraria for consulting, honoraria for giving lectures, and refunds of travel expenses from Edwards Lifesciences Inc. (Irvine, California). He has received honoraria for consulting, institutional restricted research grants, honoraria for giving lectures, and refunds of travel expenses from Pulsion Medical Systems SE (Feldkirchen, Germany). Dr. Saugel has received institutional restricted research grants, honoraria for giving lectures, and refunds of travel expenses from CNSystems Medizintechnik GmbH (Graz, Austria). He has received institutional restricted research grants from Retia Medical LLC. (Valhalla, New York). He has received honoraria for giving lectures from Philips Medizin Systeme Böblingen GmbH (Böblingen, Germany). Dr. Saugel has received honoraria for consulting, institutional restricted research grants, and refunds of travel expenses from Tensys Medical Inc. (San Diego, California). The other authors declare no competing interests.