Postoperative Hypotension after Noncardiac Surgery and the Association with Myocardial Injury

This single-center observational cohort study was derived from an ongoing clinical routine troponin registry of noncardiac surgery patients at the Erasmus University Medical Center, Rotterdam, The Netherlands.^12,18  Eligibility criteria were patients aged 60 yr or older undergoing intermediate- to high-risk noncardiac surgery, including elective and emergency procedures with an expected postoperative length of hospitalization of at least 24 h. Patients were included in the period between July 1, 2012, and July 1, 2017. The sample size was based on the available data. When clinically indicated, patients were either continuously monitored on the high-dependency ward (for the remaining day of surgery until the day after surgery), in the intensive care unit, or on the recovery ward before discharge to the wards. Patients who were admitted to the high-dependency ward were selected for analysis. In our hospital, the decision to admit patients to the high-dependency ward after surgery is made several weeks in advance before the surgery itself at the outpatient clinic. Patients are admitted to the high-dependency ward based on either the type of surgical procedure (e.g., intracranial, major abdominal) or significant comorbidity, as judged by the screening anesthesiologist. Patients with a postoperative duration less than 8 h before being discharged to the wards or with a revised procedure within the postoperative period were excluded from analysis (fig. 1). Furthermore, cases were excluded if either perioperative hemodynamic measurements or postoperative high-sensitive troponin T were unavailable. Institutional approval for this study was obtained, and no informed consent was required nor obtained according to local directives for retrospective studies. This study was not registered, was not subject to the Dutch Medical Research Involving Human Subjects Act¹⁹  due to this observational character, and complies with the Helsinki Declaration on research ethics.²⁰  This report follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) criteria for observational studies.²¹ 

At the outpatient clinic, patients were screened on medical history, physical examination, laboratory measurements, and an electrocardiogram according to local policy. Baseline characteristics were acquired from medical records and consisted of age, sex, type of surgery, emergency procedures, hypertension, insulin-dependent diabetes mellitus, chronic obstructive pulmonary disease, previous myocardial infarction (MI), coronary artery disease, congestive heart failure, cerebrovascular disease, and peripheral artery disease. Preoperative use of medication was recorded, including β-blockers, statins, angiotensin-converting enzyme inhibitors, angiotensin-II antagonists, calcium channel blockers, diuretics, aspirin, and oral anticoagulants. Additional perioperative laboratory measurements, ward vitals, and admission details were retrospectively extracted from the institution’s electronic medical record storage database.

Perioperative blood pressure recordings were extracted from the hospital anesthesia information management system data. The intraoperative period was specified as the documented start of anesthesia until the departure from the operating room. The postoperative period was defined as the departure from the operating room until the patient’s discharge from the high-dependency ward to the surgical wards. A maximum of 24 h of postoperative measurements was selected for analysis. Preoperative blood pressures were acquired from all measurements obtained at the wards or from the outpatient clinic noninvasively (oscillometrically). Intraoperative blood pressures were measured and recorded either invasively (arterial line catheter) at 1-min intervals or noninvasively at 1- to 5-min intervals. Postoperative blood pressures on the high-dependency ward were measured invasively or noninvasively and recorded at 1- to 15-min intervals. Based on a previously published algorithm, the following measurements were considered artifacts and removed from the data accordingly: a systolic blood pressure either less than 20 or greater than 300 mmHg; a diastolic blood pressure either less than 20 or greater than 200 mmHg; a diastolic blood pressure greater than the systolic blood pressure; and a diastolic blood pressure greater than 20 mmHg below the systolic blood pressure.¹⁰  Invasive blood pressure measurements were preferred above noninvasive if both measurements were available. Intervals between blood pressure measurements were linearly interpolated.

Absolute MAP thresholds were used for analyses. Multiple MAP thresholds (50 to 75 mmHg, with 5-mmHg increments)²²  were initially selected to examine postoperative thresholds for myocardial injury and to assess differences between intraoperative and postoperative thresholds.

Different characterizations of blood pressure exposures accounting for time components were calculated for each patient: lowest MAP for multiple cumulative minutes, duration, area, and time-weighted average under MAP thresholds. The lowest MAP was defined as a patient’s lowest MAP during the whole intra- or postoperative period for a minimal cumulative duration of 1, 5, 10, 15, 30, 60, 120, and 240 min. Duration was defined as the cumulative length in minutes a patient’s MAP had decreased below the threshold. Area under a MAP threshold was defined as the depth underneath the threshold multiplied by duration, expressed as mmHg • minutes, as the severity of hypotension. Additionally, to account for differences in durations of the perioperative periods and assuming measurements are not equidistant, time-weighted average under a MAP threshold, expressed in mmHg, was calculated, defined as area divided by the total duration of the intra- or postoperative period. We defined our main exposure as duration under the selected MAP threshold in minutes with area and time-weighted average under the MAP threshold as additional exposures.

Troponin measurements were routinely obtained on postoperative days 1, 2, and 3, unless discharged earlier. High-sensitive troponin T was measured using the Cobas e602 Troponin T hs STAT assay (Roche Diagnostics, Germany). A peak high-sensitive troponin T measurement of 50 ng/l and above during the first 3 postoperative days was defined as myocardial injury after surgery and used as the primary outcome, as previously applied in our similar cohorts^12,18  and within the current clinical practice of our institution.

Furthermore, postoperative MI (which was ruled in or out based on the criteria according to the third universal definition²³ ) and 30-day all-cause mortality were assessed after surgery and used as secondary endpoints. Survival status was completed in all patients by means of using the institution’s medical records or was ascertained by inquiry from the civil registries.

A data analysis and statistical plan was written after the data were accessed. No statistical power calculation was conducted before the study. Continuous variables are presented as medians with interquartile range or as means and SDs as appropriate. Categorical variables are presented as numbers and percentages. Complete case analyses were performed. Hypotension exposures between patients with and without myocardial injury were compared using the Mann–Whitney U test. Normality for baseline characteristics was assessed visually with histograms and normal quantile–quantile plots. Incidence of intra- and postoperative hypotension is presented as percentages of patients with a cumulative duration below a MAP threshold and were linearly interpolated on the graphs.

To determine the MAP threshold in the postoperative period at which risk of myocardial injury starts to increase, graphs of the estimated probabilities based on the lowest MAP were plotted and inspected using adjusted binomial logistic regression models. In response to peer review, only the two extremes from the preselected cumulative durations (i.e., 1 min and 4 h) were used for inspection to prevent overfitting. Potential confounders were selected and entered (age, sex, high-risk surgery, emergency procedures, intra- and postoperative heart rate, previous history of hypertension, insulin-dependent diabetes mellitus, coronary artery disease, congestive heart failure, cerebrovascular disease, renal disease, estimated blood loss, and length of surgery) based on known factors associated with perioperative hypotension, postoperative myocardial injury, and cardiovascular and/or mortality risk.^24,25  Restricted cubic splines analyses were applied between the lowest MAP in cumulative minutes and myocardial injury in the models and were retained when they improved the model.

Univariable and multivariable logistic regression analyses were used to assess the association between intra- and postoperative hypotension with myocardial injury. The absolute threshold of 65 mmHg was selected for intraoperative hypotension based on previous studies.^5,8  The threshold for postoperative hypotension was based on examining the plots of the lowest MAP for cumulative minutes and postoperative myocardial injury, as previously explained. One single threshold was selected to define postoperative hypotension for our analyses using three different hypotension exposures (i.e., duration, area, and time-weighted average under MAP) and the association with myocardial injury. Five additional thresholds were selected for comparing the association of the main exposure (duration under MAP threshold) and myocardial injury.

Due to nonlinearity, duration, area, and time-weighted average under the MAP threshold were divided in quartiles with their zero time spent under a threshold as the reference group. Duration under MAP was additionally divided and categorized into less than 0, 0 to 1, 1 to 2, 2 to 4, and greater than 4 h below the threshold. Similar confounders, as described previously, were used for the association between MAP exposures and myocardial injury. Multivariable analyses were first adjusted for potential confounders without and subsequently with intraoperative hypotension to determine if the previous hypotension occurrence would affect postoperative hypotension in the final model. Multicollinearity between the different occurrences of intra- and postoperative hypotension was assessed using variance inflation factor (threshold 5). The interaction between intraoperative hypotension and postoperative hypotension was tested and dropped if not statistically significant. Sensitivity analyses of our primary analysis were performed using the manufacturer’s 99th percentile reference value of high-sensitive troponin T (i.e., 14 ng/l) to define myocardial injury.²⁶  Additionally, in response to peer review, several post hoc sensitivity analyses were conducted. These included patients with only invasive blood pressure monitoring and patients with peak troponin elevations on various days after surgery.

Results are reported as odds ratios with their 95% CI. All statistical tests were two-tailed. Significance was set at P < 0.05 for comparing between groups and interaction terms. The Bonferroni correction was applied accordingly when comparing three different exposures for postoperative hypotension, resulting in a P value of 0.05/3 = 0.017 as level of statistical significance. When comparing for five multiple thresholds for postoperative hypotension, a P value of 0.05/5 = 0.01 as level of statistical significance was used. All statistical analyses were performed using R software version 3.6.0, 2018 (The R Foundation for Statistical Computing, Austria).

The initial study cohort consisted of 4,795 noncardiac surgery patients who met eligibility criteria. Inclusion and exclusion of patients are presented in figure 1. Subsequently, our final study sample comprised a total of 1,710 patients with frequent postoperative hemodynamic monitoring on a high-dependency ward. A total of 1,587 of 1,710 (92%) of the patients had invasive blood pressure monitoring. The median postoperative time (interquartile range) on the high-dependency ward before being discharged to the wards was 21 h (17.7, 22.8).

Patients’ baseline and perioperative characteristics are presented in table 1. This cohort’s median age was 70 yr. Over half of the patients had a medical history of hypertension and 22% coronary artery disease. The median length of surgery was 236 (174, 322) min, and median blood loss 300 ml (100, 650). The first troponin was routinely measured on the morning after surgery at 6:00 am. In 36% of the cases, high-sensitive troponin T measurements were available on all 3 postoperative days, 31% of the cases on 2 postoperative days, and 33% on 1 postoperative day. In more than half of our patients (53%), peak high-sensitive troponin T occurred on postoperative day 1, 30% on postoperative day 2, and 17% on postoperative day 3. Overall, postoperative myocardial injury occurred in 238 (14%) patients. Of these, 52 (22%) were defined as MI, and 20 (8%) died within 30 days of all-cause mortality.

Figure 2 shows the incidence of intra- and postoperative hypotension for each MAP threshold in cumulative minutes. There were no differences between intraoperative hypotension exposures in duration, area, and time-weighted average under MAP thresholds in patients with myocardial injury compared to patients without (P > 0.05 for all, except for duration under MAP threshold 75 mmHg; table 2). In contrast to the intraoperative period, patients with myocardial injury had longer durations of postoperative hypotension compared to patients without, at all MAP thresholds 50 to 75 mmHg (P < 0.001 for all). Likewise, both postoperative hypotension exposures area and time-weighted average under MAP were higher in patients with myocardial injury at all thresholds (P < 0.001 for all).

Adjusted risk for myocardial injury based on the postoperative lowest MAP for a cumulative duration of 1 min and 4 h are shown in Supplemental Digital Content 1, figure 1(https://links.lww.com/ALN/C392). Probability of myocardial injury increased with decreasing MAP below the threshold of 75 mmHg for both cumulative durations. Longer cumulative durations of the lowest postoperative MAP were furthermore associated with exponentially increased risk of myocardial injury. A MAP threshold of 75 mmHg was hence selected based on the plots, with additional thresholds of 60, 65, 70, and 80 mmHg for comparison.

The interaction between intraoperative hypotension and postoperative hypotension was not statistically significant in all models and was subsequently removed from further analyses. Multicollinearities of intraoperative and postoperative hypotension were all minor with a variance inflation factor less than 2. Postoperative duration under a MAP threshold of 75 mmHg was associated with increased risk of myocardial injury after adjusting for potential confounders (fig. 3) and, moreover, remained statistically significant after intraoperative hypotension was added to the model; adjusted odds ratio (95% CI) was 2.68 (1.46 to 5.07) for the fourth quartile (table 3). In contrast, intraoperative hypotension as cumulative duration under the predefined MAP threshold of 65 mmHg was not associated with myocardial injury. Additional characterization of hypotension, area, and time-weighted average under MAP yielded similar unadjusted and adjusted results. After dividing and categorizing duration under MAP in hours, more than 4 h under a MAP of 70 mmHg (adjusted odds ratio, 2.18; 95% CI, 1.37 to 3.51, P = 0.001) was associated with myocardial injury (table 4). In our total cohort, patients who suffered myocardial injury were mainly male patients, had a higher incidence of previous cardiovascular history, and were taking more cardiovascular medication (Supplemental Digital Content 2, table 1, https://links.lww.com/ALN/C392). Compared to the total cohort of patients, patients in the lowest postoperative MAP quartile more often underwent major abdominal as well as high-risk procedures. However, they were predominantly not different in previous medical history or use of preoperative medication compared to the total cohort. Within the lowest postoperative MAP quartile, for those patients who suffered postoperative myocardial injury, similar differences can be seen in age, emergency surgery procedures, and cardiovascular medical history. Associations for additional thresholds of 65 and 60 mmHg showed similar patterns. There was no association with myocardial injury for all durations under a MAP of 75 and 80 mmHg (P > 0.01 for all). After substitution of the primary outcome with the manufacturer’s 99th percentile suggested reference threshold of 14 ng/l, associations were similar.

Furthermore, several post hoc additional analyses based on reviewer questions were conducted. A sensitivity analysis in patients where peak troponin was measured on day 2 or 3 after surgery as well as a sensitivity analysis in patients where blood pressure was measured with invasive blood pressure monitoring did not influence our main findings. Additionally, we conducted sensitivity analyses where the analyses were replicated separately in patients admitted to the ward or admitted to the ICU. The findings that postoperative hypotension, but not intraoperative hypotension, was associated with myocardial injury were consistent in these analyses.

In this study of patients undergoing intermediate- to high-risk noncardiac surgery, we investigated the association of postoperative hypotension with myocardial injury. Postoperative hypotension ranging from MAP 50 to 75 mmHg for multiple cumulative duration in minutes was common. A postoperative MAP below a threshold of 75 mmHg was found to increase the risk of myocardial injury, with shorter durations at lower thresholds being likewise harmful. After adjusting for potential confounders and intraoperative blood pressure, postoperative hypotension was independently associated with myocardial injury.

This current study confirms former results correlating the lowest MAP on the remaining day of surgery with postoperative high-sensitive troponin T levels¹²  and expands on our previous findings by augmenting with frequent blood pressure measurements, multiple thresholds, characterizations for severity, and duration to define postoperative hypotension.

In a substudy of the Vascular Events in Noncardiac Surgery Patients Cohort Evaluation (VISION), investigating the effects of withholding angiotensin-converting-enzyme inhibitors and angiotensin-II antagonists in noncardiac surgery, postoperative hypotension was investigated as a secondary objective.¹³  Similarly, the substudy of the Perioperative Ischemic Evaluation 2 (POISE-2) trial also investigated postoperative hypotension, but in a period-dependent matter in relation with postoperative MI and death.¹⁴  In both studies, postoperative hypotension was subjectively defined as clinically important hypotension, i.e., when systolic blood pressure dropped to less than 90 mmHg for any duration requiring intervention during postoperative day 0 to 3. In the VISION’s substudy, postoperative hypotension occurred less often in 19.5% of the patients and was associated with myocardial injury after noncardiac surgery (adjudicated as ischemicof origin, adjusted relative risk, 1.63; 95% CI, 1.4 to 1.8, P < 0.001). Postoperative hypotension during the remaining day of surgery occurred in 32% of the patients in POISE-2 and was associated with an increased risk of MI per duration of 10 min (adjusted odds ratio, 1.03; 99.2% CI, 1.0 to 1.05, P = 0.002). Differences in the incidence of hypotension with our study can be related to a younger, healthier population of both these cohorts, infrequent routine monitoring, and the use of a less sensitive fourth-generation troponin T assay.

Interestingly, we were unable to assess the association between intraoperative hypotension and postoperative myocardial injury described in previous studies.^5–7  This can be due to selection and treatment bias or due to the fact that the selected threshold of 65 mmHg might have been too high for our cohort (high-dependency ward patients at our center might be more vigorously treated). However, in the substudy of the VISION, the association of intraoperative hypotension and myocardial injury after noncardiac surgery was similarly no longer statistically significant after adjusting for postoperative hypotension (adjusted relative risk, 1.04; 95% CI, 0.9 to 1.2, P < 0.58). After adjustment for postoperative hypotension in the POISE-2’s substudy, the association of intraoperative hypotension per 10-min increase and MI did not remain (adjusted odds ratio, 1.03; 99.2% CI, 0.9 to 1.1, P = 0.162). These results parallel our findings demonstrating the robust association between postoperative hypotension and myocardial injury. Given the likelihood that patients who are hypotensive during surgery are also hypotensive after surgery,¹⁴  previous studies on intraoperative hypotension and myocardial injury might have been confounded by the effect of postoperative hypotension, which was not accounted for. Due to the complexity of the perioperative process influencing the patient’s blood pressure, it remains unknown whether the observed hypotension is refractory, a direct cause of adverse outcomes, a symptom of an underlying disease, or a sign of cardiac decompensation. Regardless of the cause, postoperative hypotension was associated with myocardial injury and remained associated after adding intraoperative hypotension to the models. The recently published Intraoperative Norepinephrine to Control Arterial Pressure (INPRESS) study investigated whether individualized blood pressure management during and in the early hours after surgery would lead to less postoperative organ dysfunction.⁴  The primary endpoint in this study was the dysfunction of at least one organ system of the renal, respiratory, cardiovascular, coagulation, and neurologic systems during the first 7 days after surgery. In this randomized controlled trial, an individualized strategy (aiming to achieve a systolic blood pressure within 10% of the reference value during and the first 4 h after surgery) was associated with decreased incidence in postoperative organ dysfunction (38.1% vs. 51.7%; relative risk, 0.73; 95% CI, 0.56 to 0.94, P = 0.02). Surprisingly, the incidence of myocardial injury or infarction in this high-risk population was extremely low (0.003%) compared to our study, which might potentially be explained due to their absence of systematic troponin surveillance. If blood pressure management during surgery might prevent perioperative organ injury, it is not unreasonable that postoperative management might also have the potential to improve the patient’s postoperative outcome.

One of our study’s strengths is exploring multiple MAP thresholds instead of one single cut-off threshold to define postoperative hypotension. Currently, there is no universal accepted definition of perioperative hypotension. The current consensus states both absolute MAP thresholds and relative MAP percentage reduction from baseline to be equally acceptable for the predictive value of myocardial injury when defining hypotension.¹⁶  We chose to report the absolute MAP threshold values for the ease of interpretation in clinical practice. Further strengths are the use of frequent blood pressure monitoring measurements of 1- to 15-min intervals during the postoperative period. Importantly, we were able to quantify multiple characterizations of hypotension, as myocardial injury is a function of both hypotension duration and severity.¹⁶  Both additional severity (area under MAP threshold) and averaged (time-weighted average under MAP threshold) characterizations were comparable with our main characterization of duration below the threshold, confirming the strength of our primary analysis.

In our study, several important limitations need to be considered. First, the recorded frequent postoperative blood pressures were from a high-dependency ward reflecting a higher-risk population, and the data were measured in one single university hospital, which limits generalizability. Second, since blood pressure readings during and after surgery were real-time displayed on the screens and visible to clinical staff, we cannot account for the fact that some blood pressure recordings were accepted by clinical judgment. Since no validation takes place of the recorded blood pressures, artifacts after applying a data cleaning filter may still be present. Third, we did not account for intravenous fluids, volatile anesthesia, inotropes, and vasopressors, which potentially have a substantial contribution to intraoperative hypotension, though this is also dependent on the institution’s protocols and the anesthetists’ individual preferences. Fourth, preoperative high-sensitive troponin T as well as the precision of the measurement were unfortunately unavailable; hence, we were unable to validate how many patients had elevated troponin concentrations before surgery. Fifth, although we have adjusted for potential confounders and conducted multiple sensitivity analyses, as with all observational data, our cohort is still accountable for residual confounding and bias. After all possible adjustments, hypotension after surgery is clearly associated with an increased risk for myocardial injury in our cohort. However, given the fact that myocardial injury is still common in patients without severe postoperative hypotension, residual unmeasured confounding should still be expected.

On surgical wards, patients’ vital signs are commonly monitored every 4 to 6 h, in which hypotensive episodes may occur more frequently. In a recent prospective blinded observational study of 312 patients, postoperative hypotension after abdominal surgery was measured with a continuous noninvasive monitor and compared to routine vital signs monitoring.²⁷  In this study, postoperative hypotension was also common and prolonged. More importantly, many of the hypotensive events were not detected by their institution’s standard routine monitoring. Given the association between postoperative hypotension and organ dysfunction, and the fact that hypotension is common and easily overlooked on surgical wards, further studies are warranted to understand this critical period. With the implementation of continuous monitoring, physicians may be able to detect those who are most vulnerable. Subsequently, clinical trials could further elucidate what factors contribute to postoperative hypotension and how to treat it effectively to improve outcome.

Postoperative hypotension during the first 24 h after noncardiac surgery on a high-dependency ward is common and associated with myocardial injury when decreasing from a MAP threshold less than 75 mmHg. Multiple characterizations and longer durations of postoperative hypotension were independently associated with myocardial injury at multiple thresholds.

Steven S. Roubos, B.N., clinical research coordinator, and Casper J. Steenweg, B.Sc., medical student, Department of Anesthesiology, Erasmus Medical Center, Rotterdam, The Netherlands, contributed to the clinical troponin registry by aiding in the acquisition and administration of the data.

This study was funded by Stichting Lijf en Leven and the Department of Anesthesiology, Erasmus Medical Center, Rotterdam, The Netherlands.

The authors declare no competing interests.