Relationship between Intraoperative Hypotension, Defined by Either Reduction from Baseline or Absolute Thresholds, and Acute Kidney and Myocardial Injury after Noncardiac Surgery: A Retrospective Cohort Analysis

THE perioperative period is characterized by hemodynamic instability. Various degrees of hypotension are common during anesthesia and surgery and may cause organ ischemia. For example, hypotension contributes to oxygen supply–demand mismatch, which appears to be an important cause of postoperative myocardial infarction.^1–3  Furthermore, ischemia and reperfusion may contribute to postoperative acute kidney injury (AKI).^4–10  Myocardial perfusion is dependent on pressure gradient created by diastolic blood pressure¹¹ ; vasomotor responses and regional ischemia in response to decreased blood pressure and cardiac output also contribute to ischemic renal injury.^12,13 

A systematic review of interventions to decrease the incidence of postoperative AKI demonstrated that avoiding hypotension reduced the incidence of AKI.¹⁴  Consistent with the theory that intraoperative hypotension contributes to organ injury, hypotension, defined in various ways, is weakly associated with AKI^8,10  and strongly associated with myocardial infarction^8,15  and death.^9,16 

How best to characterize hypotension remains unclear, and there is no universal definition of hypotension. In a systematic review, for example, Bijker et al.¹⁷  found 140 definitions for hypotension in 130 articles. A consequence was that the incidence of intraoperative hypotension ranged from 5 to 99% depending on the selected definition.

Several recent studies report associations between low mean arterial pressure (MAP) and organ injury, with hypotension defined in terms of minutes or integrated pressures below various absolute thresholds.^8–10,15  This approach differs from classical anesthesia teaching, which suggests keeping blood pressure within a relative 20% of preoperative values, apparently based on the theory that hypertensive patients require higher than normal pressures to adequately perfuse organs habituated to high pressures. Despite the frequency of this recommendation, it does not appear to be based on credible outcome evidence. Which characterization of blood pressure, absolute versus relative hypotension, is most related to organ injury remains unknown.

Therefore, we assessed the relationship between various absolute and relative characterizations of hypotension and myocardial injury after noncardiac surgery (MINS)¹⁸  and AKI in adults having inpatient surgery. Absolute thresholds were characterized by the lowest MAP maintained for various durations and by time under various MAP thresholds. Relative hypotension was characterized by maximum percentage MAP decrease from baseline maintained for various durations and by time under various percentage reductions from baseline. We then evaluated the interaction between preoperative MAP and the relationships between intraoperative hypotension and MINS or AKI. Finally, we determined whether absolute or relative characterizations best predict MINS and AKI.

We conducted a retrospective cohort study using data from the Cleveland Clinic Perioperative Health Documentation System and Epic, electronic medical record-based registries of noncardiac surgery patients who had undergone surgery between January 6, 2005, and March 1, 2014, at the Cleveland Clinic, Cleveland, Ohio.

Inclusion criteria were as follows: (1) adults who had inpatient noncardiac surgery between January 6, 2005, and March 1, 2014; (2) preoperative and at least one postoperative serum creatinine measurement available within the first 7 postoperative days; (3) blood pressure recorded in the preanesthesia care evaluation clinic or other preoperative appointments within 6 months before surgery.

Exclusion criteria were as follows: (1) patients with chronic kidney disease defined as preoperative estimated glomerular filtration rate of less than 60 ml × min⁻¹ × 1.73 m⁻² or patients who were on dialysis; (2) urologic procedures including relief of urinary obstruction (International Classification of Diseases, Ninth Revision [ICD-9] codes of 5501, 5502, 5503, 5504, 5511, 5512, 560, 570, 5741, 5749, 602, 6021, 6029, 6096, 6097, 603, 604, 605, 6061, 6062, and 6069), nephrectomy (ICD-9 codes of 554, 5551, 5552, 5553, and 5554), or renal transplantation (ICD-9 codes of 5561 and 5569); (3) patients who had anesthesia for less than 60 min or missing baseline variables; (4) patients with invalid or unavailable data for more than 10 consecutive minutes.

1. MINS was defined as at least one increased postoperative value of either fourth-generation troponin T or creatine kinase-MB above the upper limit of normal in the 7 days after operation. The upper limit of normal was defined as 0.03 ng/ml for troponin T¹⁸ and 8.8 ng/ml for creatine kinase-MB.³  Eligible patients without postoperative cardiac enzyme determinations were assumed not to have acute myocardial injury.

2. Postoperative AKI was defined by increases in serum creatinine between preoperative and postoperative values. Preoperative creatinine was taken to be the last before surgery. Postoperative creatinine was taken to be the highest concentration measured within 7 postoperative days. According to the Acute Kidney Injury Network definition, patients were considered to have AKI if the postoperative value was either more than 1.5-fold or more than 0.3 mg/dl before surgery.⁸ 

Intraoperative MAPs recorded in the Perioperative Health Documentation System cannot be modified by clinicians, but can be identified as artifactual. Invasive pressures were recorded at 1-min intervals; noninvasive pressures were recorded at 1- to 5-min intervals. We removed artifacts using the following rules, in order: (1) blood pressures documented as artifacts; (2) pressures out-of-range defined by (a) SBP greater than or equal to 300 or SBP less than or equal to 20 mmHg, (b) SBP less than or equal to DBP + 5 mmHg, or (c) DBP less than or equal to 5 mmHg or DBP greater than or equal to 225 mmHg; (3) abrupt changes defined by SBP change greater than or equal to 80 mmHg within 1 min in either direction or abrupt SBP changes greater than or equal to 40 mmHg within 2 min in both directions. Pressures between measurements were linearly interpolated.

Baseline MAP is described as the average of all MAP readings in the 6 months before surgery, excluding measurements during a hospital stay. Anesthesia time was defined as the interval between induction and emergence.

Potentially confounding variables are listed in table 1. We defined preexisting medical conditions using ICD-9 billing codes and included only those fulfilling at least one of the following: (1) appeared in the patient “problem list” with a date preceding the date of surgery; (2) appeared in an ICD-9 list before the index surgery; or (3) were flagged as a chronic ICD-9 condition based on Healthcare Cost and Utilization Project definitions. Because there were many types of surgical procedures, we characterized each procedure code into one of 231 clinically meaningful categories using the Agency for Healthcare Research and Quality’s Clinical Classifications Software for Services and Procedures.¹⁹  We then aggregated low-frequency event or nonevent categories (n < 10) into one group and used that as the reference group (a low-risk group).²⁰ 

We first determined the absolute and relative (percent below baseline) thresholds below which MINS and AKI began to increase. Specifically, we assessed the relationships between MINS or AKI and the lowest MAP or the lowest percent decrease from baseline for a cumulative case total of 1, 3, 5, and 10 min, and time-weighted average under absolute thresholds (i.e., less than 55, less than 60, less than 65, less than 70, less than 75 mmHg) or relative thresholds (i.e., greater than 10%, greater than 15%, greater than 20%, greater than 25%, greater than 30% decrease from baseline).

We first assessed the univariable relationship between each MAP threshold and MINS and AKI using moving-average smoothing plots. Relationships were then studied further using multivariable logistic regression to adjust for confounding and model the relationships; linearity between each MAP exposure and response was modeled by a restricted cubic spline function with three knots located at 10th, 50th, and 90th percentiles. The univariable moving-average plots and multivariable smoothed cubic spline curves were studied to find optimal thresholds based on the data. We further evaluated interactions between baseline MAP and the relationship between exposure and outcome.

Based on inspection of exposure versus outcome curves, we determined that absolute thresholds of 65 mmHg and lower and relative thresholds of 20% or more decrease from the baseline MAP were associated with the increased risk of both MINS and AKI. We then defined our main absolute and relative exposures to be (1) number of minutes under each threshold and (2) area under each threshold. Since all relationships were found to be nonlinear, we categorized patients as belonging to either a reference group who spent no time under a given threshold or to one of four groups based on quartiles of nonzero time spent under the threshold.

Specifically, we defined the absolute MAP reference group as patients whose intraoperative MAPs were never less than 65 mmHg. For the remaining patients, we counted the number of minutes within the lowest achieved category per patient: less than 50, 50 to 55, 55 to 60, and 60 to 65 mmHg. That is, each patient was assigned uniquely to one of the four hypotension categories. We then categorized cumulative minutes of exposure into 1, 2 to 4, or greater than equal to 5 min for a total of 12 groups (i.e., four pressure ranges by three durations) and compared each to the reference group.

We similarly defined the relative MAP reference group as patients whose intraoperative MAPs were never more than 20% below the preoperative reference pressure. For the remaining patients, we counted the number of minutes within the lowest achieved category per patient: 20 to 30%, 30 to 40%, 40 to 50%, and greater than 50% below baseline. Thus, each patient was again assigned uniquely to one of 12 groups (i.e., four pressure ranges by three durations) and compared each to the reference group.

Multivariable logistic regression was used to assess the association between the above MAP exposures and postoperative MINS or AKI. All potentially confounding variables listed in table 1 were forced into the models regardless of statistical significance. Bonferroni correction was used to adjust for four main comparisons within each exposure of interest, with P < 0.0125 (i.e., P < 0.05/4 = 0.0125) considered statistically significant. Interactions between baseline MAP and exposures were considered significant if P < 0.05. All analyses were performed with the use of SAS software, version 9.4 (SAS Institute, USA).

We expected to have between 50,000 and 150,000 patients meeting all study criteria. With at least 50,000 patients and the incidence of MINS or AKI of 2% or more, we had good statistical power (80% or more) to detect moderately small odds ratios, especially given the continuous/ordinal nature of the predictor variables.

Of 164,514 patients having noncardiac surgery between 2005 and 2015, analysis included 57,315 patients who met our inclusion and exclusion criteria (fig. 1). Different subsets of these patients were included in studies by Walsh et al.⁸  and Mascha et al.¹⁶  The overall incidence of MINS was 3.1% and of AKI was 5.6% among qualified patients. Only 8,558 patients (15%) had postoperative troponin tests, and we assumed that patients without the test did not have MINS.

Nearly all demographic, medical history, procedural, medicine, preoperative, and intraoperative factors were associated with both MINS and AKI (table 1). Descriptive statistics for baseline MAP and all MAP exposures are displayed in table A1. Baseline MAP was based on a mean of 5 ± 3 values per patient in the 6 months before surgery. Average baseline MAP was 93 ± 10 mmHg; preinduction MAP averaged 101 ± 16 mmHg, and intraoperative time-weighted average MAP was 84 ± 10 mmHg.

Univariable analyses showed that patients having postoperative MINS or AKI had higher time-weighted average, area under threshold, and number of minutes under all thresholds compared to those with no evidence of AKI or MINS (all P < 0.001; table 2).

Univariable moving-average and multivariable spline smoothing plots for the lowest observed MAPs for a patient are shown for MINS in fig. 2 and for AKI in fig. 3. Odds for both MINS and AKI increased for decreasing thresholds of MAP less than 65 mmHg for any of 1, 3, 5, or 10 min. A relative MAP threshold of 20% below baseline was not an obvious change-point for AKI (fig. 3), but it was for MINS (fig. 2). We thus selected an absolute reference threshold of 65 mmHg and a relative reference threshold of 20% below baseline for further analysis.

Increasing time-weighted average MAP under various absolute and relative thresholds was associated with increased odds of MINS (fig. 4) and AKI (fig. 5), both univariably and multivariably. Further, the relationships strengthened at lower thresholds. For example, the observed slope for less than 60 mmHg is steeper than that for less than 65 mmHg, and the observed slope for less than 25% below baseline is steeper than that for pressures less than 20% below baseline (figs. 4 and 5).

There was no interaction between baseline MAP and the relationship between the TWA under various relative thresholds for either MINS or AKI. Furthermore, there was no interaction with TWA under absolute thresholds for AKI (all P > 0.40). There was some evidence of interaction between baseline MAP and the relationship between TWA under absolute thresholds and MINS (table 3). Investigating further, univariable moving-average and multivariable spline smoothing plots by quartile of baseline MAP showed that there were no clinically important interactions at MAPs less than 65 mmHg (fig. 6).

Using different absolute or relative thresholds did not increase discriminative ability, as evidenced by similar C-statistic values. The full multivariable model for MINS (table A2), including all baseline and intraoperative covariables mentioned in table 1, has a C statistic of 0.86. In contrast, blood pressure alone had a C statistic between 0.55 and 0.66. Whether hypotension exposure was defined by cumulative minutes of the lowest MAP (0.62 to 0.65) or duration under various thresholds (0.55 to 0.62), the C statistic values were essentially the same for absolute and relative exposures. There was thus no advantage to using relative thresholds for myocardial injury.

For AKI, the full multivariable model, including all baseline and intraoperative covariables mentioned in table 1, had a C statistic of 0.81 (table A3). In contrast, blood pressure alone had a C statistic between 0.54 and 0.59. Whether hypotension exposure was defined by cumulative minutes of lowest MAP or duration under various thresholds, the C statistic was nearly identical for absolute and relative thresholds. There was thus no advantage to using relative thresholds for AKI either.

Time spent under the absolute threshold of MAP less than 65 mmHg had increased odds of MINS, with an odds ratio (OR; 98.75% CI) of 1.34 (1.06 to 1.68) for the third quartile and 1.60 (1.28 to 2.01) for the fourth quartile (table 4). Results were similar when hypotension exposure was characterized by area (rather than minutes) under absolute thresholds. In contrast, there were no significant associations between minutes or area under the relative threshold of 20% below baseline and MINS. Hypotension exposure was also characterized by various blood pressure ranges and exposure durations within each range. For instance, MAP less than 50 mmHg for at least 1 min or a 50% decrease from baseline for at least 1 min increased the odds of MINS after Bonferroni correction.

Time spent under the absolute threshold of MAP less than 65 mmHg had increased odds of AKI compared to patients never going less than 65 mmHg, with an OR (98.75% CI) of 1.20 (1.02 to 1.40) for the third quartile and 1.35 (1.14 to 1.58) for the fourth quartile (table 5). When hypotension exposure was characterized by area (rather than time) under absolute thresholds, odds were higher than reference only for the fourth quartile, with OR (98.75% CI) of 1.34 (1.15 to 1.58). For a relative threshold of 20% below baseline, again the fourth quartile had significantly higher odds of AKI with OR (98.75% CI) of 1.27 (1.01 to 1.61). The lowest hypotension exposure was also characterized by various blood pressure ranges and by exposure durations within each range. For instance, absolute categories of 50 to 55 mmHg for at least 1 min and less than 50 mmHg had higher odds of AKI compared to those never less than 65 mmHg. A relative decrease of greater than 50% from baseline MAP had higher odds of AKI compared to those never reaching less than 20% of baseline.

We first characterized hypotension exposure by the lowest MAP maintained for various durations and by time under various absolute MAP thresholds. MAP less than 65 mmHg for greater than equal to 13 min (characterizing 50% of the patients who ever went less than 65 mmHg) was associated with significantly higher odds of myocardial and kidney injury. Injury was more common at lower absolute thresholds, and when hypotension was prolonged. At a MAP of 50 mmHg, for example, just 1 min significantly increased the risk for both myocardial and kidney injury.

Our results are broadly consistent with the results of previous reports. Based on previous studies, MAP less than absolute thresholds of 49 to 60 for various durations ranging from 1 to 30 min increases the risk of myocardial and kidney injury and mortality.^8–10,15,16  Available analyses thus suggest that even short periods of hypotension below MAPs thresholds of 50 to 65 mmHg are associated with kidney and myocardial injury. While causality cannot be determined from analysis of purely observational data, all results suggest that anesthesiologists should avoid unnecessary hypotension. In this context, it is sobering that therapeutic hypotension was used for decades—often for nonessential reasons.

We also characterized hypotension exposure by time under various relative MAP thresholds. Injury was more common at lower absolute thresholds, and when hypotension was prolonged. For example, a cumulative time exceeding 90 min (highest quartile of patients) with MAP less than 20% below preoperative values was needed to increase the odds of kidney injury, and total minutes less than 20% was not significant for myocardial injury. When MAP was more than 50% below preoperative values, just 5 min significantly increased the risk for both myocardial and kidney injury.

Again, our results are broadly consistent with the results of previous reports. Monk et al.⁹  showed that blood pressure measurements less than 50% below baseline was associated with increased 30-day mortality although their analysis was limited in that one third of their patients lacked baseline blood pressures. Van Waes et al.¹⁵  showed that a relative decrease in MAP to values less than 40% below preinduction blood pressure for more than 30 min was associated with the increased incidence of myocardial injury. Available analyses thus suggest that sufficient time with pressures less than 20% or even short periods of hypotension to less than 40 to 50% below preoperative MAPs are associated with kidney and myocardial injury. The classical teaching that intraoperative pressures should be maintained within 20% of preoperative values thus appears justified.

The interaction between preoperative blood pressure and the relationship between intraoperative blood pressure and postoperative outcome was evaluated by Levin et al.²¹  They found that hypertensive patients had more intraoperative blood pressure lability and that lability decreased mortality. In our study, however, there was no interaction between baseline pressure and the relationship between intraoperative hypotension and AKI. Intraoperative hypotension was thus proportionately related to AKI over the entire range of preoperative pressures.

In contrast, there was a significant interaction between baseline pressure and the relationship between intraoperative pressure and myocardial injury. However, the interaction was only substantive at intraoperative MAPs exceeding 65 mmHg. In the clinically relevant range of hypotensive pressures less than 65 mmHg, there was no important interaction. Preoperative blood pressure thus had no important effect on the relationship between intraoperative hypotension and myocardial injury.

From a clinical perspective, our interaction analysis thus indicates that anesthesiologists can manage intraoperative blood pressure without reference to preoperative values—a conclusion that differs starkly from classical anesthesia teaching that patients with high preoperative pressures should be maintained at relatively high pressures throughout surgery. A caveat, of course, is that we evaluated only two organs. It remains possible that preoperative pressures do matter for the brain and other physiologic functions such as gut permeability.

A novel aspect of our study is comparison between absolute and relative thresholds. Both were predictive. However, there was no advantage to using relative over absolute thresholds for AKI or myocardial injury. Absolute thresholds are easier to use since a reliable baseline pressure is not required. Furthermore, absolute thresholds are far easier to incorporate into decision support systems that would not normally have access to individual preoperative reference values. Therefore, we conclude that clinicians can use absolute thresholds to guide intraoperative blood pressure management.

We defined myocardial injury on the basis of increased cardiac enzymes. However, cardiac enzymes were not routinely measured even in relatively high-risk patients during the study period. Consequently, our analysis was mostly based on clinically apparent myocardial infarctions, thus underestimating the actual incidence of myocardial injury by about a factor-of-three.¹  Whether the relationships between hypotension and myocardial injury that we report apply comparably to silent injury remains unknown. However, the physiology is probably similar, suggesting that the relationships are probably similar.

As in any retrospective analysis, confounding and bias are concerns. For example, patients who experienced MINS or AKI were generally sicker and had more preoperative comorbidities. However, our large sample size and detailed registry allowed us to statistically adjust for many potential confounding factors. Our results are nonetheless surely somewhat degraded by both unknown and known but poorly characterized confounders. The extent to which either contributes is hard to assess.

About 60% of our patients had blood pressure measured oscillometrically at 1- to 5-min intervals. We linearly interpolated between measurements to provide reasonable estimates of intervening values, but is obviously less accurate than values from arterial catheters that were available at 1-min intervals. It seems unlikely that more frequent measurements would much change the harm thresholds we identified.

Pressures that until recently were considered clinically acceptable, for instance, a MAP of 65 mmHg, were associated with both myocardial and renal injuries. At lower pressures, the association was stronger and only brief exposures were required. Associations based on relative thresholds were no stronger than those based on absolute thresholds. Furthermore, there was no clinically important interaction with preoperative pressure. The extent to which the associations we observe are causal remains to be determined. But to the extent that they are, a strategy aimed at maintaining MAP above 65 mmHg appears to be as good as one based on the percentage reduction from baseline. This result is fortuitous because absolute thresholds are easier to use in that they do not require a reliable baseline pressure and can thus more easily be incorporated into decision support systems. While retrospective analyses cannot assess causality, our results suggest that maintaining intraoperative MAP greater than 65 mmHg may reduce the risk of AKI and myocardial injury—the leading cause of 30-day postoperative mortality.

Support was provided solely from institutional and/or departmental sources.

The authors declare no competing interests.