Reich, David L. MD; Hossain, Sabera MA; Krol, Marina PhD; Baez, Bernard MD; Patel, Puja; Bernstein, Ariel; Bodian, Carol A. DrPH
The influence of hemodynamic aberrations during anesthesia on adverse outcomes is an important clinical issue. There is evidence that hypotension and hypertension during general anesthesia are independently associated with adverse outcomes in patients having both noncardiac and cardiac surgery (1,2).
One of the intervals of general anesthesia during which hypotension is prevalent is the period after the induction of anesthesia but before the onset of surgical stimulation. This period is particularly prone to decreased vigilance with regard to hemodynamic changes and inaccuracies in handwritten anesthesia records, probably because of increased workload for anesthesia practitioners (3,4). The advent of computerized anesthesia information management systems, with unbiased and automated data collection process, has enabled the accurate measurement of hemodynamic trends during anesthesia (5–8).
We hypothesize that choice and dosage of IV anesthetic drugs influence the occurrence of hemodynamic instability in the period after anesthetic induction. To test this hypothesis, we undertook a retrospective database review to identify the predictors of hemodynamic instability in the 10-min period after the induction of general anesthesia. To consider the clinical relevance of hemodynamic instability, we also explored these data for the association of postinduction hemodynamic instability with adverse outcomes.
Under an IRB-approved protocol, computerized anesthesia records were queried for arterial blood pressure (BP), demographic, and drug data and stripped of identifying information so that the data were not identifiable to source, in compliance with privacy regulations. Computerized records for the period of December 1999 through March 2000 were analyzed, resulting in 5244 records with at least one BP recording before the induction of general anesthesia, where general anesthesia was the primary anesthetic technique.
BP recordings were obtained automatically from oscillometric BP cuffs or intraarterial lines and transducer systems and recorded by the CompuRecord anesthesia information management system (Philips, Andover, MA). Of all BP recordings before the induction of anesthesia (the preinduction period), the median mean arterial blood pressure (MAP) was chosen as the most representative number. Similarly, the median of all MAP recordings was determined for the first 5-min interval after the induction of anesthesia (0–5 min postinduction) and again for the second 5-min interval (5–10 min postinduction). The median numbers were chosen as the best method for filtering out artifactual data. Furthermore, any MAP <30 mm Hg or >200 mm Hg was excluded to further guard against artifactual data. The data were scrutinized for outliers, but no cases were eliminated on this basis beyond those with missing BP data during the postanesthetic induction intervals.
Demographic data, including age, ASA physical status, emergency status, gender, and weight were also extracted from the anesthesia records. Anesthetic drug totals were measured for the period beginning 2 min before and ending 2 min after the time recorded on the anesthesia record as the anesthetic induction time. The sum total of these medications was considered the anesthetic induction regimen. The medical record numbers of all patients were checked in the hospital admission/discharge/transfer computer system for duration of postoperative hospital stay and hospital discharge status.
We restricted the analysis to commonly used IV anesthetic induction regimens in adult patients. Specifically, we excluded all patients who underwent an inhaled induction of anesthesia, patients <18 yr of age, patients who had received more than one of the usual drugs to induce anesthesia (i.e., etomidate, thiopental, or propofol), or rarely used regimens (e.g., ketamine). Large dose opioid-benzodiazepine inductions were also excluded. This resulted in 4096 records that were included in the final analysis.
The doses of drugs commonly used for IV induction of anesthesia and adjuvants (i.e., propofol, etomidate, thiopental, midazolam, and fentanyl) were normalized to body weight and classified into groups according to dosage range. These data were examined for an association of the dose level with hypotension, and categories were combined accordingly.
For every case, an investigator classified the patient’s chronic preoperative medications into 10 categories: diuretic, β-adrenergic blocker, calcium channel blocker, angiotensin-converting enzyme inhibitor, angiotensin receptor blocker, other antihypertensive, thyroid replacement, glucocorticoid, antidepressant, or hypoglycemic. Some of these categories were chosen based on the preoperative drug history of a sample of patients from a different time interval whose BP had declined precipitously after the induction of anesthesia (i.e., thyroid replacement, glucocorticoid, and antidepressant) whereas others were chosen as proxies for the underlying conditions of diabetes mellitus and hypertension.
A χ2 test was used to test for an association of hypotension during either 0–5 min or 5–10 min after anesthetic induction with adverse outcomes. For the purpose of this analysis, adverse outcome was defined as a postoperative hospital length of stay >10 days or mortality during hospital stay (9).
For the data analysis, the cases were categorized based on the median MAP during each of the two postinduction intervals and their relationships to the preinduction median MAP. All MAPs analyzed were the medians for the respective intervals. Clinically important hypotension was defined according to the following criteria:
1. Preinduction: MAP <70 mm Hg.
2. Postinduction: A MAP decrease >40% and a MAP <70 mm Hg; or a MAP <60 mm Hg.
Clinically important hypertension was defined according to the following criteria:
1. Preinduction: A MAP >110 mm Hg.
2. Postinduction: A MAP increase >40% AND a MAP >110 mm Hg.
χ2 tests, including Fisher’s exact tests and tests for trend, were used for separate tests of association among demographic factors, preoperative drug therapy, and anesthetic drug doses (normalized to body weight) with hypotension and with hypertension. Initial examination of these results revealed no clinically important associations of potential predictor variables with hypertension, so no further analyses of predictors of hypertension were undertaken. For hypotension, initial examination of the univariate tests showed a strong influence of the patient’s ASA physical status classification on outcome and on other predictors. For this reason, the univariate and multivariate analyses were performed separately for patients with ASA status I–II, and for patients with ASA status III-V.
For the multivariate model, we tested for predictors of hypotension at any time during the first 10 min after induction of anesthesia. Stepwise multiple logistic regression was initially performed for demographic variables and preinduction MAP and the best model were selected. The anesthetic induction regimen was added as two additional variables: propofol versus other induction drugs, and fentanyl dosage. The ultimate multivariate model that was selected was used to calculate probabilities of hypotension for persons with specified values of the risk factors.
There were 2962 patients with ASA physical status classification I–II and 1134 patients with ASA physical status classification III–V. The prevalence of hypotension is presented in Table 1. In both ASA class groupings, the incidence of hypotension was more frequent during the 5–10 min postinduction period compared with the 0–5 min period (P < 0.001 for each) (Table 2).
As described above, the univariate analyses were performed separately for patients with ASA physical status classification I–II and for patients with ASA classification III–V. The univariate associations of variables with hypotension at 0–5 and 5–10 min after the induction of anesthesia are presented in Tables 3 and 4.
To explore the influence of drug dose on the incidence of hypotension, we compared the incidence in the subgroups of patients who received doses more than and less than the median dosage. The median dosages were as follows: for propofol, 2.4 mg/kg; for etomidate, 0.24 mg/kg; for thiopental, 4.7 mg/kg. Fentanyl dosage was divided according to clinical criteria into small (0–1.50 μg/kg), medium (1.51–5.0 μg/kg), and large (>5.0 μg/kg) dosage groups. Fentanyl was the only drug that demonstrated a consistent dose effect on the incidence of hypotension (Table 5). Further analysis was therefore restricted to IV anesthetic induction drug independent of the dosage, with the exception of fentanyl. Potential predictors were defined as those with P values <0.20 and are detailed below.
The potential preoperative factors associated with hypotension 0–5 min after the induction of anesthesia in patients with ASA I–II included baseline MAP <70 mm Hg, age ≥ 50 yr, use of propofol during induction, and use of midazolam during induction. The potential preoperative factors associated with hypotension 5–10 min after the induction of anesthesia in patients with ASA I–II included baseline MAP < 70 mm Hg, age ≥ 50 yr, use of propofol during induction, and magnitude of fentanyl dose during induction.
The potential factors associated with hypotension 0–5 min after the induction of anesthesia in patients with ASA III–V included preoperative β-blockade, preoperative thyroid replacement therapy, baseline MAP <70 mm Hg, and use of propofol during induction. The potential preoperative factors associated with hypotension 5–10 min after the induction of anesthesia in patients with ASA III–V included male gender, age ≥50 yr, and propofol use during induction.
In separate analyses of ASA I–II and ASA III–V patients, we found the same predictors in both sets of patients with the exception of fentanyl dosing (which was only significant in the ASA I–II group). When the groups were combined, the final model included ASA status, baseline MAP <70 mm Hg, age >50 yr, fentanyl dosing, and propofol induction. There was no significant interaction effect between ASA status and fentanyl dosing. (Table 6).
Based on the findings in Table 6, combinations of independent predictors with a calculated probability of hypotension more than 10% are shown in Table 7. In the absence of propofol induction, a high calculated probability of hypotension was obtained when the patient’s baseline MAP was low, the patient was more than 50 yr old, and the patient was either ASA III–V or had received medium or large-dose fentanyl. With normal or high baseline MAP, a high calculated probability of hypotension occurred when the patient was older than 50 yr of age, had received propofol, and was either ASA I–II with large-dose fentanyl, or ASA III–V with medium- or large-dose fentanyl. Within the current data set, 841 of 4096 (21%) patients had combinations of independent predictors that were associated with a calculated probability of postinduction hypotension exceeding 10%
In 2406 patients with retrievable outcome and hemodynamic data, the proportion of adverse events (prolonged postoperative stay and/or death during hospitalization) in patients with hypotension at any time during the first 10 min after anesthetic induction was 13.3% versus 8.6% of patients without postinduction hypotension (P = 0.012). We performed a multivariate analysis, which confirmed that the relationship was still present (P < 0.02), controlling for the effects of age, hypotension before induction of anesthesia, anesthetic induction drug, and ASA status.
Exploration of the data in detailed 2×2 tables, however, showed that the effect of early postinduction hypotension on morbidity and mortality seems to arise mainly from the group of patients that had ASA status III and were <65 yr old. Of note, the statistical analyses revealed less morbidity and mortality among the propofol patients compared with those in the etomidate/pentothal group, and the use of propofol decreased with increasing ASA physical status scores (P < 0.0001, χ2 test for trend). Therefore, we have a constellation of overlapping risk factors that complicates definitive multivariate examination of the data. Nevertheless, there is preliminary evidence that hypotension 0–10 min after anesthetic induction may have an independent association with morbidity and/or mortality in some patient groups.
Severe hypotension after induction of anesthesia is quite common and is more prevalent during the late postinduction period (5–10 minutes after induction) than at other times. The incidence of hypotension postinduction of anesthesia is strongly predicted by age ≥50 years, hypotension before induction, and propofol used at induction of anesthesia. Increasing fentanyl dosage during the anesthetic induction also appears to have a significant association with hypotension in patients with less systemic disease (ASA physical status I–II).
Hypotension after induction of anesthesia is a process variable that is commonly thought to be clinically irrelevant. We found a statistically significant association between postanesthetic induction hypotension and death or morbidity (prolonged hospital stay postoperatively) in the subset of patients who underwent surgery as inpatients or were admitted to the hospital after surgery. This retrospective analysis may be limited by the absence of other risk factors that are critical to outcomes. Also, the influence of milder degrees of hypotension was not analyzed. Nevertheless, the commonly held view that transient marked hypotension after anesthetic induction is inconsequential to outcomes is challenged by these data.
Our results are difficult to compare with others’ results because this topic is not well addressed in previous research. Hug et al. (10) performed a retrospective analysis of Phase IV trials of propofol use in 25,981 patients from 1722 institutions; the patients were 18–80 years of age and ASA physical status I–III. In their analysis, 15.7% of patients had a systolic BP <90 mm Hg and 77% of these episodes occurred within 10 minutes of anesthetic induction. That study differs from the present one in that ASA IV–V patients, emergency procedures, and patients with low baseline MAP were excluded. Benson et al. (11) investigated changes in MAP, heart rate and Spo2 in 8078 patients receiving anesthetic regimens similar to those of the present study. They reported that propofol was associated with the largest reduction in MAP after anesthetic induction with ASA class >II. The use of a different statistical approach in that study did not permit them to analyze the predictors of severe BP changes, but the magnitude and direction of changes after anesthetic induction drugs are consistent with the results of the current study.
We have previously found associations between intraoperative hypotension and adverse outcomes in cardiac surgical and liver transplantation patients (12). Previous work suggesting that angiotensin-converting enzyme inhibitors are associated with hypotension was not confirmed in the present study (13,14). Yet we found no independent associations between preoperative drug therapy (including a wide variety of antihypertensive drugs) and postinduction hypotension, despite early literature suggesting that intraoperative BP lability was more prevalent in poorly controlled hypertensive patients (15). A possible explanation for the lack of association is that the effect of ASA physical status classification as a measure of severity of systemic disease was a much stronger predictor than preoperative systemic hypertension (using preoperative drug therapy as a surrogate for the diagnosis).
As a retrospective review, this study suffers certain limitations. We have no way of knowing whether prolonged or difficult airway instrumentation, fluid therapy, positional changes, or a variety of other potential factors influenced these results. In these university hospital cases, it is very unlikely that surgical stimulation occurred during the 10 minutes after anesthetic induction, but mildly stimulating procedures, such as urinary bladder catheterization or examinations under anesthesia, may have occurred. If there was any effect of noxious stimuli during this period, it would be expected to decrease the incidence of severe hypotension.
The challenges in analyzing hemodynamic data of this type include artifacts, limited numbers of observations (i.e., BP recordings), and different measurement modalities (i.e., noninvasive versus invasive). It is therefore impossible to capture the true hemodynamic state throughout the continuity of the postinduction interval. To characterize the presence of a severe hypotensive or hypertensive exposure during a large part of the postinduction period, we chose the median BP. For example, a median MAP of 50 mm Hg for the period 0–5 minutes postinduction means that at least half of the BP recordings during that interval were at or less than 50 mm Hg.
If we had used the nadir or maximum BP, rather than the median MAP, as the outcome of interest, the incidence of BP abnormalities would have been more frequent. The anesthesia practitioners were cognizant of these BP abnormalities in nearly all cases as a result of monitoring alarms and may have treated them in various ways, such as with inhaled anesthetics, IV fluid challenges, and vasoactive drugs. We considered successfully treated BP abnormalities as “normals” in this study because we presumed that more persistent BP abnormalities would have more clinical significance. Therefore, this study does not address the predictors or clinical impact of brief and successfully corrected BP abnormalities after anesthetic induction.
The period before induction of general anesthesia is a time of anxiety, and the baseline BP may have been higher than normal for most patients despite the use of anxiolytic therapy (midazolam) in the majority of cases. We attempted to control for increased baseline BP by using the following criteria to define hypotension: (a) BP decrease >40% AND MAP <70 mm Hg; or (b) MAP <60 mm Hg.
In conclusion, 9% of patients experienced clinically important hypotension in the period 0–10 minutes after anesthetic induction in common clinical practice. A better understanding of the predictors and consequences of these BP abnormalities should enhance the care of patients undergoing general anesthesia. Given these preliminary data, it is prudent to consider alternatives to using propofol to induce anesthesia in patients older than 50 years of age with ASA physical status ≥III. It is prudent to avoid propofol induction in patients who present with baseline MAP <70 mm Hg. The incidence of hypotension may be less when smaller doses of fentanyl are used with propofol for the induction of general anesthesia.