Angiotensin converting enzyme (ACE) inhibitors and angiotensin receptor antagonists (ARAs) are used to treat hypertension, congestive heart failure (CHF) and myocardial infarction (MI), reducing mortality and improving quality of life.1 Inhibitors of ACE and ARAs are popular and effective with over 130,000 patients studied in randomized controlled trials (RCTs).2 However, the conclusions of these RCTs might not be applicable to the situation of surgical patients, because none of the RCTs examined the effects of anesthesia and perioperative care.

Perioperative blockade of the renin-angiotensin system (RAS), defined as preoperative exposure to ACE inhibitors and ARAs, remains controversial because the area has been poorly studied. The 2007 American College of Cardiology – American Heart Association guidelines3 advise caution using ACE inhibitors perioperatively in renally impaired patients citing increased risk of renal failure.4 Some physicians recommend holding ACE inhibitors or ARAs prior to surgery to prevent intraoperative refractory hypotension.5,6

Anesthetized patients depend heavily on the RAS to maintain blood pressure.7 Hypovolemia, hyponatremia, and anemia activate the RAS to maintain blood pressure.8 Intraoperative hypotension induced by ACE inhibitors has been reported in one large study9 and in a meta-analysis.10

The ACE inhibitors have been studied in patients with CHF undergoing vascular surgery.11 Correction for thirteen medical factors showed that ACE inhibitor exposure was protective in patients with left ventricular dysfunction (odds ratio 0.33; 95% confidence intervals [CI95] 0.12-0.91; P = 0.03).11 However, Miceli et al. have shown that perioperative ACE inhibitor therapy is associated with a doubling of the mortality rate in patients undergoing coronary artery bypass grafting surgery (CABG).12 The purpose of this study was to test the hypothesis that preoperative RAS blockade is safe in vascular surgery patients, using 30-day postoperative mortality as the primary end point.

Methods

Following Research Ethics Board approval, consent was obtained from the attending surgeons to review the medical charts of patients undergoing abdominal aortic aneurysm (AAA) repair at two teaching hospitals from January 1, 1998 to December 31, 2005. Patients were identified for possible inclusion into the cohort based on operative bookings and the following inclusion criteria: age > 50 yr, elective AAA repair, or urgent surgery with adequate medical workup. Patients were excluded for the following reasons: trauma, emergency surgery, sepsis, combined AAA and non-vascular surgery, and surgery on bypass.

One physician reviewed all charts and abstracted information onto data sheets. Data were entered into Microsoft Excel 2003 (Microsoft Corp., Redmond, OR, USA) using Filemaker Pro 5.0 (Filemaker Inc., Santa Clara, CA, USA) computer software for conversion to a data file for statistical analysis.

The information collected included age, sex, medical history (cardiac, respiratory, renal, neurological, and endocrine), preoperative consultation and testing, vital signs, weight, renal function, preoperative medications, surgeon, and anesthetic techniques. Postoperative complications were also entered: death, arrest, arrhythmia, CHF, increased troponin, renal failure, stroke, and other complications.

Medications were grouped using the following classes: ACE inhibitors, ARAs, beta-blockers, calcium channel blockers (CaCB), diuretics, nitrates, non-steroidal anti-inflammatory drugs (NSAIDs), cyclooxygenase II inhibitors (COX-2 s), and HMG-CoA inhibitors (statins). An assumption was made that the patients were taking the medications as directed by the preoperative assessment clinic staff. The normal recommendation was to take all cardiac medications with the exception of diuretics, which were held on the day of surgery; NSAIDs and COX-2 inhibitors were held for five days prior to surgery.

A patient was considered to have a given medical condition under any of the following circumstances: self report, positive diagnostic testing, diagnosis in old charts, and report of the condition in preoperative consultations or in the attending anesthesiologist’s note. Creatinine clearance (Cr Cl) was calculated using the Cockcroft-Gault formula.13

Postoperative complications were recorded based on physician diagnosis recorded in the medical chart. Infections were diagnosed based on positive laboratory culture results, even if not recorded in the medical notes. A postoperative MI – increased troponin was recorded as the result of a physician diagnosis of MI or a laboratory troponin I concentration > 0.3 ng.L−1 and either patient symptoms or ECG changes.

Study end point and sample size

Thirty-day mortality was the primary end point used to assess the safety of RAS blockade. A pilot study of the first 122 charts was undertaken. Seventy-one patients met the inclusion criteria with an observed death rate of 9.8% (7/71 patients), 29% in ACE/ARA exposed patients (5/17) vs 3.7% in unexposed patients (2/54) (P = 0.002). However, the subjects were not well matched. The observed death rate was greater than the expected rate of 4%. There was considerable uncertainty in the estimated mortality rate, especially in RAS blocked patients.

Modelling a range of death rates using PASS 2002 (NCSS LLC, Kaysville, UT, USA) showed that 780 pairs would be needed to detect a difference of 4% in death rate associated with RAS blockade (power 0.8; alpha = 0.05), assuming 15% of pairs were discordant. This number would decrease to 350 pairs if the difference in death rate was increased to 6%. It was decided to review eight years of consecutive cases (approximately 1,300 charts) expecting to find approximately 900 to 1,000 patients comprising cohort for analysis.

Statistical analysis

Data was analyzed using NCSS 2007 (NCSS LLC, Kaysville, UT, USA) and SAS 9.1 (SAS Inc, Cary, NC, USA). Continuous data variable differences were tested by Student’s t test. Dichotomous data differences were tested by Chi square test or Fisher’s modification. SAS 9.1 was used to perform the propensity score match and subsequent statistics on the matched data set, and the McNemar score was used to assess differences between exposed and unexposed subjects. Odds ratios and confidence intervals for matched pairs were estimated using a standard method.14 Comparisons of 30-day mortality (primary end point) and major postoperative complications were planned at the start of the study. Data are reported as mean ± standard deviation unless otherwise indicated.

To avoid selection bias and confounding factors, RAS blocked subjects and unblocked patients were matched using a propensity score system.15 A logistic regression analysis of the data set (883 patients and 49 possible factors) using SAS 9.1 (SAS Inc., Cary, NC, USA) identified predictors (all P < 0.05) of RAS blockade: hypertension, stroke, CHF, diabetes, and heart disease (a combined term comprised of any history of previous MI, previous CABG, arrhythmias, and/or angina). A second logistic regression analysis of the data set identified predictors (all P < 0.05) of mortality: age, creatinine, diabetes, ACE/ARA, and diuretic exposure. Factors affecting perioperative mortality were considered for inclusion in the model, especially those differing between the RAS blocked and unblocked groups. The model developed predicting preoperative RAS blockade included the following terms: age, sex, creatinine, CHF, heart disease, hypertension, stroke, and preoperative cardiovascular medication exposures (beta-blockers, CaCB, diuretics, and HMG-CoA reductase inhibitors [statins]). A propensity score (from 0 to 1) representing the likelihood of RAS blockade was calculated for each subject. An unexposed subject was matched to an RAS blocked subject using the MEANS procedure of the SAS 9.1 program (propensity score within 0.1). In 19 patients, the preoperative serum creatinine was estimated as 103.15 μmol·L−1.

Results

Subjects

According to operating room bookings, 1,320 patients underwent AAA repair from January 1, 1998 to December 31, 2005. Manual review of charts identified 883 subjects meeting inclusion criteria. There were 437 exclusions for the following reasons: other vascular procedure performed (259), emergency surgery (157), age < 50 yr (8), missing charts (4), cancelled surgery (2), and other (7). Subject characteristics, medical history, and the results of the propensity matching process are summarized in Table 1.

Table 1 Patient characteristics – unmatched data and propensity score matched subjects
Full size table

Preoperative medication exposures

Examination of preoperative medication exposures revealed that 306 of 883 subjects (34%) were exposed to ACE inhibitors, and 60 of 883 subjects (7%) were exposed to ARA agents. Eight subjects were exposed to both ACE inhibitors and ARAs. The exposures of subjects to different classes of medications are reported in Table 2, including propensity score matched subjects. The details of the propensity matching process are presented in Table 3.

Table 2 Preoperative medications
Full size table
Table 3 Matched pairs analysis – distribution of concordant and discordant matched pairs for preoperative factors in propensity score matching and postoperative outcomes. The number of pairs was 261
Full size table

Surgeons and anesthetic techniques

Prior to considering further analysis, we searched for differences in mortality related to surgeon or choice of anesthetic techniques employed. None were found. All 883 patients received a general anesthetic with inhaled agents being used in 882 (99.9%); one patient had malignant hyperthermia. Thoracic epidurals were placed in 715 (81%) patients and used during the anesthetic in 243 (28%) cases. Mean estimated blood loss was 1,267 ± 1,083 mL. Volume of fluid administered was 5,583 ± 3,282 mL, and fluid balance was positive by 3,918 ± 2,303 mL. Blood products were received and/or a cell saver was used in 521 patients (59%).

Postoperative outcomes

Postoperative outcomes for unmatched and propensity matched pairs are summarized in Table 4. Twenty-one of 359 patients with RAS blockade (5.8%) died within 30 days of surgery vs 10/524 (1.9%) of unblocked subjects (P < 0.002). Data for secondary outcomes and common postoperative complications for ACE/ARA exposed subjects and unexposed subjects are reported in Table 4.

Table 4 Postoperative outcomes and RAS blockade
Full size table

Propensity matched pairs outcomes are also reported in Table 4 and distribution of pairs in Table 3. The matching process afforded 261 pairs of subjects, and analysis revealed 16 deaths (6.1%) in RAS blocked patients and four deaths (1.5%) in unblocked patients. One pair was concordant for death. The McNemar Score was calculated as 8.0 (P = 0.008). The estimated odds ratio for increased mortality in the matched sample was 5; CI95 1.4-27. However, the matching did not control for hypertension; 94% of pairs were concordant (222 hypertensive and 24 not hypertensive) and 15 pairs discordant (two RAS blocked and 13 unexposed) (P = 0.005). Causes of death and associated conditions are reported in Table 5.

Table 5 Causes of death and associated factors
Full size table

Mortality associated with medical conditions and medication exposures

Unadjusted odds ratios for 30-day mortality were calculated for medical conditions and medication exposures using the entire data set of 883 patients. Only two medical conditions were associated with 30-day mortality: diabetes (odds ratio 2.8; CI95 1.2-6.1) and CHF (odds ratio 4.1; CI95 1.3-12.3). Only two medication exposures were associated with increased mortality: ACE inhibitor/ARA exposure 21/359 (5.8%) vs 10/524 (1.9%) (odds ratio 3.2; CI95 1.5-6.9), and diuretic exposure 15/164 (9.1%) exposed vs 16/688 (2.3%) unexposed (odds ratio 3.9; CI95 1.90-8.11).

Discussion

Perioperative use of ACE inhibitors or ARAs remains controversial. Randomized controlled trials demonstrating the clinical effectiveness of ACE inhibitors and ARAs in the medical setting may not be applicable perioperatively, because none of the RCTs examined the effects of anesthesia and perioperative care on RAS blockade. It is not known if a drug interaction exists between RAS blocking medications and inhaled anesthetic agents. A possible explanation for intraoperative hypotension in RAS blocked patients is impaired Ca2+ binding in vascular smooth muscle cells. Vascular myocyte contraction can be initiated by angiotensin 2 binding to the angiotensin 1 (AT1) receptor.16 Protein kinase C (PKC) is involved in multiple intracellular pathways, including sensitizing the AT1 and angiotensin 2 (AT2) receptors.17 Inhaled anesthetics inhibit PKC causing decreased activity of the AT1 receptor and increased activity of the AT2 receptor causing decreased vascular tone.16-19 Bradykinin has also been implicated as a possible intermediary.20

In the present study, RAS blockade was associated with an increased mortality within 30 days of surgery in vascular surgery patients (uncorrected odds ratio 3.2; CI95 1.5-6.9). Differences in underlying medical conditions and medication exposures between RAS blocked patients and unexposed patients were identified. Separation of the effects of a medication from the medical condition for which it is prescribed is problematic. The possibility exists that ACE/ARAs are prescribed to patients who are medically less fit, which would account for the observed differences. Differences in underlying medical conditions were addressed by propensity score matching predicting the preoperative likelihood that a patient would be exposed to an ACE inhibitor or ARA prior to surgery.

Matching based only on age and sex, as was done in other studies,21 predisposes towards poor matching of subjects. To reduce selection bias and confounding, a propensity matching system was developed to match exposed and unexposed subjects. Logistic regression analysis of the 883 cases identified five predictors of ACE inhibitor/ARA exposure: hypertension, stroke, CHF, diabetes, and heart disease. Hypertension was the strongest predictor of preoperative ACE/ARA exposure. Additional analysis of the data set identified preoperative diuretic exposure as a factor associated with mortality. Cardiovascular medication exposures were different between RAS blocked and unblocked patient populations. A propensity score model was developed that incorporated the above information and known preoperative predictors thought to be associated with mortality in this patient population.22-30 The model incorporated the following terms: age, sex, creatinine, hypertension, heart disease, CHF, diabetes, stroke, and cardiovascular medication exposures (beta-blocker, CaCB, diuretic, and statins). The patients were then matched on the calculated propensity score predicting the likelihood that patients would be preoperatively prescribed an ACE inhibitor or ARA to block the RAS.

This matching method afforded 261 pairs of subjects. The 30-day mortality of the RAS blocked patients was 16/261 deaths (6.1%) compared with 4/261 (1.5%) for the unblocked patients (odds ratio 5.0; CI95 1.4-27; P = 0.008). The patients were well matched for all medical conditions except hypertension, despite having 94% of the pairs matched concordantly. There were 15 discordant pairs for a preoperative diagnosis of hypertension, two pairs RAS blocked vs 13 pairs unblocked (McNemar score 8; P = 0.008).

Analysis of known medical risk factors22-30 was performed on the entire cohort to assess perioperative mortality. In this data set only, CHF, diabetes, and poor renal function (Cr Cl < 45 mL·min−1) were predictors of 30-day mortality. Hypertension was not a predictor of mortality in this data set nor is hypertension a known predictor of mortality in this patient population. Given the background information from the medical literature and the high percentage of concordant pairs matched with hypertension, it seems unlikely that hypertension played a major role in the observed increased mortality, but nonetheless, hypertension remains a possible confounding variable.

Nine diabetics died in this cohort of patients. Preoperative diagnosis of diabetes was used as a matching variable, and the matching methodology appears to have controlled for diabetes. However, none of the RAS unblocked diabetics in the data set died. The possibility exists that diabetes may be playing some role in the observed increased rate of death due to RAS blockade despite good matching for diabetes.

Patients exposed to ACE inhibitors/ARA agents are likely to be exposed to other cardiovascular medications: beta-blockers, diuretics, and statins (all P < 0.01). Although controversial, possible protective effects of some cardiovascular medications were considered. In the present study, exposure to beta-blockers,31-34 statins,35 or CaCB36 showed no effect on mortality. Only ACE inhibitors/ARA exposure and diuretic exposure were predictive of increased 30-day mortality.

Keterpal et al. 9 showed that exposure to ACE inhibitors and diuretics increased intraoperative hypotension but did not increase the rate of postoperative complications. In the present study, patients exposed to ACE/ARA agents were more likely to be exposed to diuretics: 103/359 (29%) in the ACE/ARA exposed patients vs 76/524 (15%) in the non ACE/ARA exposed (P < 0.001). Perioperative hypotension may be a possible contributor to increased mortality, but it was not tracked well in our data set.

In this study, preoperative diuretic exposure (all types grouped together) was a significant predictor of death within 30 days of surgery (odds ratio 3.9; CI95 1.9-8.1). The propensity matching system achieved good matching for all preoperative cardiovascular medication exposures, including diuretics. Based on this data set, it appears that diuretics are not playing a role in perioperative RAS blockade associated mortality. However, the data set may not be large enough to rule out an interaction with diuretics and medications that block the RAS.

In regards to postoperative outcomes, approximately 80% of patients suffered a postoperative complication, albeit the majority of complications were minor. In addition to increased mortality, RAS blocked patients were more likely to need postoperative hemodialysis (McNemar score 7.14; P < 0.013) despite not having a higher rate of postoperative renal failure. The matched group of patients did not have increased rates of other common major postoperative complications: MI, CHF, renal failure, and stroke. A higher percentage of RAS blocked patients died with multi-system organ failure (Table 5).

There are several limitations related to this study. The use of retrospective data from chart review to study clinical questions has marked limitations, for example, only association but not causality can be demonstrated. Another weakness of chart review studies is the inability to adequately control for exposures other than the exposure of interest. The best way to test the perioperative safety of RAS blockade would be to conduct a prospective randomized trial in patients undergoing major surgeries with sufficient washout of active metabolites (circa ten days).

In conclusion, we assessed patients with preoperative exposure to ACE inhibitors and ARA agents in terms of their outcomes with perioperative blockade of the RAS. Our data suggest that blockade of the RAS is associated with increased 30-day in-hospital mortality. Hypertension could not be excluded as a possible confounding variable. An association between chronic blockade of the RAS and increased 30-day mortality was found. It is not known if stopping these medications prior to surgery would alter this finding. A careful evaluation of the risks and benefits of ACE inhibitors or ARA agents should be factored into the decision for their continuance up to the time of major surgery. More studies are needed to evaluate the mechanism of the drug interaction of agents that block the renin-agiotensin system and anesthetic agents. The potential roles played by diabetes and medication exposures (especially diuretics) should also be considered in the design of future related studies.