- Email: [email protected]
emergency surgical patients†
British Journal of Anaesthesia 86 (4): 506±12 (2001)
Intercurrent drug therapy and perioperative cardiovascular mortality in elective and urgent/emergency surgical patients² J. W. Sear1*, S. J. Howell1, Y. M. Sear1, D. Yeates2, M. Goldacre2 and P. Foex1 1
Nuf®eld Department of Anaesthetics, University of Oxford, Oxford, UK. 2Unit of Health Care Epidemiology, University of Oxford, Oxford, UK
*Corresponding author: Nuf®eld Department of Anaesthetics, John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK The Oxford Record Linkage Study (ORLS; an epidemiological database) was used to examine relationships between intercurrent cardiovascular drug therapy and cardiac death within 30 days of elective or emergency/urgent surgery under general anaesthesia. Cases identi®ed from the ORLS were paired with matched control patients. Clinical details were obtained from the patients' medical notes. In elective surgical patients, there was no effect of b-adrenoceptor or calcium entry channel blockade, diuretics or digoxin on cardiac death after adjusting for confounding variables. Use of nitrates was associated with an odds ratio of 4.79 [95% con®dence interval (CI) 1.01±22.72] for cardiac death after adjustment for confounding by a history of angina and residual age difference. In emergency/urgent patients, there were signi®cant univariate associations with cardiac death for intercurrent use of angiotensin converting enzyme (ACE) inhibitors (odds ratio 1.18) and diuretics (odds ratio 4.95; 95% CI 1.82±13.46). However, neither maintained signi®cance after adjustment for the confounding effect of cardiac failure. We conclude that, with the possible exception of the use of nitrates in elective surgical patients, chronic intercurrent drug treatment alone does not signi®cantly affect the odds of cardiac death within 30 days of surgery. Br J Anaesth 2001; 86: 506±12 Keywords: anaesthesia, general; complications, death; complications, cardiovascular disease; anaesthesia, audit; records, anaesthesia; complications, intercurrent drug therapy Accepted for publication: November 29, 2000
In high-risk patients undergoing major non-cardiac surgery, there is evidence that treatment with b-adrenoceptor blocking antagonists started before surgery and continued into the post-operative period can reduce cardiac mortality up to 2 yr after operation.1 2 There are no published data on the effects of chronic b-adrenoceptor blockade on cardiac mortality and no data for any associations between other cardiovascularly active drugs (especially diuretics, digoxin, nitrates, calcium channel entry blocking drugs and angiotensin converting enzyme (ACE) inhibitors) and outcome from elective non-cardiac surgical procedures under general anaesthesia. Furthermore, there are no studies examining the interaction between drug therapy and cardiac death in urgent/emergency surgical patients. We have reported previously on the use of a large epidemiological database [the Oxford Record Linkage Study (ORLS)] to examine the risk factors for cardiovascular death in the 30 days after elective and urgent/ emergency surgery under general anaesthesia.3±5 In these studies, we identi®ed three risk factors associated with
cardiac death in elective non-cardiac surgical patients (a previous myocardial infarction, a history of arterial hypertension, and renal failure), while in urgent or emergency surgical patients only the history or presence of congestive heart failure was a signi®cant risk factor. Many of our patients were receiving intercurrent drug therapies for the management of underlying medical problems. We have re-analysed our data to search for any associations between chronic b-adrenoceptor blockade or other intercurrent therapies and peri-operative cardiovascular death within 30 days of surgery in both elective and urgent/emergency surgical patients. ² The Oxford Record Linkage Study is funded by the South East Regional Of®ce of the NHS Executive. This study was supported in part by a grant from the Oxford District Clinical Quality and Audit Group (YMS), and for part of the study period by the Anglia and Oxford Regional Health Authority (SJH). Some of these data were presented at the November 1998 meeting of the Anaesthetic Research Society, and published in abstract form in Br J Anaesth 1999; 82: 458.
Ó The Board of Management and Trustees of the British Journal of Anaesthesia 2001
Intercurrent drug therapy in surgery under general anaesthesia
Methods The ORLS covers the former Oxford Regional Health Authority area and contains data on all hospital admissions and deaths in the area. We used the database to identify patients who died within 30 days of anaesthesia and surgery after elective or urgent/emergency admission to hospital, together with matched controls. The population chosen for this study comprised patients who were admitted to one of three hospital groups within the Oxford region and who died of a cardiac cause after either elective surgery (n=115) or urgent/emergency surgery (n=73) under general anaesthesia between 1979 and 1992. Because of the non-availability of some hospital records, the timescales of the two study populations overlapped but were not identical. We included only patients aged over 18 yr who were undergoing non-cardiac or non-neurological operations (predominantly abdominal, vascular and orthopaedic surgery). For the area covered by the ORLS (relative to England as a whole), the standardized mortality ratios (SMRs) for diseases of the circulatory system were 85 for men and 89 for women in 1990; the corresponding SMRs for ischaemic heart disease were 83 and 81.
Cases For each study, patients who had died within 30 days of anaesthesia and surgery and whose death certi®cate included ICD9 codes (9th International Classi®cation of Diseases) in the range 401±438 inclusive at any position on the death certi®cate were identi®ed. The diseases covered by these codes are hypertensive heart disease (401±405), ischaemic heart disease (410±414), diseases of the pulmonary circulation (415±417), other forms of heart disease (420±429) and cerebrovascular disease (430±439). We excluded patients dying from pulmonary embolism (415.1). There are also speci®c codes within ICD9 relating to complications associated with anaesthesia (e.g. 995.4 for shock as a result of anaesthesia, and 997.1 for cardiac arrest or insuf®ciency resulting from an operative procedure). As reported previously,3±5 these codes were not used as a certi®ed cause of death in any of the patients in the study population.
Controls For each patient, at least one control was identi®ed through the ORLS. Where possible, two or more controls were sought to allow for the possible non-availability of the medical case records. For the elective surgical patients, controls were matched according to the following criteria: (i) underwent the same operation as the patient; (ii) under the care of the same consultant as the patient; (iii) age within 5 yr of that of the patient; (iv) same sex as the case patient. For the urgent/emergency surgery patients, the matching criteria for the controls were slightly differentÐage was within 10 yr of the case and there was no matching for sex.
These criteria were relaxed to generate a larger number of control subjects, as it was felt that our earlier study of elective surgery suffered from over-matching.4 When two control patients were identi®ed, one was chosen at random and the notes were obtained. If these were not available, or if this control was excluded for other reasons, the notes of the second control patient were examined.
Examination of the medical records Before the study, we obtained consent to examine case and control patient records from the consultants caring for the patient during the surgical admission. Where a consultant had died or retired, the consent of his or her successor was sought. All notes were examined in detail by a research nurse (YMS). Whenever possible, the cause of death was con®rmed from the notes to ascertain that the cause on the death certi®cate was supported by the medical records. If the patient had not died from a cardiovascular cause, he or she was excluded from the study. Physical characteristics of the patients and information on cardiovascular risk factors were collected using criteria described elsewhere.4 5 In the case of drug therapies, these were noted where patients had received them for more than 1 month before hospital admission.
Data handling A computer database was written using Microsoft Access (version 2.0) to store and examine the data. Microsoft Excel (version 6.0) was used to examine the tables produced by Access. Statistical analysis was performed using Stata (version 5.0) (STATA Corporation; College Station, TX, USA). Data were exchanged between the various programs using DBMScopy (version 5.10) (DBMScopy Software Inc.; Houston, TX, USA). All programs were run on a Viglen 233 MHz Pentium II desktop computer.
Statistical analysis For matched data, univariate analysis was carried out and crude odds ratios were calculated as the ratio of the discordant pairs. Because subjects were matched for age over a wide interval (65 yr in the elective surgical patients, and 610 yr in the urgent/emergency surgical patients), there was potential for residual confounding by age within pairs. Adjustment was therefore made for within-pair age difference, using conditional logistic regression. This provides a useful statistical correction for within-pair age difference but does not give a direct description of the effect of age on peri-operative risk. In the urgent/emergency surgery patients, adjustment was similarly made for within-pair sex differences. Multivariate analyses were performed to examine the associations between cardiac death and the various
507
Sear et al. Table 1 Results for the association of individual drug therapies and cardiac death in elective surgical patients. Data are crude and age-adjusted odds ratios together with the 95% con®dence intervals (CI). P values refer to the age-adjusted odds ratio, and were calculated using the Wald test. *Lower CI estimate; no point estimate calculable. n.s.=not signi®cant
b-Adrenoceptor blockade Calcium channel entry blockade ACE inhibitors* Diuretics Digoxin Nitrates
Odds ratio (95% CI)
Age-adjusted odds ratio (95% CI)
P
1.64 (0.81±3.45) 4.67 (1.30±25.32) (0.19) 2.36 (1.23±4.77) 3.50 (1.10±14.60) 5.50 (1.87±21.96)
2.04 (1.00±1.17) 4.35 (1.20±15.79) (0.19) 2.51 (1.28±4.92) 4.13 (1.24±13.79) 6.04 (1.97±18.29)
0.05 0.026 n.s. 0.008 0.021 0.002
Table 2 Conditional logistic regression model for the effects of badrenoceptor blockade and calcium entry channel blockade on cardiac death within 30 days of elective surgery, adjusted for hypertension, angina and residual confounding by age. Data are odds ratios together with the 95% con®dence intervals (CI). P values were calculated using the Wald test
b-Adrenoceptor blockade Calcium entry channel blockade Hypertension Angina Age difference
Odds ratio (95% CI)
z score (Wald test)
P
1.00 1.61 2.04 2.57 0.82
0.007 0.62 1.97 1.8 3.2
1.0 0.53 0.049 0.074 0.001
(0.41±2.44) (0.36±7.12) (1.0±4.13) (0.91±7.27) (0.73±0.93)
cardiovascular medications after adjusting for confounding. For each drug examined, statistical adjustment was made for the indications for prescription of that therapy. For example, in the case of b-adrenoceptor blockers and calcium channel entry blocking drugs, adjustment was made for the effects of arterial hypertension and angina, and in the case of diuretics adjustment was made for the effects of congestive cardiac failure and hypertension. Alternative regression models were compared using the log likelihood ratio test. The signi®cance of individual terms in the ®nal models was examined with the Wald test.6 This test was performed by dividing the coef®cient ratio obtained from the regression model by its standard error to obtain a standard normal deviate. The P value was obtained using a table of probabilities. By this method, we examined the effects of chronic treatment with b-adrenoceptor blockers, calcium channel entry blockers, diuretics, nitrates and digoxin. The small numbers of patients receiving ACE inhibitors prevented the drug being examined in a regression model. The results generated by the analyses are shown as odds ratios and their 95% con®dence intervals (CI).
Results Elective surgical patients Among the elective surgical patients, we identi®ed 177 patients who died of a cardiovascular cause with 30 days of surgery under general anaesthesia; 115 of these were studied
further after matching with controls. Of the other 62 patients, 37 were excluded because either the notes of the patient, control or both were missing, and the other 25 because no matched control could be identi®ed. The mean age of the cases was 73 yr (range 45±93) and that of the controls 71 yr (43±90). In 74 (64%) of the pairs, the case and the control were male. Twenty-four cases and 15 controls were taking b-adrenoceptor-blocking drugs, 16 cases and ®ve controls were taking calcium channel entry blocking drugs, two cases and no controls were taking ACE inhibitors; 48 cases and 29 controls were taking diuretics, 24 cases and six controls were taking nitrates, and 15 cases and ®ve controls were taking digoxin. A signi®cantly greater number of the cases were receiving calcium channel entry blockers, diuretics, digoxin and nitrates. The causes of death in the cases were myocardial infarction/ischaemic heart disease (78 deaths), heart failure (23) and cerebrovascular accident (14). The crude odds ratios for the association of individual drug therapies and adverse outcome and the adjusted ratios for confounding by residual age difference between the case-control pairs are shown in Table 1. Univariate ageadjusted ratios showed signi®cant associations with cardiac death for calcium channel entry blockade, diuretics, digoxin and nitrates (P<0.05), and a signi®cance of P=0.05 for b-adrenoceptor blocking drugs. The effects of b-adrenoceptor blockade and calcium channel blockade were examined together in a conditional logistic regression model, which also contained terms for arterial hypertension and angina, and a term to allow for the effects of residual confounding by age. Full details of the model are given in Table 2. In this model, the odds ratio for the association between the use of b-adrenoceptor blockade and postoperative cardiac death was 1.00 (95% CI 0.41±2.44); z=0.007, P=1.0. The odds ratio for the association of calcium channel entry blockers and cardiac death was 1.61 (0.36±7.12); z=0.62, P=0.53. The addition of a term for interaction between b-adrenoceptor blocking drugs and calcium channel antagonists did not improve the model signi®cantly. That is to say, there was no evidence in elective surgical patients of interaction between b-adrenoceptor antagonists and calcium entry blocking drugs.
508
Intercurrent drug therapy in surgery under general anaesthesia Table 3 Results for the association of individual drug therapies and cardiac death in urgent/emergency patients. Data are crude and age- and sex-adjusted odds ratios together with the 95% con®dence intervals (CI). P values refer to the age- and sex-adjusted odds ratio, and were calculated using the Wald test. *Lower CI estimate; no point estimate calculable
b-Adrenoceptor blockade Calcium channel entry blockade ACE inhibitors* Diuretics Digoxin Nitrates
Odds ratio (95% CI)
Age- and sex-adjusted odds ratio (95% CI)
P
5.00 (0.56±236.45) 3.50 (0.67±34.53) (1.18) 4.60 (1.71±15.49) 2.50 (0.92±7.86) 1.75 (0.45±8.25)
5.11 (0.59±44.20) 3.57 (0.73±17.48) (1.18) 4.95 (1.82±13.46) 2.61 (0.99±6.90) 1.93 (0.54±6.82)
0.138 0.117 0.032 0.002 0.053 0.310
For the other drug therapies, there were no signi®cant associations between diuretics or digoxin and cardiac death after the confounding effects of arterial hypertension, heart failure and other drug therapies (i.e. b-adrenoceptor blocking drugs and calcium channel entry blocking drugs) had been taken into account. There was, however, a signi®cant association between administration of nitrates and cardiac death after adjustment for the confounding effects of preexisting angina and residual age difference [odds ratio 4.79 (95% CI 1.00±22.72); z=1.943, P=0.049]. After adjustment for the effect of any associated administration of badrenoceptor blockade, the odds ratio increased to 5.72 (1.09±25.54); z=2.065, P=0.039.
Urgent and emergency patients In the urgent/emergency patients, we identi®ed 202 cases; controls were found for 73 of them. Of the 129 patients excluded, 34 underwent surgery under local or regional anaesthesia, 24 were admitted as emergency patients but initially received medical treatment rather than surgical care (surgery taking place some days after admission), 25 had untraceable notes or no matched control and 45 did not ful®l the study criteria (18 having pulmonary emboli, three noncardiovascular causes of death and 24 being incorrectly coded); the remaining patient died on the way to the operating theatre. Death was due to myocardial infarction/ ischaemic heart disease in 34 patients, cardiac failure in 16, cerebrovascular accident in 14, and other cardiac causes in the remaining nine. The mean age of the cases was 81 yr (range 40±98) and that of the controls was 81 yr (43±100). Five cases were receiving b-adrenoceptor blocking drugs and seven were receiving calcium channel entry blockers; the corresponding controls included one patient on b-adrenoceptor blockers and two on calcium channel entry blockers. For the other drug treatments, six cases and no controls were receiving ACE inhibitors, seven cases and four controls were receiving nitrates, 39 cases and 21 control were receiving diuretics, and 16 cases and seven controls were receiving digoxin. In the urgent/emergency patients, more cases were receiving diuretics and ACE inhibitors.
There were no signi®cant univariate associations between b-adrenoceptor blocking drugs, calcium channel entry blocking drugs or nitrates and cardiac death in the urgent/ emergency surgical patients; however, after adjusting for sex and residual age difference between the case±control pairs, there were signi®cant associations between diuretics and ACE inhibitors and cardiac death (Table 3). As with the elective surgical patients, there was no interaction between b-adrenoceptor blockade and calcium channel entry blocking drugs After adjusting for the effects of other confounders (such as angina, arterial hypertension and heart failure) by logistic regression analysis, there were no associations between drug therapies and cardiac death. In the case of diuretic therapy, the addition of a history of heart failure to the regression model reduced the odds ratio for diuretics from 4.29 to 1.94, while a history of heart failure reduced the odds ratio for digoxin from 2.61 to 1.34 (both ®nal odds being non-signi®cant).
Discussion In this case±control study, we found no signi®cant association in elective or emergency surgical patients between chronic b-adrenoceptor blockade and cardiac death within 30 days of anaesthesia. The analyses presented here were prompted by the con¯icting ®ndings of several other groups of researchers.1 2 7 8 In a randomized controlled trial of atenolol versus placebo, Mangano and colleagues demonstrated a reduction in mortality 6, 12 and 24 months after hospital discharge in patients randomized to receive the b-adrenoceptor blocking drug before operation and for 7 days after surgery.1 The main effect of atenolol treatment was seen in the ®rst 6 months after surgery, but no difference in in-hospital mortality was observed. In a second study, Poldermans and colleagues reported on 112 high cardiac risk patients with evidence of dobutamine stress-inducible echocardiographic (ECHO) abnormalities scheduled for major vascular surgery.2 After randomization, the treatment group received oral bisoprolol for at least 1 week before surgery, and continued until 30 days after operation. The incidence of cardiac deaths in the bisoprolol
509
Sear et al.
group during the 30-day peri-operative period was only 3.4% compared with 17% in the control group. There was also a difference in the incidence of non-fatal myocardial infarction (0% in the bisoprolol group vs 17% in the control group). In a subgroup consisting of 53 patients already maintained on b-adrenoceptor blocking drugs, there were four post-operative cardiac deaths (7.5%) and no non-fatal infarctions. These results are in contrast to the data of Mangano and colleagues, who found no effect of acute atenolol compared with placebo on in-hospital deaths.1 In two further studies in patients undergoing vascular surgery under general or regional anaesthesia, Yeager and colleagues found b-adrenoceptor blockade was associated with a 50% reduction in peri-operative myocardial infarction, but they do not report on peri-operative mortality,7 while Oliver and colleagues also found no effect of badrenoceptor blockade on the incidence of either myocardial infarction or cardiac death in patients receiving mivazerol or placebo.8 There are differences in study design between our study and these other studies. The latter used randomization of patients from high-risk populations for cardiovascular disease to ensure a comparable distribution of risk factors between the treatment and placebo groups. On the other hand, our observational study depended on examining the possible effects of chronic b-adrenoceptor therapy, having made statistical adjustments for the in¯uence of other cardiovascular risk factors. Speci®cally, adjustment was made for the indication for the prescription of the various medications. In the analysis of the data for elective surgery, we included terms for the two main indications for the prescription of b-adrenoceptor blocking drugs: arterial hypertension and angina. These differences may help explain the contrast between our ®ndings and those of Mangano and colleagues1 and Poldermans and colleagues.2 More recently, studies have shown that use of badrenoceptor blockade in patients with mild to moderate heart failure may have a signi®cant effect on cardiac outcome in these high-risk patients.9 Two studies have found b-adrenoceptor blockade to be associated with a reduction in total cardiac mortality and all-cause hospitalizations in patients with heart failure.10 11 The interaction of anaesthesia, b-blockade and heart failure has not yet been evaluated fully. In our present study populations, there were few patients who had a history of heart failure and were receiving b-adrenoceptor blocking drugs (four among the elective surgical patients and two in the urgent/emergency patients). However, there were no differences in death rates between patients who had a history of heart failure and who were receiving b-adrenoceptor blocking drugs compared with patients not receiving this treatment. Another possible explanation for the difference between acute and chronic b-blockade is that patients on long-term cardioselective b1 antagonists do not maintain sustained blockade as a result of b1 receptor adaptation. This could result from either receptor up-regulation or a compensatory
increase in cardiac b2 receptors, or a combination of the two.12 An alternative explanation might be that of inverse agonism by chronic b-blockade at G-protein-coupled receptors, as has been demonstrated in vitro in rat ®broblasts by Milligan and colleagues.13 There have been a number of studies examining the effects of medication with calcium channel antagonists on cardiac death in patients with stable angina or after infarction.14±16 There are few comparable data on cardiac outcomes in non-cardiac surgical patients receiving intercurrent calcium channel entry blocking drugs. We found an increased incidence of post-operative silent myocardial ischaemia in patients receiving calcium channel entry blocking drugs.17. Mangano and colleagues demonstrated an association between pre-operative calcium entry blocking drugs and post-operative ischaemia in non-cardiac surgical patients.18 However, multivariate regression analysis failed to support this association, or an association with either in-hospital or long-term mortality in US Veterans Administration male patients undergoing non-cardiac surgery.19 There have been reports that short-acting calcium channel entry blocking drugs (such as nifedipine) may be associated with an increased risk of cardiovascular events, including cardiac death. Thus, Psaty and colleagues found an association between short-acting calcium channel entry blocking drugs and the increased risk of myocardial infarction in hypertensive patients.20 However, this was not borne out in two further studies in patients receiving calcium channel entry blocking drugs for chronic coronary artery disease21 or hypertension.22 It is unlikely that the regression models used in our analyses make a complete adjustment for confounding risk factors for the effects of either chronic b-adrenoceptor or channel entry channel blockade. However, the fact that the odds ratios for both treatments moved towards unity after adjustment suggests no independent effect of these drugs. The slight association between nitrate administration and adverse cardiac outcomes in the elective surgical patients was surprising. However, Ishikawa and colleagues have shown long-term nitrate therapy given to post-infarction patients to be associated with an increased incidence of cardiac events, including cardiac death.23 There are no data on post-surgical cardiac complications in patients on longterm nitrates, although ®ve studies have looked at their effects on perioperative ischaemia. Two studies show a signi®cant reduction in intraoperative ischaemia24 25 while three others show no effect.26±28 The differences in outcome may re¯ect differences in underlying anaesthetic technique (volatile vs opioid-based) or differences in study populations. Our data may indicate the need for high-risk cardiac patients receiving only nitrates to be given additional antiischaemic drugs in the perioperative period. Other researchers have noted increased perioperative mortality associated with emergency surgery.29±31 In our cases who underwent urgent or emergency surgery, there
510
Intercurrent drug therapy in surgery under general anaesthesia
were no signi®cant associations between b-adrenoceptor or calcium channel entry blockade and 30-day cardiac death. Although diuretics and ACE inhibitor therapy had signi®cant associations with cardiac death on univariate analysis, these effects were rendered non-signi®cant by adjustment of the data for confounding factors (especially heart failure). The implication is that the risk associated with these drugs was due to the conditions for which they were prescribed. Heart failure is universally accepted as a strong predictor of post-operative myocardial infarction, reinfarction and worsening left ventricular function,32 and cardiac death.33 In conclusion, our case±control study failed to show any signi®cant effect of chronic b-adrenoceptor blockade on 30day survival after either elective or urgent/emergency noncardiac surgery. However, the CIs for the death rates after urgent/emergency surgery were wide, and we suggest that further studies using large databases, such as the ORLS, are warranted. Although chronic calcium channel entry blockade was associated with adverse cardiac outcome in elective surgical patients by univariate analysis, this was not sustained after adjustment for the effects of hypertension and angina. In the emergency or urgent surgery patients, there was a signi®cant association between digoxin therapy and cardiac death, supporting the ®ndings of Hollenberg and Mangano, who showed that pre-operative digoxin therapy was a signi®cant predictor of post-operative silent myocardial ischaemia and hence cardiac events.34 However, after adjustment for confounding due to a history of cardiac failure (a well-established risk factor for cardiac death following anaesthesia and surgery),5 33 digoxin therapy did not remain a signi®cant risk factor. As with our previous publications, caution should be exercised because of the assumed low power of the study, as indicated by the small numbers of patients receiving the therapies and the resulting large CIs of the odds ratios.
References 1 Mangano DT, Layug EL, Wallace A, Tateo I, for the Multicenter Study of Perioperative Ischemia Research Group. Effect of atenolol on mortality and cardiovascular morbidity after noncardiac surgery. New Engl J Med 1996; 335: 1713±20 2 Poldermans D, Boersma E, Bax JJ, et al. The effect of bisoprolol on perioperative mortality and myocardial infarction in high-risk patients undergoing vascular surgery. New Engl J Med 1999; 341: 1789±94 3 Howell SJ, Sear YM, Yeates D, Goldacre M, Sear JW, Foex P. Hypertension, admission blood pressure and perioperative cardiovascular risk. Anaesthesia 1996; 51: 1000±4 4 Howell SJ, Sear YM, Yeates D, Goldacre M, Sear JW, Foex P. Risk factors for cardiovascular death after elective surgery under general anaesthesia. Br J Anaesth 1998; 80: 14±19 5 Howell SJ, Sear YM, Sear JW, Yeates D, Goldacre M, Foex P. Cardiovascular death within 30 days following anaesthesia and urgent or emergency surgery. Determination of risk factors using a nested case±control study. Br J Anaesth 1999; 82: 679±84 6 Armitage P, Berry G. Statistical Methods in Medical Research, edn 3. Oxford: Blackwell Science, 1994; 429
511
7 Yeager RA, Moneta GL, Edwards JM, Taylor LM, McConnell DB, Porter JM. Reducing perioperative myocardial infarction following vascular surgery. The potential role of b-blockade. Arch Surg 1995; 130: 869±73 8 Oliver MF, Goldman L, Julian DG, Holme I. Effect of mivazerol on perioperative cardiac complications during non-cardiac surgery in patients with coronary heart disease. Anesthesiology 1999; 91: 951±61 9 Abraham WT. Beta-blockers: the new standard of therapy for mild heart failure. Arch Intern Med 2000; 160: 1237±47 10 Hjalmarson A, Goldstein S, Fagerberg B, et al., for the MERIT-HF Study Group. Effects of controlled-release metoprolol on total mortality, hospitalizations, and well-being in patients with heart failure: the Metoprolol CR/XL Randomized Intervention Trial in Congestive Heart Failure (MERIT-HF). J Am Med Assoc 2000; 283: 1295±302 11 Bonet S, Agusti A, Arnau JM, et al. Beta-adrenergic blocking agents in heart failure: bene®ts of vasodilating and nonvasodilating agents according to patients' characteristics: a meta-analysis of clinical trials. Arch Intern Med 2000; 160: 621±27 12 Yndgaard S, Lippert FK, Berthelsen PG. Are patients chronically treated with b1-adrenoceptor antagonists in fact b-blocked? J Cardiothoracic Vasc Anesth 1997; 11: 32±6 13 Milligan G, MacEwan DJ, Mercouris M, Mullaney I. Inverse agonism at adrenergic and opioid receptors: studies with wild type and constitutively active mutant receptors. Receptors Channels 1997; 5: 209±213 14 Heidenreich PA, McDonald KM, Hastie T, et al. Meta-analysis of trials comparing beta-blockers, calcium antagonists and nitrates for stable angina. J Am Med Assoc 1999; 281: 1927±36 15 Singh BN. When is drug therapy warranted to prevent sudden cardiac death? Drugs 1991; 41 Suppl 2: 24±46 16 Leitch JW, McElduff P, Dobson A, Heller R. Outcome with calcium channel antagonists after myocardial infarction: a community-based study. J Am Coll Cardiol 1998; 31: 111±7 17 Sear JW, Foex P, Howell SJ. Effect of chronic intercurrent medication with b-adrenoceptor blockade or calcium channel entry blockade on postoperative silent myocardial ischaemia. Br J Anaesth 2000; 84: 311±5 18 Hollenberg M, Mangano DT, Browner WS, et al., for the Study of Perioperative Ischemia Research Group. Predictors of postoperative myocardial ischemia in patients undergoing noncardiac surgery. J Am Med Assoc 1992; 268: 205±9 19 Browner WS, Li J, Mangano DT, for the Study of Perioperative Ischemia Research Group. In-hospital and long-term mortality in male veterans following non-cardiac surgery. J Am Med Assoc 1992; 268: 228±32 20 Psaty BM, Heckbert SR, Koepsell TD, et al. The risk of myocardial infarction associated with antihypertensive drug therapies. J Am Med Assoc 1995; 274: 620±5 21 Braun S, Boyko V, Behar S, et al. Calcium antagonists and mortality in patients with coronary artery disease: a cohort study of 11,575 patients. J Am Coll Cardiol 1996; 28: 7±11 22 Jick H, Derby LE, Gurewich V, Vasilakis C. The risk of myocardial infarction associated with antihypertensive treatment in persons with uncomplicated essential hypertension. Pharmacotherapy 1996; 16: 321±6 23 Ishikawa K, Kanamasa K, Ogawa I, et al. Long-term nitrate treatment increases cardiac events in patients with healed myocardial infarction. Japan Circ J 1996; 60: 779±88 24 Coriat P, Daloz M, Bousseau D, et al. Prevention of intraoperative myocardial ischemia during noncardiac surgery with intravenous nitroglycerin. Anesthesiology 1984; 63: 193±6 25 Fusciardi J, Godet G, Bernard J, et al. Roles of fentanyl and
Sear et al.
26
27 28 29
nitroglycerin in prevention of myocardial ischemia associated with laryngoscopy and tracheal intubation in patients undergoing operations of short duration. Anesth Analg 1986; 65: 617±24 Thomson J, Mutch W, Culligan J. Failure of intravenous nitroglycerin to prevent intraoperative myocardial ischemia during fentanyl±pancuronium anesthesia. Anesthesiology 1984; 61: 385±93 Gallagher J, Moore R, Jose A, et al. Prophylactic nitroglycerin infusions during coronary artery bypass surgery. Anesthesiology 1986; 64: 785±9 Dodds T, Stone G, Coromilas J. Prophylactic nitroglycerin infusion during noncardiac surgery does not reduce perioperative ischemia. Anesth Analg 1993; 76: 705±13 Fowkes FGR, Lunn JN, Farrow SC, Robertson IB, Samuel P. Epidemiology in anaesthesia. III: Mortality in patients with coexisting physical disease. Br J Anaesth 1982; 54: 819±25
512
30 Goldman L, Caldera DL, Nussbaum SR, et al. Multifactorial index of cardiac risk in noncardiac surgical procedures. New Engl J Med 1977; 297: 845±50 31 Pedersen T, Elliasen K, Henrikson E. A prospective study of mortality associated with anaesthesia and surgery: risk indicators of mortality in hospital. Acta Anaesthesiol Scand 1990; 34: 176±82 32 Larsen SF, Olesen KH, Jacobsen E, et al. Prediction of cardiac risk in non-cardiac surgery. Eur Heart J 1987; 8: 179±85 33 Dirksen A, Kjoller E. Cardiac predictors of death after noncardiac surgery evaluated by intention to treat. Br Med J 1988; 297: 1011±13 34 Hollenberg M, Mangano DT. Therapeutic approaches to postoperative ischemia. The Study of Perioperative Ischemia Research Group. Am J Cardiol 1994; 73: 30B±33B
Intercurrent drug therapy and perioperative cardiovascular mortality in elective and urgent/emergency surgical patients² J. W. Sear1*, S. J. Howell1, Y. M. Sear1, D. Yeates2, M. Goldacre2 and P. Foex1 1
Nuf®eld Department of Anaesthetics, University of Oxford, Oxford, UK. 2Unit of Health Care Epidemiology, University of Oxford, Oxford, UK
*Corresponding author: Nuf®eld Department of Anaesthetics, John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK The Oxford Record Linkage Study (ORLS; an epidemiological database) was used to examine relationships between intercurrent cardiovascular drug therapy and cardiac death within 30 days of elective or emergency/urgent surgery under general anaesthesia. Cases identi®ed from the ORLS were paired with matched control patients. Clinical details were obtained from the patients' medical notes. In elective surgical patients, there was no effect of b-adrenoceptor or calcium entry channel blockade, diuretics or digoxin on cardiac death after adjusting for confounding variables. Use of nitrates was associated with an odds ratio of 4.79 [95% con®dence interval (CI) 1.01±22.72] for cardiac death after adjustment for confounding by a history of angina and residual age difference. In emergency/urgent patients, there were signi®cant univariate associations with cardiac death for intercurrent use of angiotensin converting enzyme (ACE) inhibitors (odds ratio 1.18) and diuretics (odds ratio 4.95; 95% CI 1.82±13.46). However, neither maintained signi®cance after adjustment for the confounding effect of cardiac failure. We conclude that, with the possible exception of the use of nitrates in elective surgical patients, chronic intercurrent drug treatment alone does not signi®cantly affect the odds of cardiac death within 30 days of surgery. Br J Anaesth 2001; 86: 506±12 Keywords: anaesthesia, general; complications, death; complications, cardiovascular disease; anaesthesia, audit; records, anaesthesia; complications, intercurrent drug therapy Accepted for publication: November 29, 2000
In high-risk patients undergoing major non-cardiac surgery, there is evidence that treatment with b-adrenoceptor blocking antagonists started before surgery and continued into the post-operative period can reduce cardiac mortality up to 2 yr after operation.1 2 There are no published data on the effects of chronic b-adrenoceptor blockade on cardiac mortality and no data for any associations between other cardiovascularly active drugs (especially diuretics, digoxin, nitrates, calcium channel entry blocking drugs and angiotensin converting enzyme (ACE) inhibitors) and outcome from elective non-cardiac surgical procedures under general anaesthesia. Furthermore, there are no studies examining the interaction between drug therapy and cardiac death in urgent/emergency surgical patients. We have reported previously on the use of a large epidemiological database [the Oxford Record Linkage Study (ORLS)] to examine the risk factors for cardiovascular death in the 30 days after elective and urgent/ emergency surgery under general anaesthesia.3±5 In these studies, we identi®ed three risk factors associated with
cardiac death in elective non-cardiac surgical patients (a previous myocardial infarction, a history of arterial hypertension, and renal failure), while in urgent or emergency surgical patients only the history or presence of congestive heart failure was a signi®cant risk factor. Many of our patients were receiving intercurrent drug therapies for the management of underlying medical problems. We have re-analysed our data to search for any associations between chronic b-adrenoceptor blockade or other intercurrent therapies and peri-operative cardiovascular death within 30 days of surgery in both elective and urgent/emergency surgical patients. ² The Oxford Record Linkage Study is funded by the South East Regional Of®ce of the NHS Executive. This study was supported in part by a grant from the Oxford District Clinical Quality and Audit Group (YMS), and for part of the study period by the Anglia and Oxford Regional Health Authority (SJH). Some of these data were presented at the November 1998 meeting of the Anaesthetic Research Society, and published in abstract form in Br J Anaesth 1999; 82: 458.
Ó The Board of Management and Trustees of the British Journal of Anaesthesia 2001
Intercurrent drug therapy in surgery under general anaesthesia
Methods The ORLS covers the former Oxford Regional Health Authority area and contains data on all hospital admissions and deaths in the area. We used the database to identify patients who died within 30 days of anaesthesia and surgery after elective or urgent/emergency admission to hospital, together with matched controls. The population chosen for this study comprised patients who were admitted to one of three hospital groups within the Oxford region and who died of a cardiac cause after either elective surgery (n=115) or urgent/emergency surgery (n=73) under general anaesthesia between 1979 and 1992. Because of the non-availability of some hospital records, the timescales of the two study populations overlapped but were not identical. We included only patients aged over 18 yr who were undergoing non-cardiac or non-neurological operations (predominantly abdominal, vascular and orthopaedic surgery). For the area covered by the ORLS (relative to England as a whole), the standardized mortality ratios (SMRs) for diseases of the circulatory system were 85 for men and 89 for women in 1990; the corresponding SMRs for ischaemic heart disease were 83 and 81.
Cases For each study, patients who had died within 30 days of anaesthesia and surgery and whose death certi®cate included ICD9 codes (9th International Classi®cation of Diseases) in the range 401±438 inclusive at any position on the death certi®cate were identi®ed. The diseases covered by these codes are hypertensive heart disease (401±405), ischaemic heart disease (410±414), diseases of the pulmonary circulation (415±417), other forms of heart disease (420±429) and cerebrovascular disease (430±439). We excluded patients dying from pulmonary embolism (415.1). There are also speci®c codes within ICD9 relating to complications associated with anaesthesia (e.g. 995.4 for shock as a result of anaesthesia, and 997.1 for cardiac arrest or insuf®ciency resulting from an operative procedure). As reported previously,3±5 these codes were not used as a certi®ed cause of death in any of the patients in the study population.
Controls For each patient, at least one control was identi®ed through the ORLS. Where possible, two or more controls were sought to allow for the possible non-availability of the medical case records. For the elective surgical patients, controls were matched according to the following criteria: (i) underwent the same operation as the patient; (ii) under the care of the same consultant as the patient; (iii) age within 5 yr of that of the patient; (iv) same sex as the case patient. For the urgent/emergency surgery patients, the matching criteria for the controls were slightly differentÐage was within 10 yr of the case and there was no matching for sex.
These criteria were relaxed to generate a larger number of control subjects, as it was felt that our earlier study of elective surgery suffered from over-matching.4 When two control patients were identi®ed, one was chosen at random and the notes were obtained. If these were not available, or if this control was excluded for other reasons, the notes of the second control patient were examined.
Examination of the medical records Before the study, we obtained consent to examine case and control patient records from the consultants caring for the patient during the surgical admission. Where a consultant had died or retired, the consent of his or her successor was sought. All notes were examined in detail by a research nurse (YMS). Whenever possible, the cause of death was con®rmed from the notes to ascertain that the cause on the death certi®cate was supported by the medical records. If the patient had not died from a cardiovascular cause, he or she was excluded from the study. Physical characteristics of the patients and information on cardiovascular risk factors were collected using criteria described elsewhere.4 5 In the case of drug therapies, these were noted where patients had received them for more than 1 month before hospital admission.
Data handling A computer database was written using Microsoft Access (version 2.0) to store and examine the data. Microsoft Excel (version 6.0) was used to examine the tables produced by Access. Statistical analysis was performed using Stata (version 5.0) (STATA Corporation; College Station, TX, USA). Data were exchanged between the various programs using DBMScopy (version 5.10) (DBMScopy Software Inc.; Houston, TX, USA). All programs were run on a Viglen 233 MHz Pentium II desktop computer.
Statistical analysis For matched data, univariate analysis was carried out and crude odds ratios were calculated as the ratio of the discordant pairs. Because subjects were matched for age over a wide interval (65 yr in the elective surgical patients, and 610 yr in the urgent/emergency surgical patients), there was potential for residual confounding by age within pairs. Adjustment was therefore made for within-pair age difference, using conditional logistic regression. This provides a useful statistical correction for within-pair age difference but does not give a direct description of the effect of age on peri-operative risk. In the urgent/emergency surgery patients, adjustment was similarly made for within-pair sex differences. Multivariate analyses were performed to examine the associations between cardiac death and the various
507
Sear et al. Table 1 Results for the association of individual drug therapies and cardiac death in elective surgical patients. Data are crude and age-adjusted odds ratios together with the 95% con®dence intervals (CI). P values refer to the age-adjusted odds ratio, and were calculated using the Wald test. *Lower CI estimate; no point estimate calculable. n.s.=not signi®cant
b-Adrenoceptor blockade Calcium channel entry blockade ACE inhibitors* Diuretics Digoxin Nitrates
Odds ratio (95% CI)
Age-adjusted odds ratio (95% CI)
P
1.64 (0.81±3.45) 4.67 (1.30±25.32) (0.19) 2.36 (1.23±4.77) 3.50 (1.10±14.60) 5.50 (1.87±21.96)
2.04 (1.00±1.17) 4.35 (1.20±15.79) (0.19) 2.51 (1.28±4.92) 4.13 (1.24±13.79) 6.04 (1.97±18.29)
0.05 0.026 n.s. 0.008 0.021 0.002
Table 2 Conditional logistic regression model for the effects of badrenoceptor blockade and calcium entry channel blockade on cardiac death within 30 days of elective surgery, adjusted for hypertension, angina and residual confounding by age. Data are odds ratios together with the 95% con®dence intervals (CI). P values were calculated using the Wald test
b-Adrenoceptor blockade Calcium entry channel blockade Hypertension Angina Age difference
Odds ratio (95% CI)
z score (Wald test)
P
1.00 1.61 2.04 2.57 0.82
0.007 0.62 1.97 1.8 3.2
1.0 0.53 0.049 0.074 0.001
(0.41±2.44) (0.36±7.12) (1.0±4.13) (0.91±7.27) (0.73±0.93)
cardiovascular medications after adjusting for confounding. For each drug examined, statistical adjustment was made for the indications for prescription of that therapy. For example, in the case of b-adrenoceptor blockers and calcium channel entry blocking drugs, adjustment was made for the effects of arterial hypertension and angina, and in the case of diuretics adjustment was made for the effects of congestive cardiac failure and hypertension. Alternative regression models were compared using the log likelihood ratio test. The signi®cance of individual terms in the ®nal models was examined with the Wald test.6 This test was performed by dividing the coef®cient ratio obtained from the regression model by its standard error to obtain a standard normal deviate. The P value was obtained using a table of probabilities. By this method, we examined the effects of chronic treatment with b-adrenoceptor blockers, calcium channel entry blockers, diuretics, nitrates and digoxin. The small numbers of patients receiving ACE inhibitors prevented the drug being examined in a regression model. The results generated by the analyses are shown as odds ratios and their 95% con®dence intervals (CI).
Results Elective surgical patients Among the elective surgical patients, we identi®ed 177 patients who died of a cardiovascular cause with 30 days of surgery under general anaesthesia; 115 of these were studied
further after matching with controls. Of the other 62 patients, 37 were excluded because either the notes of the patient, control or both were missing, and the other 25 because no matched control could be identi®ed. The mean age of the cases was 73 yr (range 45±93) and that of the controls 71 yr (43±90). In 74 (64%) of the pairs, the case and the control were male. Twenty-four cases and 15 controls were taking b-adrenoceptor-blocking drugs, 16 cases and ®ve controls were taking calcium channel entry blocking drugs, two cases and no controls were taking ACE inhibitors; 48 cases and 29 controls were taking diuretics, 24 cases and six controls were taking nitrates, and 15 cases and ®ve controls were taking digoxin. A signi®cantly greater number of the cases were receiving calcium channel entry blockers, diuretics, digoxin and nitrates. The causes of death in the cases were myocardial infarction/ischaemic heart disease (78 deaths), heart failure (23) and cerebrovascular accident (14). The crude odds ratios for the association of individual drug therapies and adverse outcome and the adjusted ratios for confounding by residual age difference between the case-control pairs are shown in Table 1. Univariate ageadjusted ratios showed signi®cant associations with cardiac death for calcium channel entry blockade, diuretics, digoxin and nitrates (P<0.05), and a signi®cance of P=0.05 for b-adrenoceptor blocking drugs. The effects of b-adrenoceptor blockade and calcium channel blockade were examined together in a conditional logistic regression model, which also contained terms for arterial hypertension and angina, and a term to allow for the effects of residual confounding by age. Full details of the model are given in Table 2. In this model, the odds ratio for the association between the use of b-adrenoceptor blockade and postoperative cardiac death was 1.00 (95% CI 0.41±2.44); z=0.007, P=1.0. The odds ratio for the association of calcium channel entry blockers and cardiac death was 1.61 (0.36±7.12); z=0.62, P=0.53. The addition of a term for interaction between b-adrenoceptor blocking drugs and calcium channel antagonists did not improve the model signi®cantly. That is to say, there was no evidence in elective surgical patients of interaction between b-adrenoceptor antagonists and calcium entry blocking drugs.
508
Intercurrent drug therapy in surgery under general anaesthesia Table 3 Results for the association of individual drug therapies and cardiac death in urgent/emergency patients. Data are crude and age- and sex-adjusted odds ratios together with the 95% con®dence intervals (CI). P values refer to the age- and sex-adjusted odds ratio, and were calculated using the Wald test. *Lower CI estimate; no point estimate calculable
b-Adrenoceptor blockade Calcium channel entry blockade ACE inhibitors* Diuretics Digoxin Nitrates
Odds ratio (95% CI)
Age- and sex-adjusted odds ratio (95% CI)
P
5.00 (0.56±236.45) 3.50 (0.67±34.53) (1.18) 4.60 (1.71±15.49) 2.50 (0.92±7.86) 1.75 (0.45±8.25)
5.11 (0.59±44.20) 3.57 (0.73±17.48) (1.18) 4.95 (1.82±13.46) 2.61 (0.99±6.90) 1.93 (0.54±6.82)
0.138 0.117 0.032 0.002 0.053 0.310
For the other drug therapies, there were no signi®cant associations between diuretics or digoxin and cardiac death after the confounding effects of arterial hypertension, heart failure and other drug therapies (i.e. b-adrenoceptor blocking drugs and calcium channel entry blocking drugs) had been taken into account. There was, however, a signi®cant association between administration of nitrates and cardiac death after adjustment for the confounding effects of preexisting angina and residual age difference [odds ratio 4.79 (95% CI 1.00±22.72); z=1.943, P=0.049]. After adjustment for the effect of any associated administration of badrenoceptor blockade, the odds ratio increased to 5.72 (1.09±25.54); z=2.065, P=0.039.
Urgent and emergency patients In the urgent/emergency patients, we identi®ed 202 cases; controls were found for 73 of them. Of the 129 patients excluded, 34 underwent surgery under local or regional anaesthesia, 24 were admitted as emergency patients but initially received medical treatment rather than surgical care (surgery taking place some days after admission), 25 had untraceable notes or no matched control and 45 did not ful®l the study criteria (18 having pulmonary emboli, three noncardiovascular causes of death and 24 being incorrectly coded); the remaining patient died on the way to the operating theatre. Death was due to myocardial infarction/ ischaemic heart disease in 34 patients, cardiac failure in 16, cerebrovascular accident in 14, and other cardiac causes in the remaining nine. The mean age of the cases was 81 yr (range 40±98) and that of the controls was 81 yr (43±100). Five cases were receiving b-adrenoceptor blocking drugs and seven were receiving calcium channel entry blockers; the corresponding controls included one patient on b-adrenoceptor blockers and two on calcium channel entry blockers. For the other drug treatments, six cases and no controls were receiving ACE inhibitors, seven cases and four controls were receiving nitrates, 39 cases and 21 control were receiving diuretics, and 16 cases and seven controls were receiving digoxin. In the urgent/emergency patients, more cases were receiving diuretics and ACE inhibitors.
There were no signi®cant univariate associations between b-adrenoceptor blocking drugs, calcium channel entry blocking drugs or nitrates and cardiac death in the urgent/ emergency surgical patients; however, after adjusting for sex and residual age difference between the case±control pairs, there were signi®cant associations between diuretics and ACE inhibitors and cardiac death (Table 3). As with the elective surgical patients, there was no interaction between b-adrenoceptor blockade and calcium channel entry blocking drugs After adjusting for the effects of other confounders (such as angina, arterial hypertension and heart failure) by logistic regression analysis, there were no associations between drug therapies and cardiac death. In the case of diuretic therapy, the addition of a history of heart failure to the regression model reduced the odds ratio for diuretics from 4.29 to 1.94, while a history of heart failure reduced the odds ratio for digoxin from 2.61 to 1.34 (both ®nal odds being non-signi®cant).
Discussion In this case±control study, we found no signi®cant association in elective or emergency surgical patients between chronic b-adrenoceptor blockade and cardiac death within 30 days of anaesthesia. The analyses presented here were prompted by the con¯icting ®ndings of several other groups of researchers.1 2 7 8 In a randomized controlled trial of atenolol versus placebo, Mangano and colleagues demonstrated a reduction in mortality 6, 12 and 24 months after hospital discharge in patients randomized to receive the b-adrenoceptor blocking drug before operation and for 7 days after surgery.1 The main effect of atenolol treatment was seen in the ®rst 6 months after surgery, but no difference in in-hospital mortality was observed. In a second study, Poldermans and colleagues reported on 112 high cardiac risk patients with evidence of dobutamine stress-inducible echocardiographic (ECHO) abnormalities scheduled for major vascular surgery.2 After randomization, the treatment group received oral bisoprolol for at least 1 week before surgery, and continued until 30 days after operation. The incidence of cardiac deaths in the bisoprolol
509
Sear et al.
group during the 30-day peri-operative period was only 3.4% compared with 17% in the control group. There was also a difference in the incidence of non-fatal myocardial infarction (0% in the bisoprolol group vs 17% in the control group). In a subgroup consisting of 53 patients already maintained on b-adrenoceptor blocking drugs, there were four post-operative cardiac deaths (7.5%) and no non-fatal infarctions. These results are in contrast to the data of Mangano and colleagues, who found no effect of acute atenolol compared with placebo on in-hospital deaths.1 In two further studies in patients undergoing vascular surgery under general or regional anaesthesia, Yeager and colleagues found b-adrenoceptor blockade was associated with a 50% reduction in peri-operative myocardial infarction, but they do not report on peri-operative mortality,7 while Oliver and colleagues also found no effect of badrenoceptor blockade on the incidence of either myocardial infarction or cardiac death in patients receiving mivazerol or placebo.8 There are differences in study design between our study and these other studies. The latter used randomization of patients from high-risk populations for cardiovascular disease to ensure a comparable distribution of risk factors between the treatment and placebo groups. On the other hand, our observational study depended on examining the possible effects of chronic b-adrenoceptor therapy, having made statistical adjustments for the in¯uence of other cardiovascular risk factors. Speci®cally, adjustment was made for the indication for the prescription of the various medications. In the analysis of the data for elective surgery, we included terms for the two main indications for the prescription of b-adrenoceptor blocking drugs: arterial hypertension and angina. These differences may help explain the contrast between our ®ndings and those of Mangano and colleagues1 and Poldermans and colleagues.2 More recently, studies have shown that use of badrenoceptor blockade in patients with mild to moderate heart failure may have a signi®cant effect on cardiac outcome in these high-risk patients.9 Two studies have found b-adrenoceptor blockade to be associated with a reduction in total cardiac mortality and all-cause hospitalizations in patients with heart failure.10 11 The interaction of anaesthesia, b-blockade and heart failure has not yet been evaluated fully. In our present study populations, there were few patients who had a history of heart failure and were receiving b-adrenoceptor blocking drugs (four among the elective surgical patients and two in the urgent/emergency patients). However, there were no differences in death rates between patients who had a history of heart failure and who were receiving b-adrenoceptor blocking drugs compared with patients not receiving this treatment. Another possible explanation for the difference between acute and chronic b-blockade is that patients on long-term cardioselective b1 antagonists do not maintain sustained blockade as a result of b1 receptor adaptation. This could result from either receptor up-regulation or a compensatory
increase in cardiac b2 receptors, or a combination of the two.12 An alternative explanation might be that of inverse agonism by chronic b-blockade at G-protein-coupled receptors, as has been demonstrated in vitro in rat ®broblasts by Milligan and colleagues.13 There have been a number of studies examining the effects of medication with calcium channel antagonists on cardiac death in patients with stable angina or after infarction.14±16 There are few comparable data on cardiac outcomes in non-cardiac surgical patients receiving intercurrent calcium channel entry blocking drugs. We found an increased incidence of post-operative silent myocardial ischaemia in patients receiving calcium channel entry blocking drugs.17. Mangano and colleagues demonstrated an association between pre-operative calcium entry blocking drugs and post-operative ischaemia in non-cardiac surgical patients.18 However, multivariate regression analysis failed to support this association, or an association with either in-hospital or long-term mortality in US Veterans Administration male patients undergoing non-cardiac surgery.19 There have been reports that short-acting calcium channel entry blocking drugs (such as nifedipine) may be associated with an increased risk of cardiovascular events, including cardiac death. Thus, Psaty and colleagues found an association between short-acting calcium channel entry blocking drugs and the increased risk of myocardial infarction in hypertensive patients.20 However, this was not borne out in two further studies in patients receiving calcium channel entry blocking drugs for chronic coronary artery disease21 or hypertension.22 It is unlikely that the regression models used in our analyses make a complete adjustment for confounding risk factors for the effects of either chronic b-adrenoceptor or channel entry channel blockade. However, the fact that the odds ratios for both treatments moved towards unity after adjustment suggests no independent effect of these drugs. The slight association between nitrate administration and adverse cardiac outcomes in the elective surgical patients was surprising. However, Ishikawa and colleagues have shown long-term nitrate therapy given to post-infarction patients to be associated with an increased incidence of cardiac events, including cardiac death.23 There are no data on post-surgical cardiac complications in patients on longterm nitrates, although ®ve studies have looked at their effects on perioperative ischaemia. Two studies show a signi®cant reduction in intraoperative ischaemia24 25 while three others show no effect.26±28 The differences in outcome may re¯ect differences in underlying anaesthetic technique (volatile vs opioid-based) or differences in study populations. Our data may indicate the need for high-risk cardiac patients receiving only nitrates to be given additional antiischaemic drugs in the perioperative period. Other researchers have noted increased perioperative mortality associated with emergency surgery.29±31 In our cases who underwent urgent or emergency surgery, there
510
Intercurrent drug therapy in surgery under general anaesthesia
were no signi®cant associations between b-adrenoceptor or calcium channel entry blockade and 30-day cardiac death. Although diuretics and ACE inhibitor therapy had signi®cant associations with cardiac death on univariate analysis, these effects were rendered non-signi®cant by adjustment of the data for confounding factors (especially heart failure). The implication is that the risk associated with these drugs was due to the conditions for which they were prescribed. Heart failure is universally accepted as a strong predictor of post-operative myocardial infarction, reinfarction and worsening left ventricular function,32 and cardiac death.33 In conclusion, our case±control study failed to show any signi®cant effect of chronic b-adrenoceptor blockade on 30day survival after either elective or urgent/emergency noncardiac surgery. However, the CIs for the death rates after urgent/emergency surgery were wide, and we suggest that further studies using large databases, such as the ORLS, are warranted. Although chronic calcium channel entry blockade was associated with adverse cardiac outcome in elective surgical patients by univariate analysis, this was not sustained after adjustment for the effects of hypertension and angina. In the emergency or urgent surgery patients, there was a signi®cant association between digoxin therapy and cardiac death, supporting the ®ndings of Hollenberg and Mangano, who showed that pre-operative digoxin therapy was a signi®cant predictor of post-operative silent myocardial ischaemia and hence cardiac events.34 However, after adjustment for confounding due to a history of cardiac failure (a well-established risk factor for cardiac death following anaesthesia and surgery),5 33 digoxin therapy did not remain a signi®cant risk factor. As with our previous publications, caution should be exercised because of the assumed low power of the study, as indicated by the small numbers of patients receiving the therapies and the resulting large CIs of the odds ratios.
References 1 Mangano DT, Layug EL, Wallace A, Tateo I, for the Multicenter Study of Perioperative Ischemia Research Group. Effect of atenolol on mortality and cardiovascular morbidity after noncardiac surgery. New Engl J Med 1996; 335: 1713±20 2 Poldermans D, Boersma E, Bax JJ, et al. The effect of bisoprolol on perioperative mortality and myocardial infarction in high-risk patients undergoing vascular surgery. New Engl J Med 1999; 341: 1789±94 3 Howell SJ, Sear YM, Yeates D, Goldacre M, Sear JW, Foex P. Hypertension, admission blood pressure and perioperative cardiovascular risk. Anaesthesia 1996; 51: 1000±4 4 Howell SJ, Sear YM, Yeates D, Goldacre M, Sear JW, Foex P. Risk factors for cardiovascular death after elective surgery under general anaesthesia. Br J Anaesth 1998; 80: 14±19 5 Howell SJ, Sear YM, Sear JW, Yeates D, Goldacre M, Foex P. Cardiovascular death within 30 days following anaesthesia and urgent or emergency surgery. Determination of risk factors using a nested case±control study. Br J Anaesth 1999; 82: 679±84 6 Armitage P, Berry G. Statistical Methods in Medical Research, edn 3. Oxford: Blackwell Science, 1994; 429
511
7 Yeager RA, Moneta GL, Edwards JM, Taylor LM, McConnell DB, Porter JM. Reducing perioperative myocardial infarction following vascular surgery. The potential role of b-blockade. Arch Surg 1995; 130: 869±73 8 Oliver MF, Goldman L, Julian DG, Holme I. Effect of mivazerol on perioperative cardiac complications during non-cardiac surgery in patients with coronary heart disease. Anesthesiology 1999; 91: 951±61 9 Abraham WT. Beta-blockers: the new standard of therapy for mild heart failure. Arch Intern Med 2000; 160: 1237±47 10 Hjalmarson A, Goldstein S, Fagerberg B, et al., for the MERIT-HF Study Group. Effects of controlled-release metoprolol on total mortality, hospitalizations, and well-being in patients with heart failure: the Metoprolol CR/XL Randomized Intervention Trial in Congestive Heart Failure (MERIT-HF). J Am Med Assoc 2000; 283: 1295±302 11 Bonet S, Agusti A, Arnau JM, et al. Beta-adrenergic blocking agents in heart failure: bene®ts of vasodilating and nonvasodilating agents according to patients' characteristics: a meta-analysis of clinical trials. Arch Intern Med 2000; 160: 621±27 12 Yndgaard S, Lippert FK, Berthelsen PG. Are patients chronically treated with b1-adrenoceptor antagonists in fact b-blocked? J Cardiothoracic Vasc Anesth 1997; 11: 32±6 13 Milligan G, MacEwan DJ, Mercouris M, Mullaney I. Inverse agonism at adrenergic and opioid receptors: studies with wild type and constitutively active mutant receptors. Receptors Channels 1997; 5: 209±213 14 Heidenreich PA, McDonald KM, Hastie T, et al. Meta-analysis of trials comparing beta-blockers, calcium antagonists and nitrates for stable angina. J Am Med Assoc 1999; 281: 1927±36 15 Singh BN. When is drug therapy warranted to prevent sudden cardiac death? Drugs 1991; 41 Suppl 2: 24±46 16 Leitch JW, McElduff P, Dobson A, Heller R. Outcome with calcium channel antagonists after myocardial infarction: a community-based study. J Am Coll Cardiol 1998; 31: 111±7 17 Sear JW, Foex P, Howell SJ. Effect of chronic intercurrent medication with b-adrenoceptor blockade or calcium channel entry blockade on postoperative silent myocardial ischaemia. Br J Anaesth 2000; 84: 311±5 18 Hollenberg M, Mangano DT, Browner WS, et al., for the Study of Perioperative Ischemia Research Group. Predictors of postoperative myocardial ischemia in patients undergoing noncardiac surgery. J Am Med Assoc 1992; 268: 205±9 19 Browner WS, Li J, Mangano DT, for the Study of Perioperative Ischemia Research Group. In-hospital and long-term mortality in male veterans following non-cardiac surgery. J Am Med Assoc 1992; 268: 228±32 20 Psaty BM, Heckbert SR, Koepsell TD, et al. The risk of myocardial infarction associated with antihypertensive drug therapies. J Am Med Assoc 1995; 274: 620±5 21 Braun S, Boyko V, Behar S, et al. Calcium antagonists and mortality in patients with coronary artery disease: a cohort study of 11,575 patients. J Am Coll Cardiol 1996; 28: 7±11 22 Jick H, Derby LE, Gurewich V, Vasilakis C. The risk of myocardial infarction associated with antihypertensive treatment in persons with uncomplicated essential hypertension. Pharmacotherapy 1996; 16: 321±6 23 Ishikawa K, Kanamasa K, Ogawa I, et al. Long-term nitrate treatment increases cardiac events in patients with healed myocardial infarction. Japan Circ J 1996; 60: 779±88 24 Coriat P, Daloz M, Bousseau D, et al. Prevention of intraoperative myocardial ischemia during noncardiac surgery with intravenous nitroglycerin. Anesthesiology 1984; 63: 193±6 25 Fusciardi J, Godet G, Bernard J, et al. Roles of fentanyl and
Sear et al.
26
27 28 29
nitroglycerin in prevention of myocardial ischemia associated with laryngoscopy and tracheal intubation in patients undergoing operations of short duration. Anesth Analg 1986; 65: 617±24 Thomson J, Mutch W, Culligan J. Failure of intravenous nitroglycerin to prevent intraoperative myocardial ischemia during fentanyl±pancuronium anesthesia. Anesthesiology 1984; 61: 385±93 Gallagher J, Moore R, Jose A, et al. Prophylactic nitroglycerin infusions during coronary artery bypass surgery. Anesthesiology 1986; 64: 785±9 Dodds T, Stone G, Coromilas J. Prophylactic nitroglycerin infusion during noncardiac surgery does not reduce perioperative ischemia. Anesth Analg 1993; 76: 705±13 Fowkes FGR, Lunn JN, Farrow SC, Robertson IB, Samuel P. Epidemiology in anaesthesia. III: Mortality in patients with coexisting physical disease. Br J Anaesth 1982; 54: 819±25
512
30 Goldman L, Caldera DL, Nussbaum SR, et al. Multifactorial index of cardiac risk in noncardiac surgical procedures. New Engl J Med 1977; 297: 845±50 31 Pedersen T, Elliasen K, Henrikson E. A prospective study of mortality associated with anaesthesia and surgery: risk indicators of mortality in hospital. Acta Anaesthesiol Scand 1990; 34: 176±82 32 Larsen SF, Olesen KH, Jacobsen E, et al. Prediction of cardiac risk in non-cardiac surgery. Eur Heart J 1987; 8: 179±85 33 Dirksen A, Kjoller E. Cardiac predictors of death after noncardiac surgery evaluated by intention to treat. Br Med J 1988; 297: 1011±13 34 Hollenberg M, Mangano DT. Therapeutic approaches to postoperative ischemia. The Study of Perioperative Ischemia Research Group. Am J Cardiol 1994; 73: 30B±33B