Randomized controlled trials and meta-analyses provide conflicting guidance on the role of beta-adrenergic receptor blockers (beta-blockers) in reducing perioperative complications. We hypothesize that variability in trial results may be due in part to heterogeneous properties of beta-blockers. First, we propose that the extent of beta-blocker metabolism by cytochrome P-450 and the time available to titrate the dosage before surgery (titration time) may interact; dependence on P-450 may be most harmful when titration time is short. Second, beta-blockers vary in their selectivity for the beta-1 receptor and reduced selectivity may contribute to cerebral ischemia.
We used meta-analysis and meta-regression of existing trials to explore the role of these pharmacological properties.
We found that both of these pharmacological factors are significantly associated with reduced efficacy of beta-blockers.
Pharmacological properties of beta-blockers may contribute to heterogeneous trial results. Many trials have used metoprolol, which is extensively metabolized by cytochrome P450 and is less selective for the beta-1 receptor. For these two reasons, the efficacy of metoprolol to prevent perioperative cardiac complications should be compared with the efficacy of other beta-blockers.
What We Already Know about This Topic
❖ The role of β-blockers in reducing overall mortality is unclear.
What This Article Tells Us That Is New
❖ The extent of β-blocker metabolism by CYP2D6, the time available to titrate β-blocker dosage preoperatively, and variations in β-blocker selectivity for the β-1 adrenergic receptor may contribute to the heterogeneous results of randomized controlled trials of perioperative β-blockade.
❖ Metoprolol should probably not be used for perioperative β-blockade when there is insufficient time to titrate its dose.
THE 2009 American College of Cardiology Foundation and American Heart Association practice guidelines recommend “careful titration” of β-adrenergic receptor blockers (β-blockers) in selected patients.1In the most comprehensive meta-analysis of perioperative β-blockers, Bangalore et al. 2concluded that there was no reduction in total mortality and that heterogeneity in results regarding benefit was likely due to variable presence of bias in the trials. In that meta-analysis, trials reporting the greatest benefit from β-blockers were those deemed to be at most risk of bias. A more recent meta-analysis that included the Dutch Echocardiographic Cardiac Risk Evaluation Applying Stress Echocardiography (DECREASE) IV trial and selected trials from the meta-analysis by Bangalore et al. 2provided an alternative interpretation.3Poldermans et al. 3concluded that β-blockers are safe if adequate time exists to titrate the dose before surgery.3Although we agree with this claim, it does not address the use of β-blockers when the titration time is short.
We propose two additional hypotheses. First, the benefit of β-blockers is reduced for β-blockers metabolized by the CYP2D6 isoenzyme of cytochrome P-450. This reduction occurs because individual variations in CYP2D6 activity, as a result of inheritance and drug interactions, may cause both insufficient and excessive β-blockade. Second, the reduced β-1 to β-2 adrenergic receptor selectivity may reduce the physiologic response to surgical anemia. If these hypotheses are correct, then use of β-blockers such as metoprolol, which are metabolized by CYP2D6 and have relatively low β-1 to β-2 selectivity, may not be appropriate choices for perioperative β-blockade, especially when little time for titration is available before surgery.
Review of Selected Pharmacological Aspects of β-Blockers
Potentially relevant polymorphisms identified to date affect β-blockade via β adrenergic receptors,4,5G-protein-coupled receptor kinases,‡and metabolism by cytochrome P-450.§Polymorphisms of β adrenergic receptors are the most extensively studied.5In the β Blocker in Spinal Anesthesia trial,4bisoprolol-related bradycardia and the c.16G>A polymorphism of ADRB2 were associated with hypotension. Among patients with acute coronary syndrome, c.46G>A and c.79C>G polymorphisms of ADRB2 may affect mortality among patients treated with β-blockers.5Although polymorphisms of the β adrenergic receptors are clinically important, they have not yet been shown to differentially affect specific β-blockers.
P-450 polymorphisms may be uniquely important, because we already know that β-blockers differ in their dependency on cytochrome P-450 for metabolism. Many (metoprolol, propranolol, carvedilol, labetalol, timolol) are metabolized by the P-450 CYP2D6 isoenzyme, and metoprolol is the most dependent, with 70–80% of its metabolism by CYP2D6.6The CYP2D6 isoenzyme may be the most problematic of the many cytochrome P-450 isoenzymes. CYP2D6 is estimated to metabolize 25% of prescribed drugs7and to underlie 38% of the most frequently cited adverse drug reactions.8Patients using β-blockers metabolized by CYP2D6 are more susceptible to bradycardia caused by lower functioning polymorphisms as well as drug interactions. Among European and American subjects, 5–10% have reduced function of CYP2D6.§Other patients may be hyperfunctioning metabolizers.9Although these variations may not be important in managing chronic illness, where doses can be slowly titrated,9they may be very relevant in acute settings and have not been studied perioperatively.
Ratio of β-1 to β-2 Adrenergic Receptor Selectivity.
Even among β-blockers that are β-1 cardioselective, variations occur in the ratio of β-1 to β-2 receptor affinity. Among the β-blockers used perioperatively, the β-1/β-2 affinity ratios range from 13.5 for bisoprolol to 4.7 for atenolol and 2.3 for metoprolol.10The benefit of β-blockers is probably due to blunting tachycardia mediated by β-1 receptors.11,12The role of β-2 adrenergic receptors in perioperative care is uncertain. On one hand, β-2 blockade may prevent hypotension by blocking systemic vasodilation, whereas, on the other hand, β-2 blockade might cause cerebral ischemia by blocking cerebral vasodilation. An animal study has addressed the effects of metoprolol during acute anemia. Without metoprolol, brain oxygenation was preserved during acute anemia but kidney oxygenation fell.13However, during β-blockade with metoprolol, acute anemia led to loss of both cerebral and kidney oxygenation.13Although observational studies in humans have found that adverse effects from β-blockers are increased in patients with anemia,14it has not been established that the interaction between anemia and β-blockers is related to the ratio of β-1 to β-2 adrenergic receptor selectivity. In summary, β-blockade may be more harmful during acute anemia and the harm may be worse with increasing β-2 blockade.
β-Blocker Duration of Effect.
Starting in the 1990s, it was recognized that some β-blockers, such as atenolol, should be dosed twice a day for hypertension.15We have no reason to think that dosing principles should be different in the perioperative setting. This may complicate interpretation of trials in which atenolol was used once daily.16,17
The benefits and risks of shorter- versus longer-acting β-blockers need consideration. We do not know whether longer-acting drugs may be more beneficial because they prevent rebound if doses are missed. Alternatively, shorter-acting β-blockers may be better perioperatively to avoid short-term toxicity and allow better titration, as has been shown in one study of angiotensin-converting enzyme inhibitors in heart failure.18
Materials and Methods
For our hypothesis-generating analysis, we included only studies of patients at increased risk of perioperative cardiac complications in accord with practice guidelines.19,20Thus, we included from the Bangalore meta-analysis studies of patients who had a Revised Cardiac Risk Index of 1 or greater without counting surgery type as one of the criteria. This led to excluding the trials by Brady et al. 21and Yang et al .22Because perioperative ischemia occurs both during and after surgery,23we restricted our analysis to trials in which β-blocker administration was begun before induction of anesthesia and continued postoperatively. Therefore, seven trials met our inclusion criteria4,16,17,24–27: six from the meta-analysis by Bangalore et al. 2and the more recent DECREASE IV trial.25Compared with the recent meta-analysis by Poldermans et al. ,3we excluded the trials by Brady et al. 21and Yang et al. 22and included the DECREASE IV trial25and the trial by Neary et al. 17
We restricted the outcomes to total mortality and stroke during hospitalization or at 30 days. In determining which trials had high risk of biased results, we used the judgments from the meta-analysis by Bangalore et al .2Bias was considered high if any of the first three items of the modified Cochrane Collaboration tool (randomization method, allocation concealment before and during enrollment, blinding) were not adequate.
Meta-analysis was done with RevMan software (The Cochrane Collaboration, Copenhagen, Denmark). We used more conservative random effects models instead of fixed effects models to yield conservative results because of heterogeneity in some of the analyses. Studies were weighted by the Dersimonian and Laird variation of the inverse of the variance.
Heterogeneity of study results was measured using the I2statistic, as recommended by the Cochrane Collaboration.28The importance of heterogeneity in results as measured by the I2is described by the Cochrane as follows: 0–40%, might not be important; 30–60%, may represent moderate heterogeneity; 50–90%, may represent substantial heterogeneity; and 75–100%, may represent considerable heterogeneity.
To explore reasons for heterogeneous results and to test for interactions between subgroups of trials, we used meta-regression, also using Dersimonian and Laird weighting.∥To analyze the interaction of length of titration period and metabolism by CYP2D6, we created an interaction variable whose value for studies was as follows: 0 if titration period was short and metoprolol was used; 1 if the titration period was short and a β-blocker without metabolism by CYP2D6 was used; and 2 if the titration period was long. Meta-regression was done with the R programming language (R: A Language and Environment for Statistical Computing. Version 2.10.1. Vienna, Austria) using the rmeta module.#
An evidence table of the trials is tabulated and maintained online.**When the seven included trials are pooled, there is moderate heterogeneity (I2= 42%) in the mortality reported in the trials. We explored four possible sources for the heterogeneity.
The Role of Study Bias
Banglore et al. 2proposed that a major source of heterogeneity was study quality. Trials with adequate description of methods of allocation and blinding based on the scale of the Cochrane Collaboration showed significant harm from β-blockers (pooled odds ratio, 1.34; 95% confidence interval [CI], 1.04–1.69). In our analysis of the nine trials meeting our inclusion criteria (fig. 1), meta-regression showed that the presence of bias was a significant predictor of drug efficacy (P = 0.038).
The Role of Length of Titration Period
Poldermans et al. 3meta-analyzed the rates of perioperative myocardial infarction and stroke on the basis of the amount of time to titrate the β-blocker.3When we replicated this analysis, using the studies meeting our inclusion criteria, the resulting two groups of studies contain the same trials as the analysis based on bias, because the two DECREASE studies are the only two studies with a long duration of titration and the only two studies deemed to have high bias by the Banglore et al. 2version of the Cochrane tool (fig. 2). Thus, among studies with a short titration period, the pooled odds ratio is again 1.34 (95% CI, 1.04–1.69), and by meta-regression, the P value is again significant at 0.038.
A New Analysis Based on the Metabolism of β-Blockers
We proposed that the degree of metabolism by the CYP2D6 isoenzyme of cytochrome P-450 is another plausible explanation for heterogeneity (fig. 2). In figure 3, the studies are further grouped by both metabolism and duration of titration. This analysis suggests that increased mortality from perioperative β-blockers is confined to trials that combined a short titration period with CYP2D6 metabolism (pooled odds ratio, 1.34; 95% CI, 1.05–1.72). Meta-regression of the interaction between these two factors yields a statistically significant correlation (P = 0.044), with the highest mortality in the short titration-CYP2D6 trials, intermediate mortality in the short titration-no CYP2D6 trials, and the lowest mortality in the long titration trials (fig. 4).
We found no interaction between the route of metabolism and the odds ratio of stroke from β-blockers. Data were sparse because two trials that did not use a β-blocker metabolized by CYP2D6 did not report the incidence of perioperative stroke.16,17
A New Analysis Based on Ratio of β-1/β-2 Selectivity
We proposed that the ratio of β-1/β-2 selectivity is another possible cause of heterogeneity. Figure 5is a meta-regression of the β-1/β-2 affinity ratios. This analysis suggests that benefit from β-blockers correlates with cardioselectivity (P = 0.046).
We found no interaction between the ratio of β-1 to β-2 selectivity and the odds ratio of stroke from β-blockers. Again, data for stroke are not reported in all trials.
β-Blockers vary in their pharmacology, and, in our analysis, both the metabolic pathway and the degree of β-1 selectivity of the β-blocker showed a statistically significant interaction with benefit on total mortality. These observations provide alternative or additional explanations for the source of heterogeneous trial results. These findings contrast with the conclusion by Bangalore et al. 2that there is “potentially increased mortality” from using any β-blocker. In the analysis of mortality by Bangalore et al. ,2after grouping trials by bias, substantial heterogeneity remained in the subgroup of trials with high bias (I2= 62%), whereas low heterogeneity remained in the subgroup of trials with low bias (I2= 28%)29; however, sources for this heterogeneity were not explored.
Our finding that metabolic pathway is associated with efficacy of β-blockers is consistent with a prior meta-analysis of trials through 2003.11Beattie et al. 11concluded that controlling heart rate correlated with fewer perioperative myocardial infarctions and that heart rate was less consistently controlled with metoprolol. Our finding that the degree of β-1 selectivity is associated with efficacy of β-blockers is consistent with the two major trials of early β-blocker use for acute coronary syndrome. In acute coronary syndrome, quick titration of β-blockade is needed. The Clopidogrel and Metoprolol in Myocardial Infarction Trial (COMMIT)30found no benefit from metoprolol, whereas the First International Study of Infarct Survival (ISIS-1)31found benefit from atenolol. Both trials reported a 5% rate of hypotension from β-blockade; however, atenolol reduced vascular mortality whereas metoprolol did not.
Both findings, reduced efficacy associated with both CYP2D6 metabolism and low ratio of β-1 of β-2 selectivity, are consistent with a large retrospective cohort study.32Redelmeier et al. 32found that atenolol was associated with greater reduction in mortality compared with metoprolol in a cohort of 37,151 patients. Whether this might be due to more sympathetic rebound with the shorter half-life of metoprolol, as proposed by Redelmeier et al .,32inconsistent sympatholysis by metoprolol in patients with abnormal cytochrome P-450 CYP2D6 activity, or reduced cerebral protection from low β-1 selectivity is not known.
Our stroke results are consistent with the meta-analysis by Poldermans et al. 3and show that the risk of stroke does not vary by method of β-blockade. However, two of the seven studies, both using drugs not metabolized by CYP2D6, reported no strokes, and so our results may exaggerate the rate of stroke in this group as a result of selective outcome reporting bias.33The impact of method of β-blockade on stroke needs more investigation.
Bangalore et al. 2concluded that the heterogeneity across trials was due to biases, including lack of blinding. Although we recognize the importance of blinding in general, it may be less important for the clinical question of perioperative β-blockers for two reasons. First, trials of β-blockers are difficult to blind if the values for heart rates are not masked.34In addition, blinding may actually interfere with achieving goal heart rates.11Thus, we are not certain that the studies with stated blinding are better than those without stated blinding. Moreover, if efficacy is measured by an objective outcome such as total mortality, then the role of blinding is less important.35
Other important markers of study quality should be added to blinding. The DECREASE I trial was stopped early. We doubt that the lack of blinding in the DECREASE I trial would change the direction of the results in our analysis of total mortality. However, we agree that the early termination exaggerates the benefit of treatment.36This may contribute to the moderate heterogeneity found when pooling the two DECREASE trials in group 1.31 of figure 3.
All four explanations (study bias, duration of titration period, presence of metabolism by CYP2D6, and β-1 selectivity) show statistically significant ability to explain heterogeneity in total mortality. With the studies to date, we believe that none of these explanations can be rejected and all may interact. For example, when patients are not anemic and sufficient time exists to titrate a β-blocker (such as 5 weeks, as in the DECREASE studies), the choice of β-blocker may not matter. Studies of metoprolol in outpatients have found increased adverse effects37and dose adjustments38among patients who are genetically slow metabolizers of metoprolol; however, when sufficient time permits dose adjustment, small studies do not report an increase adverse events among slow metabolizers.9,39However, our analysis suggests that when little time is available, metoprolol may be hazardous because of drug interactions and genetic variability in metabolism.
We note the contrasting results in the two trials that used atenolol, the Multicenter Study of Perioperative Ischemia Research Group trial and the trial of Neary et al. ,17in the middle group of figure 3. Neither trial found a statistically significant impact on mortality, and the different results could be due to small study size and random error. However, we note that Neary et al. ,2who used a lower dose of atenolol, found a trend toward benefit, whereas the Multicenter Study of Perioperative Ischemia Research Group trial used a higher dose and found a trend toward increased mortality. Both studies dosed atenolol once per day, although twice a day may be more effective.40,41
A randomized controlled trial comparing outcomes after starting β-blockers with and without favorable pharmacological properties, with patients stratified by length of the titration period, could test this hypothesis. If atenolol is used in the study, we believe it should be dosed twice a day. The trial could be limited to patients with known abnormal alleles. However, an observational study may be easier. For example, a case control study of patients using metoprolol could compare the prevalence of abnormal P-450 alleles and other medications metabolized by CYP2D6 among patients with and without perioperative complications. The study design could be more efficient by only including patients at high cardiac risk or taking other medications metabolized by CYP2D6 who recently started taking metoprolol. Likewise, a cohort design could be used, as was done in investigating cardiovascular events due to patients taking clopidogrel combined with proton pump inhibitors42or the presence of hypofunctioning CYP2C19 alleles.43
We believe that future studies and meta-analyses of β-blockers for preventing perioperative morbidity should consider pharmacological properties of β-blockers. How should clinicians manage perioperative β-blockade pending further research to clarify the predictors of benefit from β-blockers? The only trials with significant results used bisoprolol. Our analysis supports two physiologic reasons why bisoprolol and atenolol may be safer than metoprolol. Thus both empirical and theoretical evidence favor medications other than metoprolol. Using the Diamond and Kaul44schema of assessing evidence, we propose that there is “reasonable suspicion” of harm from metoprolol and that clinicians should consider not starting metoprolol preoperatively unless a long period of titration is available or a patient needs hepatic metabolism as a result of reduced renal function.
The authors acknowledge the role of the anonymous reviewer who suggested that the ratio of β-1/β-2 selectivity may be a cause of heterogeneity and who referred us to an important study regarding this hypothesis.