Preoperative β-Blocker Use and Mortality and Morbidity Following CABG Surgery in North America FREE

During the past 2 decades, β-adrenergic blockade has been demonstrated to improve acute outcomes and long-term prognosis in ischemic heart disease.^{1 - 9} β-Blocker therapy has also been demonstrated to reduce perioperative events among high-risk patients undergoing major noncardiac and vascular surgery.^{10 - 13}

To date, however, randomized studies have not examined whether β-blocker therapy is beneficial when used preoperatively in patients undergoing coronary artery bypass graft surgery (CABG). Although extrapolation of the cardioprotective benefits of β-blockers from major noncardiac and vascular settings to cardiac surgery is plausible, concerns exist that this treatment in CABG patients may be detrimental due to depression of myocardial contractility¹⁴ and/or exacerbation of underlying reactive airway disease.¹⁵ Accordingly, many cardiothoracic surgeons have not considered use of preoperative β-blocker therapy in their CABG patients.

Using the Society of Thoracic Surgeons (STS) National Adult Cardiac Surgery Database (NCD), we analyzed the use of preoperative β-blockade and predictors of use among patients undergoing isolated CABG at 497 hospitals in the United States and Canada. We also determined whether preoperative β-blockade was associated with improvement in 30-day mortality and major morbidity following CABG, and, if so, in what subsets of patients.

The NCD was established in 1989 for the purpose of outcomes assessment following adult cardiac surgery.¹⁶ Data from the NCD are collected semiannually from the majority of US and Canadian hospitals performing open heart surgery. Clinical data are entered at the sites using uniform definitions and certified software systems. Data quality standards must be met before a local data set can be entered into the aggregate national data set. Data are warehoused at the Duke Clinical Research Institute in Durham, NC, which produces semiannual site-specific reports to NCD participants for outcomes analysis and quality improvement.

Outcomes definitions in the NCD include operative mortality (death within 30 days of surgery, whether in or out of hospital, or death in hospital regardless of length of stay); permanent stroke (new-onset cerebrovascular accident persisting >72 hours); renal failure (acute postoperative renal insufficiency with ≥1 of the following: increase in serum creatinine >2.0 mg/dL (177 µmol/L), ≥50% increase in creatinine level over baseline preoperative value, or new requirement for dialysis); reoperation (reexploration for bleeding, graft occlusion, other cardiac problem, or other noncardiac problem); deep sternal infection (involving muscle, bone, and/or mediastinum and including ≥1 of the following: wound opened with excision of tissue, positive culture, or treatment with antibiotics); and prolonged ventilation (pulmonary insufficiency requiring ventilatory support for ≥48 hours). The semiannual executive summary and remaining definitions are available at http://www.ctsnet.org/doc/2167.

The study population was derived from patients in the NCD who underwent CABG between 1996 and 1999. Patients were excluded if they underwent concomitant valve surgery or other cardiac procedures. We also excluded cases for which information on β-blocker use was not available (6.4% of total patients collected). Finally, we excluded information from NCD sites that reported less than 20% β-blocker use (1.6% of total patients) due to concerns about data accuracy. Reanalysis after inclusion of these sites did not materially alter the results.

We compared baseline demographics for patients who received β-blocker therapy with those who did not. Differences among treated and nontreated patients were determined using the χ² test for categorical variables and the Wilcoxon rank sum test for continuous variables. Missing data values for dichotomous variables other than β-blocker use were assigned the most frequent value, while continuous variables were assigned the median value, except for body surface area, which was assigned the sex-specific median value.

Trends in β-blocker use over time were determined using logistic regression analysis, whereby the independent variable was treatment and the dependent variable was month of surgery. The major predictors of β-blocker use were determined using a random-effects logistic model that included 31 relevant preoperative risk factors and a random site effect. Odds ratios (ORs) and 95% confidence intervals (CIs) were generated for each variable in the model to determine the strength of its influence in predicting preoperative β-blocker therapy.

The unadjusted effects of preoperative β-blocker use on operative mortality and the 5 major postoperative morbidity causes (defined herein) were assessed using logistic regression analysis. Because treatment assignment was nonrandom, we also compared outcomes after controlling for patient and site differences using 2 techniques, direct risk adjustment and propensity-matched analysis.

For our risk-adjusted comparison, a multivariable logistic regression model was used to determine β-blocker effects after simultaneously adjusting for up to 27 preoperative patient risk factors contained in the STS CABG mortality and morbidity models.¹⁷ Treatment effects also can be confounded at the site level.^{18 - 19} For example, sites that use preoperative β-blocker therapy generally may provide higher quality of care and have better surgical outcomes. To account for this, our analysis also considered site effects, as well as the percentage of β-blockers used at a site, using a hierarchical mixed-effects logistic model.^{18 - 19} Risk-adjusted ORs with 95% CIs were generated to estimate the within-site effect of β-blocker use.

The second method of adjustment involved matching patients with similar probability of receiving β-blocker therapy (ie, their propensity score).^{20 - 21} Propensity scores were developed using 2 models; one considered patient factors alone and the other considered both patient factors and site. Matched pairs were then generated for each set of propensity scores, and baseline characteristics were again compared among those treated vs not treated. Differences in clinical variables were tested using the χ² test for categorical variables and the signed rank test for continuous variables. Conditional logistic regression was used to determine the overall effect of β-blocker therapy in these matched-pairs groups.

We examined the effect of β-blocker use on prespecified high-risk subgroups, including elderly patients (≥65 vs <65 years old); women (vs men); and those with a history of chronic lung disease, heart failure, depressed ventricular function (left ventricular ejection fraction [LVEF] ≥50%; LVEF ≥30% but <50%; or LVEF <30%), and diabetes. Treatment effects in these subpopulations were assessed by adding an interaction term to the conditional logistic model for propensity-matched patients. All analyses were performed using SAS Version 8.1 (SAS Institute Inc, Cary, NC).

The study population was derived from 684 426 patients who underwent isolated CABG and were entered into the NCD between 1996 and 1999. We excluded 43 618 (6.4%) whose β-blocker use status was unknown and another 10 931 (1.6%) who had surgery at 1 of 14 centers with less than 20% reported use of β-blockade. The final sample consisted of 629 877 patients from 497 sites.

Use of preoperative β-blockade in the NCD increased from 50.0% in 1996 to 60.0% in 1999 (P<.001 for time trend). Utilization rates, however, varied substantially during this time across the 497 sites (interquartile range, 43%-63%).

Table 1 presents the overall clinical characteristics for patients who received β-blocker therapy compared with those who did not. The age and sex distributions were similar in both groups. Preoperative β-blocker use was lower in patients with preoperative congestive heart failure, chronic lung disease, cardiogenic shock, diabetes, renal failure, and concomitant peripheral vascular disease. In contrast, β-blocker use was more common in patients with recent myocardial infarction or triple-vessel disease and those undergoing reoperation.

The 31 variables used as predictors of preoperative β-blocker use in the propensity analysis are available from the authors. The most important factors increasing likelihood of β-blocker use (aside from reoperation status) were recent MI, hypertension, worse angina, younger age, and higher LVEF. The strongest predictors of nonuse of β-blockers were chronic lung disease, history of congestive heart failure, and type 1 diabetes. In addition, more emergent surgical status and percutaneous transluminal coronary angioplasty in the 6 hours prior to surgery were strong predictors of nonuse of preoperative β-blockers (c statistic, 0.60 for the logistic regression model).

Patients who received β-blockers had lower mortality rates than those who did not; unadjusted operative mortality was 2.8% vs 3.4% (unadjusted OR, 0.80; 95% CI, 0.78-0.82; P<.001). Unadjusted morbidity rates were also lower for stroke, 1.6% vs 1.7% (OR, 0.93; 95% CI, 0.89-0.96); prolonged ventilation, 5.6% vs 6.6% (OR, 0.85; 95% CI, 0.83-0.87); reoperation, 4.9% vs 5.3% (OR, 0.94; 95% CI, 0.92-0.96); renal failure, 3.4% vs 3.8% (OR, 0.87; 95% CI, 0.85-0.89); and deep sternal infection, 0.62% vs 0.64% (OR, 0.98; 95% CI, 0.92-1.04).

For major risk-adjusted outcomes, patients receiving β-blockers had significantly lower odds of operative mortality than those not receiving β-blockers after controlling for 27 patient risk factors and site effects (Table 2). Among the major complications, β-blocker therapy was associated with lower likelihood of prolonged ventilation and renal failure after controlling for both patient and site effects.

In addition to the patient-level analysis, use of β-blocker therapy at the site level was also examined. Figure 1 shows that risk-adjusted mortality declined as the percentage of β-blocker use increased across sites.

After matching patients with similar baseline propensity scores, the major differences in baseline clinical characteristics among those treated with preoperative β-blockers vs those not treated were minimized or eliminated (Table 3). After matching patients with similar treatment propensity, those receiving preoperative β-blockers had lower operative mortality than those not treated (OR, 0.92; 95% CI, 0.89-0.94) (Table 4). Further adjustment for the effects of site on treatment propensity reduced the measured treatment effects (OR, 0.97; 95% CI, 0.93-1.00) (Table 4). In both of these analyses, the major complication outcomes of prolonged ventilation and renal failure were also improved in the β-blocker group relative to those not treated.

Figure 2 illustrates the effects of β-blocker therapy among prespecified patient subgroups. After matching patients with similar treatment propensity (both patient and site effects), 30-day mortality was lower with β-blockers in men and women, patients younger than 65 years, patients with a normal LVEF, and patients with diabetes and chronic lung disease. The only subgroup in which β-blocker use was associated with increased mortality was that of patients with an LVEF of less than 30%, but this was not statistically significant (P = .23).

This analysis is the first, to our knowledge, to report on North American population-based patterns of preoperative β-blocker use in CABG and its impact on major outcomes. We found that nearly 40% of contemporary CABG patients are not receiving this therapy. We also found that patients treated with preoperative β-blocker therapy had slightly lower operative mortality and morbidity rates, a positive effect that persisted after adjustment for multiple potential patient and site confounding factors. These data on the effect of β-blockade in bypass surgery are consistent with those previously identified for β-blockers when used prior to noncardiac surgery or percutaneous coronary intervention.^{22 - 23}

Among all sites, only 60% of patients received preoperative β-blocker therapy in 1999. While this rate improved by 10% during the study period, nearly a 4-fold difference remains among the 497 sites in their use of β-blockade (from 20% of patients at lowest-use hospitals to 85% among highest users).

We also documented that patients with higher-risk features (eg, diabetes, heart failure, underlying lung disease, older age) were less likely to be treated with β-blockers. However, despite the known higher risk of adverse effects of β-blockers, in fact, each of these subgroups had lower adjusted operative mortality when treated with β-blockers (Table 4).

To date, there have been no randomized trials evaluating preoperative β-blockade in CABG patients. In a single-institution observational study, Weightman et al²⁴ found an independent beneficial effect of preoperative β-blockade (unadjusted in-hospital mortality relative risk, 0.4; 95% CI, 0.2-0.8). Fenninger et al²⁵ identified a lower heart rate at the time of anesthesia induction for CABG as a predictor of perioperative events; however, it was not possible to determine whether the lower heart rate was due to use of β-blockers or improved hemodynamic stability. Finally, Chen et al⁶ reported that elderly CABG patients discharged receiving β-blockers had improved 1-year adjusted survival rates compared with those not receiving β-blockers. However, this study did not examine the timing of β-blocker initiation (preoperative or postoperative) or its acute benefits.

Our current analysis extends these previous findings. First, we used a variety of statistical techniques to control for baseline differences in clinical risk between those treated with β-blocker therapy vs not. In addition to patient-related risk factors, observational analyses can also be confounded by site-level influences,^{18 - 19} a factor not often accounted for in published treatment comparisons. Our study found that sites that more often used preoperative β-blocker therapy had better surgical outcomes in general (Figure 1). While a portion of the beneficial effect of β-blocker therapy can be accounted for by these patient and site factors, a small but statistically significant treatment effect persisted after controlling for these factors using both risk adjustment and matched treatment propensity analyses.

This report is also the first to examine the effects of preoperative β-blocker use on major morbidity following CABG. The incidence of prolonged ventilation and renal failure was significantly less in the β-blocker group compared with those not receiving a β-blocker after propensity-matched analysis. The trend toward a decrease in stroke in the β-blocker group is also interesting in that they are similar to the single-institution findings of Grigore et al,²⁶ who found that intraoperative use of β-blocker therapy was associated with a decreased risk of major neurologic complications following cardiac surgery.

Finally, our study provides consistent data demonstrating that β-blocker use is safe and possibly effective in many CABG patients previously thought to be at high risk for adverse effects of β-blocker therapy.^{27 - 31} This included women, elderly patients, and those with diabetes or congestive heart failure symptoms (Figure 2). In each of the high-risk subgroups, we found that those treated with preoperative β-blocker therapy tended to have lower operative mortality and morbidity than nontreated patients.

Severe LVEF impairment and severe chronic lung disease still remain relative contraindications to β-blocker therapy by American Medical Association and National Committee for Quality Assurance standards, even with more selective β₂ antagonists available.^{2 ,32} In our study, patients with chronic lung disease may have benefited from therapy, but no severity breakdown was performed. β-Blocker therapy likewise appeared safe in patients with mild-to-moderate ventricular dysfunction (LVEF ≥30%) and in patients with CHF. Among those with severe ventricular dysfunction (LVEF <30%), β-blocker therapy was associated with a trend toward worse outcomes and, thus, should be used with caution in these patients pending further randomized data. Although practitioners will continue to make individualized decisions regarding the safety of β-blocker therapy in patients undergoing CABG, these data suggest that it is safe in high-risk patients with all but the most severe forms of underlying disease states.

Our study adds supportive evidence that β-blockade may improve CABG outcomes; however, it was not designed to investigate the mechanism of action. Wallace et al¹⁰ and Mangano et al¹¹ attributed the protective effect of β-blocker therapy to a reduction of ischemic events during major procedures. A similar mechanism has been implicated for major vascular procedures.¹² Fenninger et al²⁵ have suggested that the potential benefit of β-blocker therapy may derive from its autonomic effects. Further studies will be necessary to determine if preoperative β-blockade affects markers of myocardial necrosis or perioperative ischemic events in CABG patients, as has been demonstrated in the setting of percutaneous coronary intervention.^{25 - 26} Likewise, it will be interesting to see whether the beneficial effects of β-blockers extend to those who undergo off-pump revascularization procedures.^{33 - 34}

The present study has several limitations. Most importantly, this is a large observational analysis and not a randomized trial of the efficacy of preoperative β-blocker therapy.³⁵ Accounting for both patient and site effects reduced but did not entirely eliminate the measured benefit of preoperative β-blockade. It is important to emphasize that this association may still be confounded by unmeasured selection biases.

Our NCD analyses may in fact underestimate the actual treatment benefits. For example, any miscoding of preoperative β-blocker use in the NCD would bias our measured results toward a null effect. Additionally, we defined treatment as receiving any β-blocker at any time and any dosage prior to surgery. Because therapy may not have been optimized in all patients, the measured treatment effect may have been less than if the treatment regimen had been standardized.

Because it is a voluntary database, the possibility of underreporting adverse outcomes in data submitted to the NCD is a concern. The STS ensures strict confidentiality, however, which removes much of the motivation for event underreporting. Moreover, underreporting would unlikely preferentially affect those who were indicated as having received β-blockers compared with those who were not. Finally, data quality in the NCD has recently been benchmarked: a regional data audit by a Centers for Medicare and Medicaid Services quality improvement organization showed remarkable similarity between voluntary STS data (eg, percentage of missing data, incidence, trend analyses) and audited data, and NCD outcomes have been documented to be remarkably similar to mandatory cardiac database outcomes.^{36 - 38}

We have demonstrated in a large North American observational database analysis that use of preoperative β-blocker therapy in patients undergoing surgical myocardial revascularization was associated with a reduction in adjusted mortality risk, as well as risk of major procedural complications. This benefit was demonstrated among a wide variety of patient subgroups, including those considered at high risk for adverse effects of β-blocker therapy. Although these results are quite promising, we believe that, ideally, they should be confirmed in a large, randomized clinical trial of preoperative β-blocker use. Such a trial could also further optimize therapy and better determine the mechanisms underlying this effect.

Given that CABG is one of the most commonly performed procedures in North America³⁸ and that the utilization rate for this therapy is currently only 60%, this study suggests that preoperative β-blocker therapy has the potential to be a new and useful process measure in CABG quality improvement assessment. This modest measured effect (0.11% absolute risk reduction; Table 4) of β-blockade on 30-day mortality predicts saving approximately 500 lives per year. Additional small reductions in surgical morbidity would be expected to occur as well. The results of this analysis should further raise the awareness of all cardiovascular care providers to the potential benefits of β-blockade in a majority of patients with both medical and surgical cardiovascular disease.