Reoperation for Bleeding in Patients Undergoing Coronary Artery Bypass Surgery
Incidence, Risk Factors, Time Trends, and Outcomes
Background— Reoperation for bleeding represents an important complication in patients undergoing coronary artery bypass surgery (CABG). Yet, few studies have characterized risk factors and patient outcomes of this event.
Methods and Results— We evaluated 528 686 CABG patients at >800 hospitals in the Society of Thoracic Surgeons National Cardiac Database (2004 to 2007). Clinical features and in-hospital outcomes were evaluated in patients with and without reoperation for bleeding after CABG. Logistic regression was used to identify predictors of risk of this event and to estimate weights for an additive risk score. A total of 12 652 CABG patients (2.4%) required reoperation for bleeding. These rates remained fairly stable over time (2.2%, 2.3%, 2.5%, and 2.4% from 2004 to 2007, respectively). Although overall operative mortality was 4.5-fold higher in patients requiring reoperation for bleeding versus those who did not (2.0% versus 9.1%), this mortality risk declined significantly over time (11.3%, 9.5%, 8.8%, and 8.2% from 2004 to 2007, respectively, P for trend=0.0006). Factors associated with higher risk for reoperation were identified by multivariable analysis (c statistic=0.60) and summarized into a simple bedside risk score. The risk-score performed well when tested in the validation set (Hosmer-Lemeshow P=0.16).
Conclusions— Reoperation for bleeding remains an important morbid event after CABG. Nonetheless, death in patients with this complication has decreased over time. Our risk tool should allow estimation of patients risk for reoperation for bleeding and promote preventive measures when feasible in this at-risk group.
Received February 16, 2009; accepted July 29, 2009.
Reoperation for bleeding is required in 2.3% to 8% patients undergoing coronary artery bypass graft (CABG) surgery.1–9 Prior studies have characterized the morbidity and mortality and risk factors associated with the need for reoperation for bleeding after CABG.1–4,7–9 However, this information tends to be derived single tertiary center reports1–3,5,7–9 or from selected regional consortium.4,6 Additionally, these studies also preceded the use of more potent antithrombotic therapies such as clopidogrel and glycoprotein IIb/IIIa inhibitors that have been shown to increase reoperation for bleeding in patients undergoing CABG.10–12 Thus, the incidence, risk factors, and outcomes of patients requiring reoperation after CABG in contemporary community practice at large in recent era are not known.
Using the data from the Society of Thoracic Surgeons National Cardiac Surgery Database (STS NCD),13–15 we sought to (1) characterize the clinical features and in-hospital outcomes of patients undergoing CABG requiring reoperation for bleeding; (2) identify trends in reoperations and mortality in patients with this event; (3) explore the underlying plausible mechanisms for these trends; (4) identify preoperative clinical variables associated with the risk of this event; and (5) create a bedside tool to estimate the risk of this event.
The formation, rationale, and methodology of the STS NCD have been previously published.13–15 Briefly, this registry was initiated in 1986 with a goal to provide participants with their risk-adjusted outcomes compared to the national experience. This information is used by member institutions as a part of their continuous quality improvement efforts to help design strategies to improve their outcomes.
We analyzed data on patients aged ≥20 and ≤99 years undergoing isolated CABG from January 1, 2004, through December 31, 2007, at participating sites. Patients undergoing isolated or concomitant valve, congenital heart, or aortic surgery were excluded from the analysis. We also excluded patients if they had missing information on few key variables (ie, age, gender, status of surgery, surgery date, and hospital number).
The STS NCD collects information on reoperation for bleeding as repeat surgery needed for bleeding or cardiac tamponade either before hospital discharge or between hospital discharge and within 30 days of their index operation if discharged earlier than 30 days. Information on all other adverse events was ascertained as in-hospital events. Detailed definitions of risk factors and adverse events were as provided on the STS web site (www.sts.org).
Summary statistics are presented as frequency and percentage, or as median and interquartile range. χ2 test was used for comparing categorical variables and Wilcoxon rank sum test for continuous variables, both tests were stratum adjusted by participant ID. The Cochran-Mantel-Haenszel test was used for the analysis of trends.
Time Trends in Bleeding Risk
Time trends in risk-adjusted bleeding rates were assessed using a nonparsimonious multivariable logistic regression model with adjustment for 26 patient preoperative characteristics. Variables for this model were selected based on the previously published STS mortality model as well as clinical experience of factors and treatments known to be associated with increased risk of this event.14 These variables included age; gender; race; body surface area; hypertension; diabetes; immunosuppressive therapy; peripheral vascular disease; chronic lung disease; cerebrovascular disease; atrial fibrillation; baseline myocardial infarction, unstable angina, creatinine, heart failure, shock, dialysis, and left ventricular ejection fraction; left main disease; number of coronary artery disease; number of prior cardiac operations; percutaneous coronary interventions <6 hours of CABG; intraaortic balloon pump; status of surgery; year of operation and use of oral antiplatelet agents and intravenous glycoprotein IIb/IIIa receptor antagonists <24 hours of surgery. To account for clustering of outcomes within hospitals, the model was fitted using generalized estimating equation methodology with an exchangeable working correlation structure and empirical sandwich standard error estimates.
Mortality rates for patients needing reoperation were adjusted to account for risk factors present at the time of the initial operation. A nonparsimonious logistic regression model was developed for this purpose that included all factors listed above for the reoperation model with the exception of year of surgery and the use of oral antiplatelet agents and intravenous glycoprotein IIb/IIIa receptor antagonists <24 hours of surgery. We treated operative mortality as the dependent (outcome) variable. Indirectly standardized mortality rates were then calculated as the observed number of deaths divided by the model-predicted number of deaths multiplied by the overall mortality rate in the study sample.
Development of Predictive Models
We also developed a parsimonious logistic regression model that can be used to predict an individual patient’s risk of reoperation for bleeding using a small number of predictor variables. Parsimony was felt to be desirable to reduce the data collection burden for users of the model and to facilitate ease-of-use at the patient’s bedside. We used 80% of the sample for the development of the model and the remaining 20% of the sample for model validation. To enhance usability, continuous variables were assumed to have a linear relationship (age [in 5-year increments], body mass index) or categorized (serum creatinine, left ventricular ejection fraction).
Nonsignificant variables were eliminated using a backwards selection algorithm with a P=0.05 stopping rule.
In addition, we created a bedside tool which allows a crude estimate of risk of reoperation for bleeding to be calculated using simple addition based on a minimal number of explanatory variables. To choose variables for the bedside tool, we identified the smallest subset of predictors that were able to capture at least 90% of variation in the predicted log-odds of reoperation as estimated by the logistic predictive model described above. Regression coefficients were then converted to whole integers by multiplying the coefficients by a factor of 10 and rounding to the nearest integer to create a bedside risk prediction tool. For each possible risk score, the risk of reoperation for bleeding was estimated by averaging the predicted values from the logistic predictive model among all patients having the same risk score. Validation of risk score tool was performed by comparing observed mortality in the validation set versus expected mortality among risk score categories. The heterogeneity of the risk of this event among patients having the same risk score was assessed by plotting the distribution of predicted risk based on the full model within risk score subgroups. Calibration of model was assessed by Hosmer-Lemeshow test statistic.
Records with missing values of gender (0.01%), hospital number (0.002%), and operative status (0.07%) were excluded from all data presented here. Among the variables that were used for multivariable analysis, missing data were highest for ejection fraction (4.5%). Information for the remaining variables in the multivariable analyses was missing in <1%. Missing values of body surface area (0.49%) were imputed to gender-specific median values. Ejection fraction was imputed to the conditional median after stratifying by gender, status, and heart failure. Missing values of categorical risk factors were defaulted to their most common value which in most cases was “not present.” All analyses were performed using SAS 9.1 (SAS Institute) software.
Baseline Clinical Characteristics of Patients Undergoing CABG Needing Reoperation for Bleeding
Repeat surgery for bleeding was required in 12 652 (2.39%) of 528 686 patients undergoing CABG. Patients needing reoperation for bleeding were significantly older, males, nonwhites with lower body mass index with higher comorbid conditions (Table 1). Similarly, preoperative features portending adverse prognosis (ie, myocardial infarction ≤24 hours of surgery, congestive heart failure, cardiogenic shock, lower creatinine clearance) were also significantly higher in this cohort. The use of potent antiplatelet agent (ie, thienopyridine or glycoprotein IIb/IIIa agent) within 24 hours before the operation was significantly higher in patients needing reoperation for bleeding. More patients who underwent emergent or emergent salvage surgery required reoperation for bleeding. Cardiopulmonary bypass use and the median pump and cross-clamp times were marginally higher among patients requiring reoperation for bleeding, whereas internal mammary artery bypass conduit use was less frequent among these patients. There was modest direct relationship between perfusion time and the rates of reoperation for bleeding (r=0.69). Reoperation for bleeding was required in 2.07% of patients undergoing off-pump CABG and in 2.47% of patients undergoing on-pump CABG (adjusted rates 2.06% versus 2.46%, P<0.0001). In patients who did receive clopidogrel before surgery, reoperation for bleeding was lower in the off-pump CABG patients compared with those undergoing on-pump CABG (3.68% versus 3.23%, P<0.001).
Reoperation for bleeding after CABG was associated with worse outcomes compared with those without this event (Table 2). Those requiring reoperation were at higher risk for permanent stroke (2.2-fold), pneumonia (2.9-fold), septicemia (4.3-fold), and the need for prolonged ventilation (≥48 hours; 4.6-fold). Operative mortality was >4.5-fold in patients needing reoperation for bleeding. Even after adjusting for baseline confounders, mortality remained higher among patients needing reoperation for bleeding compared with those not requiring this (adjusted rates, 5.92% versus 1.97%, P<0.0001).
Time Trends for Reoperation for Bleeding and Outcomes in Patients Undergoing CABG From 2004 to 2007
There were some minor shifts in the risk profiles of patients undergoing CABG overtime. For example, there was a clinically insignificant small increase in preprocedural thienopyridines use (9.1%, 8.9%, 9.7%, and 9.2% for years 2004 to 2007, respectively, P for trend 0.0018) and a decline in glycoprotein IIb/IIIa use over the study period (4.4%, 4.5%, 4.0%, and 3.9% for years 2004 to 2007 respectively, P for trend <0.0001). The rates of emergent salvage or emergent operations; proportion of patients undergoing CABG <24 hours of a myocardial infarction and rates of off-pump CABG remained stable over the study period. Overall, the risk profile of patients receiving CABG remained stable over the 4-year study period (predicted mortality risk 2.1%, 2.2%, 2.1%, and 2.1%, for 2004 to 2007; respectively).
Reoperation for bleeding was also fairly constant over this period (2.21%, 2.34%, 2.49%, and 2.42% for years 2004 through 2007, respectively). This clinically insignificant trend for increasing need for reoperation for bleeding over time persisted even when adjusted for baseline confounders (odds ratio, 1.04; 95% CI, 0.96 to 1.12 for 2005; odds ratio, 1.09; 95% CI, 1.00 to 1.18 for 2006; odds ratio, 1.07; 95% CI, 0.97 to 1.17 for 2004 [referent]).
Mortality over the study period improved in overall patients undergoing CABG, but particularly among those needing reoperation for bleeding (Figure 1).
Risk Prediction Model and Bedside Risk Tool
Factors independently associated with reoperation for bleeding in patients undergoing CABG are shown in Table 3 (c index, 0.60). The Hosmer-Lemeshow probability value of the full model was 0.19 providing little evidence of poor fit and suggesting high concordance between predicted and observed values. The simplified model coefficients were converted to an additive risk score (Figure 2). Figure 3 displays a comparison between predicted risk of reoperation for bleeding based on bedside risk tool and observed rates in the risk score categories in the validation set. The risk score tool performed well in the validation set with good concordance between predicted values in the risk categories and the observed risk in this sample (Hosmer-Lemeshow P=0.16). Figure 4 displays the variability between the estimate of risk based on the full model and the estimated risk determined by the bedside risk tool in various risk score categories in the derivation data-set. The average absolute difference between the risks calculated by full model compared with estimates based on the risk scores was 0.30%. Overall there was good correlation between the estimation of risk with risk score and that estimated based on the full model (correlation coefficient 0.95).
Findings of Current Investigation
Our study provides a national perspective on the need for reoperation for bleeding after CABG in recent contemporary practice when ubiquitous use of potent antithrombotic agents is a norm. Our data found that the incidence of this event is low (2.4%) and remained fairly constant overtime. Nonetheless, reoperation for bleeding remained strongly associated with greater morbidity and mortality even after accounting for baseline confounding factors. Although reoperation for bleeding was needed in <1 of every 40 patients undergoing CABG, this event was associated with 1 of every 10 deaths in these patients (10.3% of all CABG deaths). Most impressively, there was a significant decline in mortality over time in patients with this complication.
Our data provide some insight into the potential mechanisms for the stable risk of reoperation for bleeding in these patients. Unlike prior studies from the STS NCD15 and the Northern New England Cardiovascular Disease Study Group,16 we did not find an increasing risk of patients undergoing CABG in recent times. Similarly, factors shown to be associated with an increased risk of reoperation for bleeding4,5,8,9 (ie, the rates of emergent or emergent salvage procedures), CABG <24 hours of myocardial infarction and on-pump (versus off-pump) CABG remained stable over the study period. The rates of thienopyridine use before CABG also remained unchanged over this time. Thus, the overall stable risk of patients undergoing CABG in general along with stable rates of predictive factors for increased risk of reoperation described above may explain in part the stable incidence of reoperation for bleeding in patients undergoing CABG.
In contrast, we can only speculate the reasons for the decline in mortality associated with reoperation for bleeding. This decrease may have been partially related to the growing awareness of the potential detrimental bleeding related side effects of not only aspirin, but also the newer antiplatelet and antithrombotic agents,10–12 avoidance of CABG when feasible in stable patients until after the effect of these agents wear off, the increasing experience of surgeons in the management of these patients, meticulous hemostasis during surgery, the use of appropriate blood products along with ε amino caproic acid, aprotinin, or desmopressin in the perioperative period,17 and use of recently collected (rather than old) blood products18 when needed. Additionally, the quality improvement effort as part of the ongoing participation in STS NCD may also have contributed to this decline in mortality over time.
Finally, we also developed a parsimonious predictive model to identify patients with increased risk of reoperation for bleeding after CABG. This may be potentially useful to physicians in estimating patient-specific risk for reoperation for bleeding before undertaking the operation, for counseling during informed consent process as well as to promote preventive measures when feasible in this at-risk group. More importantly, this outcome varied across different centers that have different case-mix. Thus, the current risk tool specific to reoperation for bleeding would allow risk adjusted comparisons of different sites. This would facilitate targeted quality improvement initiatives at sites with high rates of reoperation.
Unlike prior studies that have demonstrated increased risk for reoperation with preoperative aspirin in patients undergoing CABG,19 we found no such risk with the use of aspirin <24 hours before CABG. These findings are consistent with that of a recent review which showed no increased perioperative risk of bleeding requiring reoperation with aspirin among observational studies and increased risk only with doses ≥325 mg per day among patients in randomized trials.20
We point out that although based on large number of patients undergoing CABG with more than 12 000 events (a much larger number than any of the previously published studies), the ability of our model was only fair in discriminating between patients needing versus those not needing reoperation for bleeding, highlighting the difficulty and challenge in accurately identifying such patients preoperatively. Many factors such as surgeon’s experience or the threshold for taking back patient to the operating room, the rapidity of bleeding, patients underlying hemodynamics, other patient comorbidities, the response to conservative medical treatments including transfusion and antifibrinolytic agents, etc that were not collected in the STS NCD may have partially accounted for this lack of good discriminatory ability. Despite this drawback, our risk score performed well in the independent validation set and provided reasonable estimate of the risk of this event.
Comparisons With Prior Studies
Many previous studies have provided insights into the need for reoperation for bleeding and have reported an incidence that varied from 2.3% to 8.0%.1–12 Our estimate of this risk is on the lower side (2.4%) of that described in these studies and perhaps reflects a stable estimate of the risk in community at large. Nonetheless, our data concurs with these previous studies in that reoperation for bleeding increases morbidity and mortality in patients undergoing CABG and remains an important target for improving outcomes.
Important differences between these previous studies and our investigation exist. Most prior studies involved patients at a single center 1–3,5,7,12 or those from a regional consortium involving few institutions,4,6 thus with few patients and fewer events. In contrast, our data involve current contemporary practice at more than 800 widely varying institutions with a very large number of patients and events allowing robust statistical power, thereby increasing the generalizability of our findings. More importantly, almost all of these studies predated the more routine use of potent antiplatelet and antithrombotic agents,1–5,8,9 an important and independent predictor of this event in this and other studies10–12 limiting their applicability in current era. Finally, few of these prior investigations focused on risk factors associated with increased risk of reoperation4,6,8 or provided a simple risk tool that could be used in routine practice for estimating risk of this adverse event in patients undergoing CABG.
Many newer antithrombotic and antiplatelet agents have completed phase III clinical trials or are on the verge of doing so and many more are in the early clinical phase of human studies.21–23 As these agents become engrained in contemporary cardiovascular practice, as accurate identification on admission of patients requiring reoperation after CABG remain a major ongoing challenge,24 and because the effectiveness of many therapies is reduced if not given early after admission,25 exposure to these potent agents before CABG (an already increasingly older and sicker population) is likely to increase. This is expected to increase the need for reoperation after CABG. Although the easiest option to minimize the risk is to avoid exposure to these agents before CABG or to delay surgery in those exposed until platelet function or hemostasis have recovered, this approach may not be pragmatic many times and may even have the potential for harm as well as increasing cost (of hospitalization) while awaiting CABG. Thus, if outcomes of CABG patients are to be improved in the future, continued ongoing efforts from surgical colleagues would be necessary to reduce the risk of perioperative bleeding. Our study is unable to provide insights into strategies that may be helpful in minimizing the risk of this adverse event in CABG patients. It is likely that understanding of the detail pharmocology and the duration of increased risk conferred by the exposure of these new agents, meticulous hemostasis in those exposed to these drugs, vigilance for bleeding, and aggressive management early on with appropriate blood products and drugs may have the potential for reducing the need for reoperation for bleeding after CABG. Whether shorter acting intravenous agents21,22 or the use of novel RNA-aptamer-drug-antidote combination23 or the use of off-pump CABG would allow efficacy upfront when needed and minimize postoperative bleeding after CABG remains to be established. Finally, the importance of quality improvement initiatives at individual institutions, regional, or national level to measure this event across various sites and compare them after accounting for patient case-mix using the current risk tool, identify the root cause of the problem at sites with high rates of this event, and then implement process changes to minimize it cannot be overemphasized to help reduce the risk of this event and improve outcomes of patients undergoing CABG.
Although, the STS NCD now represents a majority of centers in United States, there is a slight overrepresentation of participation from larger centers and those with better overall outcomes. Second, we are only able to analyze risk factors collected and cannot account for the influence of any unmeasured factors influencing the risk of reoperation for bleeding in CABG patients. We only had information on the use of thienopyrines within 24 hours of CABG, but not before this time. However, this would only have biased our observation toward null (ie, no increased risk of bleeding with clopidogrel). The current database only allowed for evaluation of short-term risk, as we did not have any information on long-term outcomes.
Reoperation for bleeding is an important morbid event after CABG that has remained stable over time. Fortunately, death in patients with this adverse event decreased over time at least among patients undergoing CABG at institutions participating in STS NCD. The future availibility of more potent antiplatelet agents is expected to increase the risk of reoperation for bleeding in CABG patients further highlighting the ongoing need for monitoring and quality improvement efforts at institutional, regional, or national level through initiatives like the STS NCD to reduce this complication in the hope of improving ouctomes after CABG. Our risk tool may allow estimation of patients risk for reoperation for bleeding and promote preventive measures when feasible in this at-risk group.
Sources of Funding
This work was funded by the STS NCD.
Sellman M, Intonti MA, Ivert T. Reoperations for bleeding after coronary artery bypass procedures during 25 years. Eur J Cardiothorac Surg. 1997; 11: 521–527.
Dacey LJ, Munoz JJ, Baribeau YR, Johnson ER, Lahey SJ, Leavitt BJ, Quinn RD, Nugent WC, Birkmeyer JD, O'Connor GT; for the Northern New England Cardiovascular Disease Study Group. Reoperation for hemorrhage following coronary artery bypass grafting-Incidence and risk factors. Arch Surg. 1998; 133: 442–447.
Munoz JJ, Birkemeyer NJO, Dacey LJ, Birkmeyer JD, Charlesworth DC, Johnson ER, Lahey SJ, Norotsky M, Quinn RD, Westbrook BM, O'Connor GT, for the Northern New England Cardiovascular Disease Study Group. Trends in rates of reexploration for hemorrhage after coronary artery bypass surgery. Ann Thorac Surg. 1999; 68: 1321–1325.
Choong CK, Gerrard C, Goldsmith KA, Dunningham H, Vuylsteke A. Delayed re-exploration for bleeding after coronary artery bypass surgery results in adverse outcomes. Eur J CardioThorac Surg. 2007; 31: 834–838.
Fox KA, Mehta SR, Peters R, Zhao F, Lakkis N, Gersh BJ, Yusuf S; Clopidogrel in Unstable angina to prevent Recurrent ischemic Events Trial. Benefits and risks of the combination of clopidogrel in patients undergoing surgical revascularization for Non-ST elevation acute coronary syndromes; The Clopidogrel in unstable angina to prevent recurrent ischemic events (CURE) trial. Circulation. 2004; 110: 1202–1208.
Kapetanakis EI, Medlam DA, Petro KR, Haile E, Hill PC, Dullum MK, Bafi AS, Boyce SW, Corso PJ. Effect of clopidogrel premedication in off-pump cardiac surgery: are we forfeiting the benefits of reduced hemorrhagic sequelae? Circulation. 2006; 113: 1667–1674.
Kapetanakis EI, Mediam DA, Boyce SW, Haile E, Hill PC, Dullum MKC, Bafi AS, Petro KR, Corso PJ. Clopidogrel administration prior to coronary artery bypass grafting surgery: the cardiologist’s panacea or the surgeon’s headache? Eur Heart J. 2005; 26: 576–583.
Ferguson TB, Hamill PG, Peterson ED, DeLong ER, Glover FL. A decade of change—risk profiles and outcomes for isolated coronary artery bypass grafting procedures, 1990–1999: a report from the STS National Database Committee and the Duke Clinical Research Institute. Ann Thorac Surg. 2002; 73: 480–490.
Sethi GK, Copeland JG, Goldman S, Moritz T, Zadina K, Henderson WG. Implications of preoperative administration of aspirin in patients undergoing coronary artery bypass grafting. Department of Veterans Affairs Cooperative Study on Antiplatelet Therapy. J Am Coll Cardiol. 1990; 15: 15–20.
Sun JC, Whitlock R, Cheng J, Eikelboom JW, Thabane L, Crowther MA, Teoh KH. The effect of pre-operative aspirin on bleeding, transfusion, myocardial infarction, and mortality in coronary artery bypass surgery: a systematic review of randomized and observational studies. Eur Heart J. 2008; 29: 1057–1571.
Chan MY, Cohen MG, Dyke CK, Myles SK, Aberle LG, Lin M, Walder J, Steinhubl SR, Gilchrist IC, Kleiman NS, Vorchheimer DA, Chronos N, Melloni C, Alexander JH, Harrington RA, Tonkens RM, Becker RC, Rusconi CP. Phase 1b randomized study of antidote-controlled modulation of factor IXa activity in patients with stable coronary artery disease. Circulation. 2008; 117: 2865–2874.
Mehta RH, Chen AY, Pollack CV, Roe MT, Zalenski R, Delgado G, Gibler WB, Ohman EM, Harrington RA, Peterson ED. Challenges in predicting the need for coronary artery bypass grafting at presentation in patients with non-ST elevation acute coronary syndromes. Am J Cardiol. 2006; 98: 624–627.
Steinhubl SR, Berger PB, Mann JT III, Fry ET, DeLago A, Wilmer C, Topol EJ; CREDO Investigators. Clopidogrel for the Reduction of Events During Observation. Early and sustained dual oral antiplatelet therapy following percutaneous coronary intervention: a randomized controlled trial. JAMA. 2002; 288: 2411–2420.