Author + information
- Received June 25, 2002
- Revision received August 6, 2002
- Accepted August 26, 2002
- Published online December 4, 2002.
- Jason H Cole, MD*,
- Ellis L Jones, MD, FACC†,
- Joseph M Craver, MD, FACC†,
- Robert A Guyton, MD, FACC†,
- Douglas C Morris, MD, FACC*,
- John S Douglas Jr, MD, FACC*,
- Ziyad Ghazzal, MD, FACC* and
- William S Weintraub, MD, FACC*,* ()
- ↵*Reprint requests and correspondence:
Dr. William S. Weintraub, Emory Center for Outcomes Research, 1256 Briarcliff Road, Suite 1-North, Atlanta, Georgia, USA 30306.
Objectives This study evaluated both short- and long-term outcomes of diabetic patients who underwent repeat coronary artery bypass graft surgery (CABG) or percutaneous coronary intervention (PCI) after initial CABG.
Background Although diabetic patients who have multivessel coronary disease and require initial revascularization may benefit from CABG as compared with PCI, the uncertainty concerning the choice of revascularization may be greater for diabetic patients who have had previous CABG.
Methods Data were obtained over 15 years for diabetic patients undergoing PCI procedures or repeat CABG after previous coronary surgery. Baseline characteristics were compared between groups, and in-hospital, 5-year, and 10-year mortality rates were calculated. Multivariate correlates of in-hospital and long-term mortality were determined.
Results Both PCI (n = 1,123) and CABG (n = 598) patients were similar in age, gender, years of diabetes, and insulin dependence, but they varied in presence of hypertension, prior myocardial infarction, angina severity, heart failure, ejection fraction, and left main disease. In-hospital mortality was greater for CABG, but differences in long-term mortality were not significant (10 year mortality, 68% PCI vs. 74% CABG, p = 0.14). Multivariate correlates of long-term mortality were older age, hypertension, low ejection fraction, and an interaction between heart failure and choice of PCI. The PCI itself did not correlate with mortality.
Conclusions The increased initial risk of redo CABG in diabetic patients and the comparable high long-term mortality regardless of type of intervention suggest that, except for patients with severe heart failure, PCI be strongly considered in all patients for whom there is a percutaneous alternative.
As the care of patients undergoing coronary artery bypass graft surgery (CABG) improves, patients expect to survive for many years after their surgery. This longevity, however, increases the likelihood of repeat coronary intervention secondary to the development of atherosclerotic disease in these patients’ bypass grafts or progression of disease in their native vessels. This problem is magnified in importance in diabetic patients, who are more likely to experience long-term problems, including failure of their bypass grafts, after CABG (1).
Patients who present in need of revascularization after previous bypass surgery have two available options—a repeat coronary surgery or percutaneous coronary intervention (PCI). In many cases, each of these options is viable, and it is not clear which is preferable for a given patient. In a meta-analysis of the large trials conducted to evaluate initial CABG versus PCI, no significant mortality benefit was observed for one procedure compared to the other (2). Instead, the CABG patients required fewer repeat procedures at the expense of greater initial cost. However, in the Bypass Angioplasty Revascularization Investigation (BARI) trial, the 19% of patients who were diabetic had 81% five-year survival rates when treated with CABG but only 66% five-year survival when treated with balloon angioplasty (3,4). Similar results were also apparent in recent data from the Coronary Angioplasty versus Bypass Revascularization Investigation (CABRI) study, which showed higher mortality for diabetic patients, but not nondiabetic patients, when treated with PCI as opposed to CABG (5). It is notable, however, that whereas this effect has been shown for the diabetic subset in these randomized trials, the patient populations only included first-time CABG patients with multivessel disease.
In another large trial, at eight-year follow-up, there is a trend—although not statistically significant—toward decreased long-term mortality in diabetic patients treated with CABG as opposed to PCI (6). Furthermore, the negative interaction between diabetes and choice of PCI as opposed to CABG has been demonstrated, although inconsistently, in observational studies from large databases (1,7)and appears in analysis of the BARI registry (8). A question that remains beyond these studies, however, is whether previous bypass surgery affects the relative risk profile for diabetic patients undergoing revascularization. Can a second CABG provide the same marginal benefit to diabetic patients?
With the advent of coronary stenting, technical success rates and long-term survival have improved in all patients undergoing PCI after prior CABG (9,10). Furthermore, evidence exists that among all patients with previous CABG, there is no overall mortality difference between patients treated with different revascularization methods (11). Thus, it may be appropriate to look at clinical and angiographic criteria specific to a given patient in an effort to choose PCI or CABG. In this setting, the present study evaluates both the in-hospital and long-term outcomes of diabetic patients who underwent PCI or redo CABG after initial CABG.
Patient population, procedures, and variables
The study population was drawn from all patients presenting to Emory University hospitals from 1985 to 1999. Patients were included if they had a diagnosis of diabetes and previously had undergone CABG. Diabetic patients were defined as those who were receiving active treatment for diabetes mellitus with either insulin or an oral hypoglycemic agent at the time of intervention. Diet-controlled diabetic patients were only included in the study if documentation showed a fasting blood glucose level of >140 mg/dl or a random blood glucose of >200 mg/dl available during hospitalization. Patients underwent revascularization procedures (either PCI or CABG) in the setting of stable or unstable angina or after several days of stabilization following acute myocardial infarction (MI). Procedures performed acutely for MI were not included in the analysis. All angioplasty procedures were performed by standard techniques that have been previously described (12). Procedures performed included balloon angioplasty, coronary stent implantation, directional atherectomy, and evaluation of other new procedures. Patients who had reoperative CABG had surgery performed with standard surgical techniques (13).
Variables defined by patient history included age, gender, number of years of diagnosed diabetes, duration of time since first CABG, insulin dependence, hypertension, prior MI, heart failure, severity of angina, ejection fraction, and number of vessels diseased. Angina was defined according to the Canadian Cardiovascular Society Classification (14), and congestive heart failure (CHF) by New York Heart Association criteria (15). Single-vessel disease was defined as ≥50% diameter luminal narrowing of the left anterior descending artery (LAD), left circumflex artery (LCX), right coronary artery, or a major branch of these arteries. Two-vessel disease required the presence of at least 50% narrowing in two of the three major epicardial vessel systems, and three-vessel disease involved all three arteries (or LAD and proximal LCX disease in left-dominant systems).
Defined complications were the development of a new Q-wave MI postprocedure or a stroke, defined as a neurologic event with persisting changes in neurologic function after the procedure.
All baseline data as well as procedural results were recorded prospectively on standardized forms and entered into a computerized database. Determination of angiographic success was based on quantitative angiographic calculation performed by angiographers involved in the procedure, but other than the primary operator. Follow-up information was obtained directly from patients, from their referring physicians, by telephone or letter follow-up, and from data assessed at each subsequent hospital admission. Key elements of information obtained in follow-up related to patient survival, cause of death (cardiac or noncardiac), recurrent hospitalization, the occurrence of MI, and any need for repeat revascularization.
Continuous data were expressed as mean ± SD and categorical data as proportions. Continuous data were compared by unpaired ttest and categorical data by chi-square test. Missing data were filled in according to the method of Harrell (16). All variables were considered potential covariates of in-hospital and long-term mortality. Correlates of in-hospital outcome were assessed by logistic regression, whereas correlates of long-term outcome were analyzed by the Cox proportional hazards survival model (17). Multivariate correlates of in-hospital and long-term mortality were expressed as odds ratio (OR) and hazard ratios (HRs), respectively. Potential nonlinear effects of each of the continuous predictor variables were checked using restricted cubic splines. Interaction terms were assessed. The discrimination of the models was assessed by the c (or concordance) index, which is the fraction of pairs of patients, one with a given end point, correctly identified. The models were validated and calibrated according to the method of Harrell (16), and overall survival (cardiac plus noncardiac) was determined and expressed by the Kaplan-Meier method (18).
Baseline characteristics of the 1,721 diabetic patients who underwent repeat revascularization are included in Table 1. Follow-up was obtained for 98% of the patients. Of note, the 1,123 patients who underwent PCI and the 598 patients who had redo CABG were similar in age, gender, years of diabetes, and insulin dependence. However, hypertension and prior MI were significantly more prevalent in the subset undergoing CABG. A majority of all patients had class III to IV angina, but this finding was noted more frequently in patients who underwent PCI. Heart failure, while noted in a minority of patients, was more common among those undergoing CABG (24.3% vs. 15.6%). It is also notable that the vast majority of patients in both groups had multivessel disease, with more of the PCI patients having two- and three-vessel disease and a higher proportion of CABG patients with left main disease. Ejection fraction was relatively well preserved within both groups, but higher in the PCI patients.
Outcomes are detailed in Table 2. Reflecting the time during which these patients were treated, only 25.02% of the PCI patients underwent coronary stenting. Also, 43.19% of the percutaneous procedures involved a saphenous vein graft (SVG) site. Notable regarding the CABG patients is that 48.5% of these patients had an internal mammary artery (IMA) graft placed in the setting of redo CABG. In-hospital mortality was significantly higher with redo CABG, and Q-wave MI was more frequent in the CABG group. Stroke was a significantly more frequent complication (4.68% vs. 0.09%, p < 0.0001) in patients undergoing CABG. Both postprocedure hospital stay and cost of initial hospitalization were significantly higher in the CABG group. Figure 1demonstrates long-term survival for the two groups of patients. Despite the higher in-hospital mortality associated with redo CABG in these patients, the long-term differences are much smaller. Mortality differences were not significant between the two groups at either 5 years (38% PCI vs. 39% CABG) or 10 years (68% PCI vs. 74% CABG, p = 0.14). An additional revascularization with CABG (Fig. 2) was more common in patients initially treated with PCI (p < 0.0001).
Multivariate analysis was performed to evaluate patient survival in each group in light of underlying differences between the patient populations. Multivariate correlates of in-hospital mortality are included in Table 3, and correlates of long-term mortality are in Table 4. The strongest correlate of in-hospital mortality was coronary surgery, with an OR of 6.51 (95% confidence interval [CI] 3.75 to 11.28) (Table 3). However, year of the revascularization procedure (OR 0.92 per 1-year increase, 95% CI 0.86 to 0.97), age (OR 1.51 per 10-year increase in age, 95% CI 1.13 to 2.50), patient gender (male gender OR 0.42, 95% CI 0.25 to 0.68), and lower ejection fraction (OR 0.71 per 10% increase in ejection fraction, 95% CI 0.59 to 0.86), also correlated with in-hospital mortality. The variables that correlated with long-term mortality (Table 4) also included age (HR 1.56 per 10 years, 95% CI 1.42 to 1.71), gender (HR for males 0.83, 95% CI 0.69 to 1.00), and ejection fraction (per 10% increase in ejection fraction, HR 0.77, 95% CI 0.72 to 0.83), along with hypertension (HR 1.30, 95% CI 1.09 to 1.56).
Class III to IV CHF was a significant correlate of long-term mortality for patients managed with PCI (HR 1.54, 95% CI 1.25 to 1.89) but not for patients treated with CABG (HR 1.19, 95% CI 0.97 to 1.45) (Table 4). Thus, the only significant interaction in the long-term mortality model was between type of revascularization and CHF. Figure 3demonstrates this strong trend, consistent with multivariate analysis, for poorer survival in CHF patients treated with PCI, which is not noted in the patients without CHF. Of note, CHF was a strong predictor of mortality after either revascularization strategy (p = 0.03 for CABG and p < 0.0001 for PCI). A final analysis demonstrated that, for patients undergoing CABG, those treated with an internal mammary graft were at lower long-term risk (HR 0.77, 95% CI 0.59 to 1.00, p = 0.048). Choice of PCI itself was not a long-term correlate of mortality, either overall (Table 4) or if PCI were compared to the CABG subgroup receiving an IMA graft (HR for CABG compared to PCI 0.88, 95% CI 0.69 to 1.14, p = 0.37).
This study analyzes short- and long-term mortality for diabetic patients who underwent repeat coronary intervention following bypass surgery. A first observation is that these patients, with long-term mortality of approximately 70% at 10 years (no matter which method of revascularization was chosen), are some of the sickest ischemic heart disease patients encountered in cardiology practice. Additionally, reflecting operative and early postoperative mortality, repeat revascularization with CABG is associated with significant early mortality excess. The variables that correlate with long-term mortality—most notably older age, hypertension, and depressed ejection fraction—relate to the underlying disease process in these patients. Multivariate analysis does not reveal choice of revascularization to correlate with long-term mortality. The only significant interaction was between patients with advanced (class III to IV) heart failure and the choice of PCI.
This trend toward worse outcomes with PCI and CHF also manifests in the long-term survival curves for CHF patients, raising the possibility that these particularly sick patients may not do as well with PCI. However, except for this small subset of patients with severe CHF, the data on long-term mortality provide no support for one method of revascularization over the other. In fact, given the demonstration of a profound initial risk for redo CABG in these diabetic patients (over 11% in-hospital mortality in our cohort), this study suggests consideration of PCI for those patients in whom there is a percutaneous alternative.
There are limitations to the generalizability of these data. Medical therapy, PCI, and CABG have changed significantly over the years of this analysis. Even the definition of diabetes has changed, and the currently accepted definition of a fasting glucose value >126 mg/dl, adopted by the American Diabetes Association in 1997 and the World Health Organization in 1999 (19), considers as diabetic some patients not included in this analysis. Medical therapy, including the aggressive use of statin therapy and angiotensin-converting enzyme inhibitors, has also evolved since the mid-1980s. Furthermore, newer interventional techniques, most notably stenting, may allow diabetic patients to have better outcomes than previously available with percutaneous intervention (20,21). Reflecting the changing practice patterns over the course of 15 years, only 25% of the patients undergoing PCI in this study had the placement of intracoronary stents, and the use of stents is even more important in a population of patients with SVGs, given the significant benefit that has been demonstrated for stenting SVGs (9). At the same time, cardiac surgical techniques continue to evolve, including the more common use of off-pump CABG with acceptable mortality (22). In combination with the natural evolution of cardiac anesthesia (23)and the use of multiple arterial grafts (24,25), evidence shows that, over the course of time, mortality outcomes with CABG have been improving. The result of these improvements in both PCI and CABG is that physicians may often have more confidence sending their patients for either of these interventions.
However, these changes in medical and surgical outcomes bias decision making for many diabetic patients toward PCI. For example, the value of intracoronary stents in a diabetic population has seemed to increase greatly with the addition of the glycoprotein IIb/IIIa inhibitors (20), which would have been used in even fewer than the 25% of patients receiving stents in our analysis. Similarly, the use of vascular brachytherapy to treat in-stent restenosis has proven to be quite effective in the diabetic population, and new strategies for distal protection make SVG intervention safer in both diabetic and nondiabetic patients. It is hoped that greater numbers of these diabetic patients will ultimately be able to benefit from percutaneous intervention after initial coronary surgery, with the potential caveats that patients who can receive an arterial conduit or patients with severe CHF may be more likely to obtain a relative benefit from another coronary surgery.
Other potential limitations of these data relate to their derivation from a large database. However, excellent follow-up and the comprehensive nature of the database give credibility to the results. Although there may be limitations to the letter and telephone follow-up techniques required to maintain the database, end points of mortality or repeat revascularization are readily assessed by these means. Nonetheless, the analysis still is dependent on observational, nonrandomized data. It remains possible that selection bias may have led to certain less sick patients undergoing PCI (with surgeons only being referred patients in whom there were no discrete lesions easily amenable to percutaneous therapy), but it is also likely that there was a group of patients selected for PCI who were too sick for CABG. Although multivariate analysis did not find choice of revascularization to be correlated with long-term survival in these patients, multivariate analysis cannot account for unmeasured confounders affecting selection. Undoubtedly, certain patients in both the PCI and redo CABG groups could not have effectively been treated with the other method of revascularization, and despite careful analysis of patient characteristics, there is no way to fully address this issue in a retrospective, nonrandomized analysis.
However, it is worthwhile remembering that the evidence favoring first-time CABG in diabetic patients is derived from a subset analysis of randomized trials and nonrandomized database studies, with only inconsistent results demonstrated in the latter (described in the previous text). There has not been a randomized trial comparing revascularization strategies in patients with previous CABG. Although the continuing evolution of revascularization techniques has prompted new trials to compare PCI and CABG (26,27), these studies were also limited to patients who have not had prior CABG. Even subset analysis from these trials more reflective of stenting and other current techniques is not as helpful in evaluating choices for the high-risk patients considered in our study. Without randomized data or plans for a randomized trial in the near future, the value of results from a large database such as the one analyzed here becomes that much greater.
Thus, neither the constant evolution of medical and surgical care nor the observational nature of these data counters the underlying message. First, compared with percutaneous intervention, there is significant early hazard for repeat CABG in diabetic patients. Second, these patients have significant long-term mortality, and it is the severity of disease and, perhaps, the profile of multiple risk factors that best explain this mortality—not a selection of PCI or CABG. The poor long-term survival of diabetic patients following revascularization emphasizes the importance of maximizing medical therapy no matter which revascularization modality is chosen. This strategy will include optimization of preventive measures to retard the progression of atherosclerosis and careful surveillance of these patients for disease progression.
As far as the fundamental question of whether to proceed with PCI or redo CABG in diabetic patients after first coronary surgery, the higher initial mortality of redo CABG and the similar long-term results seen with either revascularization method argue that for some patients it is appropriate to have an initial bias toward PCI, with redo CABG selected based on individual patient characteristics, such as the ability to provide new arterial revascularization conduits or the presence of unprotected native vessel left main disease. It may also be reasonable to have a bias toward CABG in class III to IV CHF patients. Nonetheless, interventional cardiologists and surgeons will often make revascularization decisions for these patients based on angiographic evaluation specific to each patient, and these results provide strong support for such decision making. Evidence that CABG offers better outcome than PCI for a first revascularization procedure in diabetic patients with multivessel disease cannot be generalized to diabetic patients with previous CABG. Rather, choices must be based on clinical and angiographic evaluation of each patient, along with patient preference.
- Bypass Angioplasty Revascularization Investigation
- coronary artery bypass graft surgery
- Coronary Angioplasty versus Bypass Revascularization Investigation
- congestive heart failure
- confidence interval
- hazard ratio
- internal mammary artery
- left anterior descending artery
- left circumflex artery
- myocardial infarction
- odds ratio
- percutaneous coronary intervention
- saphenous vein graft
- Received June 25, 2002.
- Revision received August 6, 2002.
- Accepted August 26, 2002.
- American College of Cardiology Foundation
- Barsness G.W.,
- Peterson E.D.,
- Ohman E.M.,
- et al.
- Ellis S.G.,
- Narins C.R.
- EAST Investigators,
- King S.B.,
- Kosinski A.S.,
- Guyton R.A.,
- Lembo N.J.,
- Weintraub W.S.
- Weintraub W.S.,
- Stein B.,
- Kosinski A.,
- et al.
- Detre K.M.,
- Guo P.,
- Holubkov R.,
- et al.
- Wong S.C.,
- Baim D.S.,
- Schatz R.A.,
- et al.
- Weintraub W.S.,
- Jones E.L.,
- Morris D.C.,
- King S.B.,
- Guyton R.A.,
- Craver J.M.
- Douglas J.S.,
- King S.B.
- Goldman L.,
- Hashimoto B.,
- Cook E.F.,
- Loscalzo A.
- ↵Harrell FE. Design: S functions for biostatistical/epidemiological modeling, testing, estimation, validation, graphics, prediction, and typesetting. Programs available from: email@example.com. Accessed April 2001
- Cox D.R.
- Gabir M.M.,
- Hanson R.L.,
- Dabelea D.,
- et al.
- Marso S.P.,
- Lincoff A.M.,
- Ellis S.G.,
- et al.
- Abizaid A.,
- Kornowski R.,
- Mintz G.S.,
- et al.
- Tu J.V.,
- We K.
- Endo M.,
- Nishida H.,
- Tomizawa Y.,
- Kasanuki H.
- Lytle B.W.,
- Loop F.D.
- Serruys P.W.,
- Unger F.,
- Sousa J.E.,
- et al.
- Stable R.H.