Coronary Bypass Graft Patency in Patients With Diabetes in the Bypass Angioplasty Revascularization Investigation (BARI)
Background— Few studies have compared long-term status of bypass grafts between patients with and without diabetes, and uncertainty exists as to whether diabetes independently predicts poor clinical outcome after CABG.
Methods and Results— Among 1526 patients in BARI who underwent CABG as initial revascularization, 99 of 292 (34%) with treated diabetes mellitus (TDM) (those on insulin or oral hypoglycemic agents) and 469 of 1234 (38%) without TDM had follow-up angiography. Angiograms with the longest interval from initial surgery and before any percutaneous graft intervention (mean 3.9 years) were reviewed. An average of 3.0 grafts were placed at initial CABG for patients with TDM (n=297; internal mammary artery [IMA], 33%) and 2.9 grafts for patients without TDM (n=1347; IMA, 34%). Patients with TDM were more likely than those without to have small (<1.5 mm) grafted distal vessels (29% versus 22%) and vessels of poor quality (9% versus 6%). On follow-up angiography, 89% of IMA grafts were free of stenoses ≥50% among patients with TDM versus 85% among patients without TDM (P=0.23). For vein grafts, the corresponding percentages were 71% versus 75% (P=0.40). After statistical adjustment, TDM was unrelated to having a graft stenosis ≥50% (adjusted odds ratio, 0.87; 95% CI, 0.58 to 1.32).
Conclusions— Despite diabetic patients’ having smaller distal vessels and vessels judged to be of poorer quality, diabetes does not appear to adversely affect patency of IMA or vein grafts over an average of 4-year follow-up. Previously observed differences in survival between CABG-treated patients with and without diabetes may be largely a result of differential risk of mortality from noncardiac causes.
Received July 29, 2002; accepted September 6, 2002.
Although there is considerable information on coronary revascularization options and subsequent clinical course for patients with diabetes mellitus, little has been published on the long-term angiographic status of coronary artery bypass grafts in this important patient group. The National Heart, Lung, and Blood Institute–sponsored Bypass Angioplasty Revascularization Investigation (BARI) is a large North American multicenter study; the primary objective was to compare long-term clinical outcome after an initial strategy of percutaneous transluminal coronary balloon angioplasty (PTCA) with coronary artery bypass graft surgery (CABG) in selected patients with multivessel coronary artery disease. Five and 7 years after randomization, patients with treated diabetes mellitus (TDM) (those on insulin or oral hypoglycemic agents) had significantly higher overall mortality and cardiac mortality compared with patients without TDM, regardless of the initial revascularization strategy received. Among patients with TDM, those randomized to an initial strategy of PTCA had significantly higher 5-year mortality than patients randomized to an initial strategy of CABG.1,2⇓
The survival advantage of CABG over PTCA among patients with TDM suggests sustained graft patency, despite the known propensity for diabetes to accelerate vascular disease. Therefore, the purpose of this study was to compare graft patency rates between patients with and without TDM who had initial CABG. Importantly, some3,4⇓ but not all studies5,6⇓ have suggested that diabetes is an independent predictor of poor clinical outcome after CABG; however, adverse outcomes could not be distinguished from graft failure, native vessel disease progression, or other causes of morbidity. The BARI randomized trial and registry included 3839 patients, of whom 40% initially received CABG. Overall, 18% of these patients had a history of treated diabetes, and more than one third had follow-up angiography, thus providing an opportunity to evaluate graft patency in patients with and without diabetes.
The BARI protocol, including details regarding objectives, patient recruitment, inclusion and exclusion criteria, data collection, procedural guidelines, angiographic definitions, and participating sites, has been published previously.1,7–10⇓⇓⇓⇓ Briefly, patients were eligible for BARI if they were severely symptomatic and had angiographically confirmed multivessel coronary artery disease suitable for initial revascularization by both CABG and PTCA. Angiographic requirements included absence of severe left main coronary artery disease and presence of multiterritory coronary disease along with the expectation that the primary source(s) of ischemia (“culprit lesions”) could be relieved by PTCA or by CABG, the latter typically acknowledged to provide more complete revascularization.
During a 3-year span beginning in August 1988, 1829 patients were enrolled in the randomized trial at 18 participating clinical centers across the United States and Canada. During the same time, 2010 patients who met all the BARI randomization criteria were enrolled in the BARI registry because they or their referring physicians elected not to participate in the randomized trial. The source population for the present analysis includes the 901 patients in the randomized clinical trial and the 625 patients in the BARI registry who received CABG as their initial revascularization. The 568 CABG-treated patients who had ≥1 follow-up angiograms compose the study population.
Follow-up angiography in BARI was performed under 2 circumstances: protocol (scheduled) and clinically indicated (nonprotocol). For protocol-directed angiography, 4 clinical sites at 1 year (Cleveland Clinic, Duke University, St Louis University, and Montreal Heart Institute)11 and at 5 years (Cleveland Clinic, Duke University, Montreal Heart Institute, and Toronto General Hospital) obtained follow-up angiograms on the consecutive series of randomized patients at their site who consented to have follow-up angiography. Of 173 surviving bypass surgery patients at 1-year follow-up sites, 133 (77%) had protocol-directed angiography; of 252 surviving bypass surgery patients at 5-year follow-up sites, 200 (79%) had protocol-directed angiography. Nonprotocol-directed angiograms were obtained at all 18 clinical sites for clinical indications. BARI trained staff entered angiographic and operative graft data into preformatted data acquisition forms. For patients who had >1 follow-up angiogram, the last available angiogram before any incremental revascularization procedure that might alter the status of a bypass graft was used in the analysis. This approach was taken to maximize the length of the follow-up period to identify possible occurrences of late graft deterioration before repeat revascularization. The limited number of repeat bypass surgery procedures typically introduced new grafts without directly altering the status of previously placed grafts. Moreover, only 25 of the 568 study patients (4.4%) had a percutaneous intervention on a graft during follow-up. Results were similar when the first angiogram available during follow-up or the absolute last angiogram (irrespective of repeat revascularizations) was investigated.
Poor distal vessel quality was defined as distal vessel site judged by the surgeon by direct inspection at surgery as severe, diffuse, intimal thickening with significant luminal compromise, or endarterectomy performed; nonpatent graft was defined as a graft with stenosis ≥50% on follow-up angiography.
Differences in demographic, clinical, and initial surgical procedural characteristics between patients with and without diabetes were compared by Student’s t tests for continuous variables and Fisher’s exact test for categorical variables. Distributions of graft stenoses determined from follow-up angiograms were grouped as 0%, 1% to 49%, 50% to 95%, and 96% to 100% and were compared between patients with and without TDM by a χ2 test for trend. To assess the potential independent effect of diabetes on loss of graft patency (stenosis ≥50%), generalized estimating equations (GEEs) were fit using the binomial distribution and logit link function. In this framework, the GEE model parallels the logistic regression model but also takes into account the correlation among multiple observations (grafts) per patient.12 Similarly, for each variable included in the GEE model, a weighted incidence of loss of graft patency during follow-up was calculated with the weight constructed as the inverse of the number of bypass grafts received by each patient. Covariates included in the GEE models included patient age, sex, type of graft conduit, time from initial CABG to follow-up angiography, repeat revascularization before follow-up angiography, distal vessel size, and distal vessel quality. In addition, the type of angiogram evaluated (protocol or nonprotocol) was included in the models (controlled for), because rates of graft patency were generally lower on nonprotocol angiograms, and patients with diabetes had proportionally more nonprotocol angiograms than patients without diabetes. Smoking history and history of hypercholesterolemia were not associated with graft patency and hence were not included in the GEE models.
Characteristics of Source Population
Of the 1526 patients who underwent initial CABG, 292 (19.1%) had TDM (Table 1). The proportion of patients selected from the randomized trial (ie, not the registry) was similar between patients with and without TDM (60% versus 59%, respectively). Patients with TDM were more likely to be female than those without TDM (42% versus 22%) and were, on average, nominally older (63±8 versus 61±10 years). The average number of grafts placed was similar between patients with and without TDM (3.0 versus 2.9), as was the use of an internal mammary artery (IMA) graft (84% versus 84%). At the graft level of analysis, patients with TDM were more likely to have grafts to small (<1.5 mm) distal vessel sites (30% versus 24%) and to have distal vessels judged by surgeons as being of poor quality (11% versus 6%).
Characteristics of Study Population
Of the 1526 patients who underwent initial bypass surgery, 568 (37%) had ≥1 angiogram during follow-up available for analysis (Table 2). This included 99 patients (17.4%) with TDM. Thus, the proportion of patients with diabetes represented in the follow-up angiography study population (17%) was similar to the proportion in the original source population (19%), although in the study population, a higher proportion of all patients was selected from the randomized trial. The mean number of angiograms during follow-up was 1.8 for patients with TDM and 1.7 for patients without TDM; 48 of 99 patients with TDM (48.5%) and 198 of 469 patients without TDM (42.2%) had multiple angiograms during follow-up. On the basis of the last angiogram during follow-up (as previously defined), patients with TDM were more likely to have undergone angiography for clinical (nonprotocol) purposes than patients without TDM (73% versus 62%, P=0.04).
Patients in the study population with TDM, as in the source population, were more likely to be female than those without TDM (51% versus 24%), although, on average, they were of similar age (61±9 years versus 60±10 years) (Table 2). The total number of grafts per patient, the percentage of patients who received an IMA graft, the number of vein grafts per patient, and the percentage of IMA conduits were similar between patients with and without TDM and were generally consistent with results from the source population. Similarly, as with the source population, patients with TDM in the study population were more likely to have smaller distal vessels and vessels judged to be of poor quality.
Follow-Up Angiography by Diabetes Status
The mean time from initial CABG to angiography was 3.9 years (±1.8 years) for patients with TDM and 3.9 years (±1.9 years) for patients without TDM. As seen in Figure 1, a disproportionate amount of the angiographic assessment occurred near the 1- and 5-year follow-up in conjunction with the protocol-directed (scheduled) follow-up angiography at 4 of 18 BARI sites. A total of 11.1% of patients with TDM and 10.4% of patients without TDM had at least 1 repeat revascularization before the last follow-up angiogram used in the analysis. Thus, the length of time to angiography and previous revascularization history were similar between patients with and without TDM.
Follow-Up Assessment of Bypass Grafts by Diabetes Status
On angiographic assessment, the percentage of IMA grafts free of stenoses was similar between patients with and without TDM (89% versus 85%; P=0.23) (Table 3). Overall, IMA grafts were more likely than vein grafts to be stenosis free on the follow-up angiogram. However, the patency rate of vein grafts was also similar between patients with and without TDM (71% versus 75%; P=0.40). Similar results by diabetes status were observed when the 64 patients with angiograms selected within the first 6 months of follow-up (11.3% of total cohort) were excluded from the analysis (Table 3, middle). Among protocol-directed angiograms, 9.9% of grafts among patients with TDM and 11.1% of grafts among patients without TDM had a stenosis ≥50% (P=0.85); among non–protocol-directed angiograms, 16.2% of grafts among patients with TDM and 17.2% of grafts among patients without TDM had a stenosis ≥50% (P=0.76) (data not shown). There was a modest suggestion that both IMA and vein graft patency rates (<50% stenosis) were higher in men than in women, but again, they were similar by diabetes status (Figure 2). Too few patients with TDM had an IMA graft with a significant stenosis (≥50%) to allow for subsidiary analyses. Among vein grafts with stenoses ≥50%, patients with TDM were more likely to have small distal vessels and vessels judged to be of poor quality (Table 3), findings consistent with all grafts placed at the initial CABG (Tables 1 and 2⇑).
Among the 1644 grafts assessed on the last angiogram available during follow-up (as previously defined), independent predictors of a nonpatent graft (≥50% stenosis) included female sex, use of vein graft conduit, angiography that was conducted for nonprotocol (clinical) purposes, and having had an intercurrent repeat revascularization (Table 4). The presence of TDM was unrelated to the likelihood of having a significant graft stenosis (adjusted odds ratio, 0.87; 95% CI, 0.58 to 1.32). As an alternative measure to graft patency, the odds of poor left ventricular function at the time of the last angiographic assessment (defined as ejection fraction <40%) were also not associated with the presence of TDM (odds ratio, 1.23; 95% CI, 0.35 to 4.35; P=0.75) after adjustment for differences in ejection fraction at study entry.
Because angiographic data were not available beyond 5 years of follow-up, cumulative rates of repeat revascularization (PTCA and/or CABG) were calculated as a surrogate measure for lack of graft patency. In the 6-year follow-up period after the last angiogram was obtained, rates of repeat revascularization were virtually identical between patients with and without diabetes (27.3% versus 27.2%, respectively).
In this study, patients with TDM had rates of coronary artery bypass graft disease similar to those of patients without TDM. This was true regardless of the conduit used, regardless of the indication for angiography (protocol-driven or clinically driven) when stratified by sex, and regardless of the fact that at surgery, vessels grafted in patients with TDM were judged, on average, to be smaller and more diffusely diseased. Among bypass grafts with angiographic evidence of disease, there was no indication of greater severity in patients with TDM. In addition, the time to follow-up angiography was similar between patients with and without TDM, although patients with TDM were somewhat more likely to have had angiography for clinical (nonprotocol) purposes. On the basis of the latter results, we might have expected to observe diminished graft patency in patients with TDM, a result not found in this study.
Over the 44 years since the first coronary bypass operation, there have been several angiographic follow-up studies, but this is the first report with the primary objective of comparing graft patency in patients with and without diabetes. The Coronary Artery Surgery Study included an angiographic follow-up substudy,13 but this analysis dealt primarily with the status of the native vessels (grafted and ungrafted) at 5 years after surgery. Graft status was not commented on. More recently, a 1-year follow-up angiographic study comparing IMA and saphenous vein conduits found that diabetes was an independent predictor of graft occlusion.14 A shorter follow-up period and different patient profiles may explain, in part, the lack of agreement between this study and BARI.
Our angiographic findings of comparable graft patency by diabetes status are consistent with the higher survival observed in CABG-treated patients with diabetes than in PTCA-treated patients with diabetes1 but at the same time appear to be inconsistent with clinical follow-up studies that have shown that diabetes is a predictor of a poor clinical outcome after CABG. In patients with multivessel disease and 5-year follow-up (both similar to BARI), Barsness et al3 found that survival in patients with diabetes was significantly reduced at 74% compared with those without diabetes at 86%. In the BARI study itself and in the same population as this angiographic follow-up study (ie, randomized and registry patients) and over approximately the same 5-year follow-up, the survival of patients with TDM was 82% and that of patients without TDM, 93%.15 How can these apparently paradoxical results be explained?
Several vasculature-related explanations can be postulated for the apparent discordance between survival and graft status. These include a higher rate of native vessel disease progression in diabetes,13 the impact of diabetes on endothelial function particularly in the coronary microvasculature,16 diabetes-associated hypercoagulability,17 and an independent myocardial factor.18,19⇓ However, we believe that a more likely explanation is the differential risk of death from noncardiac causes. Indeed, among randomized CABG-treated patients in BARI, 5-year rates of cardiac mortality were 5.8% versus 4.7% in patients with and without diabetes, respectively. In contrast, the corresponding rates of noncardiac mortality were 12.2% versus 4.8%.20 Thus, the comparable rates of graft patency by diabetes status observed in our study are consistent with an attenuation of the diabetes-associated elevated risk of mortality from cardiac causes but essentially no impact on the higher risk of noncardiac mortality associated with diabetes.
This study has several limitations. First, diabetes was not the focus of BARI at its inception, and patients were not randomly assigned by diabetes status. Although angiographic follow-up was partially protocol driven and therefore performed at prescribed intervals after bypass surgery, more than half of the angiograms were obtained for clinical reasons and performed at variable times. Therefore, the data were analyzed separately by type of angiogram and adjusted for in multivariable analysis when analyzed together. Second, the average follow-up period of 3.9 years limits direct detection of adverse effects on graft patency over the longer term. Still, we did not find any evidence of greater need for repeat revascularization among patients with diabetes in the 6-year period after the last angiographic assessment (≈10-year total follow-up), which argues further against differential late graft closure by diabetes status. Third, the BARI population was highly selected, and most importantly, all patients, including those who received CABG, had to be suitable for angioplasty. This may have excluded some patients with diabetes who had extensive diffuse disease unsuited for angioplasty. Fourth, the angiograms used in the analysis were not interpreted centrally, and assessment of the native grafted vessels was performed at surgery and subjectively. Finally, the requirement for survival implicit in undergoing follow-up angiography may have obscured an adverse effect of diabetes on graft patency. However, it is noteworthy that the proportion of patients with diabetes represented in the study population was similar to that of the source population, which argues against a survival bias.
Treated diabetes mellitus does not appear to adversely affect the patency of bypass grafts over an average of 4 years of follow-up, even though vessels grafted in patients with diabetes were judged at surgery to be of smaller caliber and of poorer quality. Long-term rates of repeat revascularization (up to 10 years) are also similar by diabetes status, further suggesting similar long-term patency of bypass grafts. A likely explanation for the poorer long-term survival in CABG-treated patients with diabetes versus those without diabetes, while in the setting of comparable graft patency, is the substantially higher risk of mortality from noncardiac causes in patients with diabetes.
This study was supported by grants HL-38493, HL-38504, HL-38509, HL-38512, HL-38514-6, HL-38518, HL-38524-5, HL-38529, HL-38532, HL-38556, HL-38610, HL-38642, and HL-42145 from the National Heart, Lung, and Blood Institute, Bethesda, Md.
- ↵Barsness GW, Peterson ED, Ohman EM, et al. Relationship between diabetes mellitus and long-term survival after coronary bypass and angioplasty. Circulation. 1997; 96: 2551–2556.
- ↵BARI Investigators. Protocol for the Bypass Angioplasty Revascularization Investigation. Circulation. 1991; 84 (suppl V): V-1–V-27.
- ↵Alderman EL, Stadius ML. The angiographic definitions of the Bypass Angioplasty Revascularization Investigation. Coron Artery Dis. 1992; 3: 1189–1207.
- ↵van der Meer J, Hillege HL, van Gilst WH, et al. A comparison of internal mammary artery and saphenous vein grafts after coronary artery bypass surgery: no difference in 1-year occlusion rates and clinical outcome. Circulation. 1994; 90: 2367–2374.
- ↵Johnstone MT, Creager SJ, Scales KM, et al. Impaired endothelium-dependent vasodilation in patients with insulin-dependent diabetes mellitus. Circulation. 1993; 88: 2510–2516.
- ↵Devereux RB, Roman MJ, Paranicas M, et al. Impact of diabetes on cardiac structure and function: the Strong Heart Study. Circulation. 2000; 101: 2271–2276.
- ↵Bypass Angioplasty Revascularization Investigation (BARI) Investigators. Influence of diabetes on 5-year mortality in a randomized trial comparing CABG and PTCA in patients with multivessel disease: the Bypass Angioplasty Revascularization Investigation. Circulation. 1997; 96: 1761–1769.