Do Angiographic Results From Symptom-Directed Studies Reflect True Graft Patency?

CARDIOVASCULAR Do Angiographic Results From Symptom-Directed Studies Reflect True Graft Patency? Brian F. Buxton, FRACS, Manoj Durairaj, MCh, David L. Hare, FRACP, Ian Gordon, PhD, Simon Moten, FRACS, Victoria Orford, FRACS, and Siven Seevanayagam, FRACS Departments of Cardiac Surgery and Cardiology, Austin Hospital, Melbourne, and Statistical Consulting Centre, University of Melbourne, Parkville, Victoria, Australia Background. Coronary artery graft patency results have been obtained from repeat angiograms in patients who presented with evidence of ischemia. The purpose of this study is to compare protocol-directed angiographic results from a randomized clinical trial with symptomdirected angiography in nontrial patients. Methods. Repeat angiography after primary isolated coronary artery bypass grafting was performed in 337 of 2,259 patients between July 1996 and September 2004. Patients were divided into two groups: 596 graft angiograms in 192 trial patients were compared with 389 graft angiograms in 142 nontrial patients. The mean interval from surgery was 1,306 800 days versus 1,119 777 days, respectively. Grafting techniques were similar in both groups except that the right internal thoracic artery was used almost exclusively as a free graft in the trial patients. Angiographic outcomes were defined as patent (stenosis <80%) or failure (stenosis >80%, occlusion, or the string sign). Comparisons of trial versus nontrial grafts were made using a generalized linear mixed model. Five-year estimates of graft patency were made using survival analyses accounting for interval censoring. Results. The odds ratio for graft failure for nontrial compared with trial patient grafts was 2.6 (95% confidence interval, 1.6 to 4.3; p < 0.001). Cumulative patency estimates for all grafts at 5 years were trial 91% versus nontrial 83%, p 0.004. Five-year estimates for individual conduits were left internal thoracic artery, 99% versus 92%, p 0.002; right internal thoracic artery, 86% versus 87%, p 0.8; radial artery, 87% versus 86%, p 0.6; and saphenous vein, 86% versus 56%, p 0.003. Conclusions. Graft patency rates were superior in the trial compared with nontrial patients. Symptom-directed graft failure rates were approximately double those of trial patients. (Ann Thorac Surg 2005;80:896 901) 2005 by The Society of Thoracic Surgeons The recent paper by Khot and colleagues [1] reported low graft patencies in a symptomatic group of patients, which was inconsistent with other published data [2 5]. A number of factors could contribute to the variability of coronary artery bypass graft patencies. Graft-related factors, such as harvesting, preservation, pharmacologic protection, grafting strategy, and surgical techniques, together with graft disease, all relate to graft patency and surgical outcome. Patientrelated factors are also important. For example, associated comorbidities, such as diabetes, dyslipidemia, hypertension, and renal failure, as well as symptoms of myocardial ischemia, may influence outcome. The reported graft patencies have been derived largely from repeat angiographic assessment of patients who present with signs or symptoms related to ischemia. These studies are nonrandomized, retrospective, and subject to bias. Shah and coworkers [6 8] published Accepted for publication March 23, 2005. Presented at the Forty-first Annual Meeting of The Society of Thoracic Surgeons, Tampa, FL, Jan 24 26, 2005. Address reprint requests to Dr Buxton, Department of Cardiac Surgery, Austin Hospital, Studley Rd, Heidelberg, Victoria 3084, Australia; e-mail: brian.buxton@austin.org.au. the patency of internal thoracic artery, saphenous vein (SV), and radial artery (RA) grafts in a noncontemporaneous symptomatic group of patients. These symptom-based repeat angiographic studies represent only a relatively small cohort of patients who received coronary artery bypass grafts. In these patients, it might be expected that graft failure rate is higher than in the patients who have no evidence of ischemia. Currently, there are a number of randomized trials assessing graft patencies and clinical events after coronary artery bypass graft surgery examining the differences between SV, RA, and internal thoracic artery conduits. All trials include programed angiography, which provides an opportunity to examine graft patency in asymptomatic and symptomatic patients. The Radial Artery Patency Study group has recently reported the 12-month outcome comparing 440 RA and 440 SV grafts from 561 patients [9]. The Complete Arterial Revascularization and Conventional Coronary Artery Surgery Study also reported 12-month patency of 458 grafts in 128 patients [10]. The Radial Artery Patency and Clinical Outcomes study has enrolled 612 patients in a two-tiered comparison of RA versus free right internal thoracic artery (RITA) or SV grafts dur- 2005 by The Society of Thoracic Surgeons 0003-4975/05/$30.00 Published by Elsevier Inc doi:10.1016/j.athoracsur.2005.03.097

Ann Thorac Surg BUXTON ET AL 2005;80:896 901 ANGIOGRAPHIC GRAFT PATENCIES Fig 1. Comparison of trial and nontrial graft patency at 1 year. (LITA left internal thoracic artery; RA radial artery; RITA right internal thoracic artery; SV saphenous vein.) ing 10 years [11]. In the first year, there were 132 graft angiograms in 46 patients. The comparison of the 1-year patency of all grafts from these protocol-directed versus symptom-directed studies [6 11] revealed that the patency of the internal thoracic artery, RA, and SV grafts was greater in the protocol-directed group compared with the symptomdirected groups. The differences between subsets of the RITA and SV graft patencies were significant (Fig 1). The purpose of this paper is to examine differences between symptom-directed angiograms and programed patency data from a randomized controlled trial within a single unit in which surgery was performed by the same group of surgeons between July 1996 and September 2004. Table 1. Patient Demographics Variable Trial Patients (n 192) Nontrial Patients (n 142) Age (y) 64.5 9.4 63.6 11.4 Age 70 y (n, %) 63 (33) 46 (32) Nonelective surgery (n, %) 56 (29) 38 (27) Female (n, %) 21 (11) 42 (30) a Diabetes (n, %) 27 (14) 52 (37) a NIDDM 21 (11) 44 (31) IDDM 6 (3) 8 (6) Hypertension (n, %) 93 (48) 92 (65) b LVEF (n, %) 0.65 47 (24) 33 (23) 0.50 0.65 95 (49) 61 (43) 0.30 0.49 47 (24) 41 (29) 0.30 3 (2) 6 (4) Missing 0 1 (1) No. of grafts (mean SD) 3.1 0.7 2.7 1.0 Years from surgery to 3.6 2.2 3.1 2.1 angiogram (mean SD) a p 0.0001; b p 0.01. NIDDM non insulin-dependent diabetes mellitus; IDDM insulin-dependent diabetes mellitus; LVEF left ventricular ejection fraction; SD standard deviation. Table 2. Graft Characteristics Characteristic Trial % Nontrial % LITA In situ LAD/D single 164 104 In situ Cx/M single 3 9 In situ Sequential graft 35 31 Free LIMA 5 5 Total 207 35 149 38 RITA In situ LAD/D single 1 4 In situ Cx/M single 4 3 In situ RC single 0 5 Free RIMA 52 18 Total 57 10 30 8 SV LAD/D single 33 8 Cx/M single 89 16 RC single 85 32 Sequential graft 21 13 Total 228 38 69 18 RA LAD/D single 2 7 Cx/M single 71 42 RC single 29 33 Sequential graft 2 59 Total 104 17 141 36 Total 596 389 Cx/M intermediate or marginal; LAD/D left anterior descending or diagonal; LIMA left internal mammary artery; LITA left internal thoracic artery; RA radial artery; RC right coronary, posterior descending or posterolateral; RIMA right internal mammary artery; RITA right internal thoracic artery; SV saphenous vein. Patients and Methods 897 Protocol-directed angiographic results from the randomized controlled Radial Artery Patency and Clinical Outcomes trial were compared with those from symptomdirected studies at the University of Melbourne Austin Hospital. The radial artery graft study was approved by the human resources and ethics committee at the Austin Hospital (H95/086). The Radial Artery Patency and Clinical Outcomes trial compares the RA patency and clinical outcomes in two groups of patients in whom the RA (experimental) or the control graft was anastomosed to the largest non left anterior descending target artery. In 397 patients ( 70 years) the RA was compared with the free RITA, and in 219 patients ( 70 years) the RA was compared with the SV. Each patient was scheduled for one angiogram during the 10-year follow-up. Time to angiography was obtained by a second randomization to 1-, 2-, 5-, 7-, and 10-year intervals, with most repeat angiograms scheduled between 5 and 10 years [11]. The Austin Hospital database contained 1,654 patients who had a primary isolated coronary artery bypass between 1996 and 2004 and who were not included in the randomized controlled trial. CARDIOVASCULAR

CARDIOVASCULAR 898 BUXTON ET AL Ann Thorac Surg ANGIOGRAPHIC GRAFT PATENCIES 2005;80:896 901 Table 3. Generalized Linear Mixed Model Variable Comparison Odds Ratio 95% Confidence Interval p Value Group Symptom versus protocol 2.6 1.6 4.3 0.001 Conduit SV compared with LITA 6.0 3.2 11.1 0.001 RA compared with LITA 2.3 1.2 4.5 0.01 RITA compared with LITA 3.7 1.5 8.9 0.004 RA compared with SV 0.4 0.2 0.6 0.001 RITA compared with SV 0.6 0.3 1.2 0.17 RITA compared with RA 1.6 0.8 3.4 0.2 Target artery Circumflex compared with LAD/Diag 1.4 0.8 2.4 0.3 RCA compared with LAD/Diag 1.1 0.6 1.9 0.9 RCA compared with circumflex 0.8 0.5 1.1 0.18 Age Per 10 years 1.3 1.0 1.7 0.04 LVEF 0.50 0.65% compared with 0.65% 0.8 0.4 1.4 0.4 0.30 0.49% compared with 0.65% 0.9 0.4 1.7 0.7 0.30% compared with 0.65% 1.7 0.4 6.8 0.5 LAD/Diag left anterior descending or diagonal; LITA left internal thoracic artery; LVEF left ventricular ejection fraction; RA radial artery; RCA right coronary artery; RITA free right internal thoracic artery; SV saphenous vein. Patients Repeat angiography was performed in 337 of 2,259 patients who underwent a primary isolated coronary artery bypass graft operation between July 1996 and September 2004. To date, 192 of the 605 (32%) patients enrolled in the randomized controlled trial received a protocol-directed angiogram. These were compared with 142 of 1,654 (8.6%) nontrial patients who had repeat angiography during the same time for symptoms of ischemia. The patient demographics are shown in Table 1. Grafts and Anastomoses A total of 985 graft angiograms were studied. There were 596 grafts in the trial patients and 389 in the nontrial patients. The conduits, anastomotic sites, and type of bypass grafts are shown in Table 2. Grafts were assessed by conventional selective coronary angiographic techniques. If there was more than one graft or anastomosis from a single conduit, each segment was assessed separately. Graft failure was defined as stenosis of 80%, total occlusion, or those with a string sign. in the symptom-directed angiography group. Age, acuity of presentation, and ventricular function were similar. The interval from surgery to angiography was 1,306 800 days in the trial patients versus 1,119 777 days in the nontrial patients (mean standard deviation; Table 1). Grafts and Anastomoses The absolute numbers of grafts and grafts per patient were greater in the trial versus nontrial patients: n 596 (3.1 0.07) versus n 389 (2.7 1.0; Tables 1 and 2). The proportion of left internal thoracic artery and RITA grafts was similar in the protocol-directed versus symptomdirected angiograms. The RITA was used almost exclu- Statistical Analysis Longitudinal assessment of factors affecting graft patency was assessed using the generalized linear mixed model, which incorporated two levels of variables those related to the patient and those related to grafts. Estimates of 5-year graft patencies were obtained using survival analyses with interval censoring techniques described by Turnbull [12] and Peto [13]. Results Demography There were significant differences between groups in many of the main patient and graft variables. Female sex, diabetes, and hypertension were significantly more likely Fig 2. Comparison of cumulative patency of trial versus nontrial grafts using the interval censoring technique of Turnbull [12] and Peto [13]. The solid line represents patients from the protocol-directed group; the dashed line represents patients from the symptom-directed group.

Ann Thorac Surg BUXTON ET AL 2005;80:896 901 ANGIOGRAPHIC GRAFT PATENCIES sively as a free graft in the trial patients. The SV graft usage was nearly double and the RITA approximately half in the trial versus nontrial patients, which reflected the randomization process and the preference to use RAs in preference to SVs in nontrial patients (Table 2). Generalized Linear Mixed Model Comparison of the overall graft failure rate of nontrial versus trial patients showed an odds ratio of 2.6 (95% confidence interval, 1.6 to 4.3; p 0.001; Table 3). These rates were adjusted for conduit, age, ventricular function, and target artery. Five-Year Estimates Cumulative patency estimates for all grafts at 5 years were 91% for trial versus 83% for nontrial (p 0.004; Fig 2). Five-year estimates for individual conduits were left internal thoracic artery, 99% versus 92% (p 0.002); RITA, 86% versus 87% (p 0.8); RA, 87% versus 86% (p 0.6); and SV, 86% versus 56% (p 0.003). Comment The angiographic graft patency results from patients who present with symptoms or signs suggestive of ischemia demonstrate approximately double the failure rate compared with those in a randomized clinical trial. Although the results of this study might appear intuitive, we attempted to quantify the differing observations in the same cohort of patients. Most published patency rates have been derived from repeat angiographic studies of symptomatic patients, which may have resulted in underestimating graft patency. It is likely, therefore, that the graft patency rates from protocol-directed patency rates, such as those from a randomized clinical trial, in which most patients are asymptomatic, are more relevant to clinical practice. As a consequence, the randomized controlled trial patency rates would be more appropriate as the clinical standard. Symptom-directed angiograms are often event-driven and the data collected retrospectively. The quantification of graft disease and details of the lesions in the native coronary arteries were not recorded routinely, and these studies were not randomized. Standard statistical models assessing patency (eg, the Cox model), when used to analyze the two levels of predictive variables, patient or graft, are inappropriate. These models assume that all observations are independent, which is not so in the case of patient-related variables, which are common to all grafts in a single patient. Furthermore, the Kaplan-Meier estimates suffer from the problem that time to graft failure is unknown and are therefore invalid. Analyses of patency from such nonstandardized data are therefore not reliable. Clinical events are likely to be related to extension of coronary artery or bypass graft disease, and therefore, clinically derived patency rates will be low. Alternatively, some asymptomatic graft failures may not be recorded, providing a relative increase, albeit small, in patency rates. The relative proportion of patients presenting with 899 symptoms is low, and therefore most series of symptomdirected angiograms represent only a small portion of the patients who have undergone coronary artery bypass graft surgery. The results of surgery in trial patients have been recognized as being superior to those not in the trial [14]. In trials, the surgical techniques are often standardized and the patient follow-up regimented. The trial patients in this study are reviewed at yearly intervals for assessment of symptoms and management of lipids, hypertension, and diabetes, which may have contributed to the improved outcome of coronary artery bypass grafts. In the current trial, angiography is scheduled at predetermined intervals. This trial is in progress; the mean follow-up is 4.5 years for this 10-year study, and only about one third of patients have had the protocoldirected angiogram because the number of patients scheduled for angiography was weighted to the latter years of the study when most events are expected. Complete angiographic follow-up will not be available until the study is concluded. Selection bias of patients is likely to be a contributing factor to graft failure. The randomized controlled trial included a wide range of patients. There were exclusions of patients with left ventricular ejection fraction less than 0.30, recent myocardial infarction, body mass index greater than 35, poor lung function, and renal failure, which might bias the trial results favorably. The nontrial group had a higher proportion of women and patients with diabetes and hypertension; this could impact graft patency negatively. Although female sex, diabetes, and hypertension were significantly more likely in the symptom-directed angiography group, there were not enough graft failure end points to be able to validly control for these differences between the two groups. Although there were differences with some major patient variables, other characteristics such as left ventricular function and acuity of presentation were similar. Longitudinal follow-up is essential. The results of the first 12 months reflect a number of factors, including surgical techniques, perioperative complications, and the properties of graft and native vessels, as well as patient variables [1 11]. The mean follow-up of the current trial and nontrial patients, who had repeat angiograms, is approximately 3 years, which is a short duration relative to the expected long-term patency, especially of arterial conduits. Further follow-up will be needed to determine the true biologic differences among the conduits [15]. Limitations A single angiogram provides restricted information about graft patency. Selective angiography is invasive, and computed tomographic angiography requires exposure to similar doses of radiation; therefore, neither of these studies lends itself to repeated observations in the same patient. Although repeated observations may be useful in obtaining a more complete picture of graft function, it is difficult to justify the increased risk to an asymptomatic patient in a clinical trial. For these reasons, CARDIOVASCULAR

CARDIOVASCULAR 900 BUXTON ET AL Ann Thorac Surg ANGIOGRAPHIC GRAFT PATENCIES 2005;80:896 901 trial patients were scheduled for only one postoperative angiogram. The precise relationship of trial data to the broader population of patients having coronary artery bypass grafting is unknown. The patency results of programed angiography demonstrate what can be obtained in a population defined by the trial entry criteria [11]. Conclusions Coronary artery bypass graft failure rate in patients with ischemia was about double that of patients in the clinical trial. Angiography, after evidence of ischemia, underestimates graft patency. Implication Angiogram results from trials, compared with symptomatic patients, is more representative of those from the overall coronary artery surgical population; the trial results are more relevant and should be considered as the standard by which grafts are compared in the future. References 1. Khot UN, Friedman DT, Pettersson G, Smedira NG, Li J, Ellis SG. Radial artery bypass grafts have an increased occurrence of angiographically severe stenosis and occlusion compared with left internal mammary arteries and saphenous vein grafts. Circulation 2004;109:2086 91. 2. Possati G, Gaudino M, Prati F, et al. Long-term results of the radial artery used for myocardial revascularization. Circulation 2003;108:1350 4. 3. Muneretto C, Negri A, Manfredi J, et al. Safety and usefulness of composite grafts for total arterial myocardial revascularization: a prospective randomised evaluation. J Thorac Cardiovasc Surg 2003;125:826 35. 4. Calafiore AM, Di Mauro M, D Alessandro S, et al. Revascularization of the lateral wall: long term angiographic and clinical results of radial artery versus right internal thoracic artery grafting. J Thorac Cardiovasc Surg 2002;123:225 31. 5. Caputo M, Reeves B, Marchetto G, Mahesh B, Lim K, Angelini GD. Radial versus right internal thoracic artery as a second arterial conduit for coronary surgery: early and midterm outcomes. J Thorac Cardiovasc Surg 2003;126:39 47. 6. Shah P, Gordon I, Fuller J, et al. Factors affecting saphenous vein graft patency: clinical and angiographic study in 1,402 symptomatic patients operated between 1977 and 1999. J Thorac Cardiovasc Surg 2003;126:1972 7. 7. Shah PJ, Durairaj M, Gordon I, et al. Factors affecting patency of internal thoracic artery graft: clinical and angiographic study in 1,434 symptomatic patients operated between 1982 2002. Eur J Cardiothorac Surg 2004;26:118 24. 8. Shah P, Seevanayagam S, Rosalion A, et al. Patency of the radial artery graft: angiographic study on 209 symptomatic patients operated between 1995 and 2002 and review of the current literature. Heart Lung Circ 2004;13:379 83. 9. Desai ND, Cohen EA, Naylor CD, Fremes SE. Randomized comparison of radial-artery and saphenous-vein coronary bypass grafts. N Engl J Med 2004;351:2302 9. 10. Simon P, Devyatko E, Rohn V, et al. Early coronary graft patency rate (Complete Arterial Revascularisation and Conventional Coronary Artery Surgery Study CARACCASS Data). Abstract 108, 2nd EACTS/ESTS Joint Meeting, Vienna, Austria, Oct 12 15, 2003. 11. Buxton BF, Raman JS, Ruengsakulrach P, et al. Radial artery patency and clinical outcomes 5-year interim results of a randomized trial. J Thorac Cardiovasc Surg 2003;125:1363 70. 12. Turnbull BW. The empirical distribution function with arbitrarily grouped, censored and truncated data. J R Stat Soc Series B 1976;38:290 5. 13. Peto R. Experimental survival curves for interval-censored data. Appl Statist 1973;22:86 91. 14. Roethlisberger FJ, Dickson WJ. Management and the worker. Cambridge: Harvard University Press, 1939. 15. Lytle BW. Prolonging patency choosing coronary bypass grafts. N Engl J Med 2004;351:2262 4. DISCUSSION DR SOON J. PARK (San Francisco, CA): I would like to congratulate Dr. Buxton and his colleagues for their excellent presentation today and other significant contributions over the years on the patency rates of coronary bypass grafts. This study compared the graft patency of 142 patients who had angiographies based on symptoms of ischemia with 192 patients who underwent angiographies on a predetermined schedule per protocol in Radial Artery Patency and Clinical Outcomes (RAPCO) trial. Graft patency was significantly higher in the trial patients. Since the majority of overall coronary artery bypass grafting patients are asymptomatic, Dr. Buxton implied that the true graft patency of the overall cohort may be more like that of trial patients rather than that of symptomatic patients. By definition, patients in the symptomatic group represented a very huge subset with a specific condition of ischemia. None of the asymptomatic patients in the rest of cohort was represented. Similarly, the RAPCO trial group consisted of very specific patients based on rather strict inclusion/exclusion criteria such as ejection fraction greater than 35%, a minimal target vessel diameter greater than 1.5 mm, or expected survival greater than 10 years. Thus, the patients in RAPCO trial were very different, yet equally skewed, and they were far from being the representative sample of the entire cohort. The sampling criteria were very different between the two groups. As expected, fundamental differences were present that could have accounted for different incidences of ischemic symptoms and graft patency. For example, the symptomatic group had a significantly higher incidence of female gender, 30% versus 11%, diabetes, 37% versus 14%, and hypertension, 65% versus 48%. Also, the number of bypass grafts was lower, 2.7 versus 3.1. A substantial portion of symptomatic patients was likely to have had small target coronary vessels given the higher incidence of female gender and diabetes. Patients with target vessel diameter less than 1.5 mm were included in the symptomatic group unlike RAPCO trial group. Therefore, the observed difference in graft patency may have been a mere reflection of two very different sampling processes. Since the samples were not equivalent, any type of statistical comparison was invalid. Furthermore, one may not draw any conclusion about the overall cohort based on these nonrepresentative samples. Ischemic symptoms may be a poor surrogate marker of graft patency. Graft failure, which may be silent... and rates of failure will therefore be underestimated wrote Dr. Buxton in his previous publication.

Ann Thorac Surg BUXTON ET AL 2005;80:896 901 ANGIOGRAPHIC GRAFT PATENCIES It is very tempting to expand our knowledge base by inferring and generalizing from a smaller observation. Of course, we are all eager to know the true patency rates of coronary bypass grafts. Unfortunately, we cannot draw such a conclusion from this study. I would like to thank the chairman and the Society for the opportunity to discuss and share my thoughts on this paper. DR BUXTON: I would like to thank Dr. Park for his criticisms and comments. The prime criticism is that we really don t have the results from the wider population of patients having coronary artery bypass grafts. All we can do is gradually get closer. We believe that perhaps the trial patients with all the exclusions and there are not that many better represent the overall population. However, I accept that criticism. 901 Secondly, there is no apology for the differences between the patient groups, that is the way it is. It is not surprising that there are more adverse risk factors in the symptomatic group of patients than in the trial patients. The number of vessels in each study was slightly skewed because the trial did not include patients with any single vessel grafts. I do agree with the fundamental thrust of Dr. Park s comments, that it really is very difficult to obtain the truth, unless we have an angiogram every day or every month in the whole population, it is not possible to obtain true graft patencies. The trial patients do represent another cohort of patients which perhaps gives us wider representation and provide graft patencies which are closer to the truth. CARDIOVASCULAR Notice From the American Board of Thoracic Surgery Regarding Trainees and Candidates for Certification Who Are Called to Military Service Related to the War on Terrorism The Board appreciates the concern of those who have received emergency calls to military service. They may be assured that the Board will exercise the same sympathetic consideration as was given to candidates in recognition of their special contributions to their country during the Vietnam conflict and the Persian Gulf conflict with regard to applications, examinations, and interruption of training. If you have any questions about how this might affect you, please call the Board office at (312) 202-5900. Timothy J. Gardner, MD Chairman The American Board of Thoracic Surgery 2005 by The Society of Thoracic Surgeons Ann Thorac Surg 2005;80:901 0003-4975/05/$30.00 Published by Elsevier Inc