Comparison of radial artery and saphenous vein as the third conduit with bilateral internal mammary arteries in coronary bypass surgery Authors: Jasleen Nagi, Chirag Goyal, Nandor Marczin, Sunil K. Bhudia, Shahzad G. Raja, Cardiac Outcomes Research Group and Initiative (CORGI) Published: 2026 ABSTRACT ObjectiveThe optimal third conduit for multiarterial coronary artery bypass grafting using bilateral internal mammary arteries (BIMA) remains debated. Trial data have been confounded by crossover and frequent radial artery (RA) use in controls. We evaluated whether adding RA versus saphenous vein (SVG) to BIMA influences outcomes.MethodsWe performed a 27-year retrospective analysis of 1727 patients with BIMA grafts receiving SVG (n = 1307) or RA (n = 420). Propensity-score matching (1:1) generated 399 balanced pairs. Preoperative characteristics, intraoperative data, and in-hospital outcomes were compared. Long-term survival was assessed by Kaplan-Meier analysis, log-rank testing, and Cox proportional hazards modeling.ResultsBefore matching, recipients of RAs had fewer comorbidities (diabetes 14.8% vs 26.0%; hypertension 62.4% vs 77.7%; renal impairment 5.5% vs 25.2%; all P < .001). In-hospital outcomes were largely comparable; however, postoperative renal complications more frequent in matched recipients of RAs (2.5% vs 0.5%; P = .042). In the unmatched cohort, 20-year survival favored BIMA+RA (log-rank χ2 = 5, P = .026). After matching, this difference was attenuated and not statistically significant (χ2 = 2.7, P = .097). Cox modeling identified RA as independently protective (hazard ratio, 0.522; 95% CI, 0.279-0.977; P = .0419); increasing age, preoperative atrial fibrillation, and ≥2 previous myocardial infarctions predicted mortality.ConclusionsRA as the third conduit alongside BIMA was associated with improved long-term survival in unmatched analyses and remained protective in multivariable modeling, but the matched cohort did not show a statistically significant survival difference, suggesting selection effects. The observed increase in postoperative renal complications warrants caution. Conduit choice should be individualized. Similar in-hospital outcomes for BIMA+SVG and BIMA+RA, except renal complications. Central MessageRadial artery use as the third conduit may offer long-term benefit, but this advantage weakens after adjusting for selection. Conduit choice should be tailored to balance durability and risk. PerspectiveIn assessing third-conduit choice for BIMA CABG, this study finds similar long-term survival between radial and saphenous vein grafts after matching, with a potential renal risk linked to radial grafting. The results support tailored, patient-specific conduit strategies aligned with modern arterial-first guidelines. Similar in-hospital outcomes for BIMA+SVG and BIMA+RA, except renal complications. Central Message Radial artery use as the third conduit may offer long-term benefit, but this advantage weakens after adjusting for selection. Conduit choice should be tailored to balance durability and risk. Perspective In assessing third-conduit choice for BIMA CABG, this study finds similar long-term survival between radial and saphenous vein grafts after matching, with a potential renal risk linked to radial grafting. The results support tailored, patient-specific conduit strategies aligned with modern arterial-first guidelines. Coronary artery bypass grafting (CABG) remains a cornerstone therapy for multivessel coronary artery disease. Contemporary European guidance supports the preferential use of arterial conduits—including bilateral internal mammary arteries (BIMAs) and the radial artery (RA)—for durability.1 Furthermore, total arterial revascularization (TAR) is supported by additional contemporary evidence in appropriately selected patients, particularly younger individuals with longer life expectancy.2 Long-term data demonstrate excellent patency for arterial grafts, including RA, whereas saphenous vein grafts (SVGs) are known to fail progressively over time.2, 3, 4, 5 Beyond conduit type, competitive flow and target-vessel stenosis strongly influence arterial graft performance; patency is superior when native stenosis is severe (typically ≥80%-90%).6,7 Randomized evidence at the patient level indicates that RA, compared with SVG, is associated with lower adverse cardiac events, superior angiographic patency, and fewer repeat revascularizations at ∼5 years, although a survival advantage has not been consistently demonstrated.8 Meta-analyses aggregating propensity-matched cohorts suggest that the addition of a third arterial graft (often RA) yields late survival benefit without increasing operative risk.9, 10, 11 Yet, among studies specifically evaluating BIMA with RA versus BIMA with SVG, findings are mixed—some show no early survival difference, with signals of benefit beyond 10 years.12, 13, 14, 15, 16 The optimal third conduit in multiarterial CABG employing BIMA remains unresolved in routine practice. We therefore examined outcomes of adding RA versus SVG to BIMA over 27 years at a single high-volume center, using propensity-score matching to mitigate selection bias. Methods Data Source This retrospective cohort study used prospectively collected data from the institutional cardiac surgery registry, which captures comprehensive preoperative, intraoperative, and postoperative variables for all cardiac surgical procedures. The registry undergoes routine internal validation and contributes to national audit datasets. Mortality status was verified using hospital records and national death-tracking systems, ensuring complete follow-up for all included patients. Ethical approval was granted by the institutional review board (QS SR 04/2024), with waiver of individual consent in accordance with national guidance for retrospective analyses of deidentified data. The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request, subject to institutional governance and privacy regulations. Study Cohort We identified all adult patients undergoing BIMA from a total cohort of 13,626 isolated CABG procedures performed between January 1996 and September 2023 at a single tertiary cardiac surgical center. Patients were included if they received BIMA in combination with either a SVG or a RA as the third conduit, with additional SVGs permitted in the BIMA+RA group. Individuals undergoing concomitant cardiac procedures were excluded. Clinical decision-making regarding conduit strategy was undertaken by the operating surgeon within a multidisciplinary framework that incorporated coronary anatomy, comorbidity profile, and conduit suitability. Operative Strategy All procedures were performed via median sternotomy. Use of cardiopulmonary bypass was at surgeon discretion and recorded for all cases. In the BIMA+SVG group, the left and right internal mammary arteries were supplemented with SVGs to achieve complete revascularization. In the BIMA+RA group, the RA was harvested using a standardized technique and used as the third conduit (Appendix E1), typically targeting the circumflex or right coronary territories. After harvest, the radial artery was flushed with an antispasm solution of glyceryl trinitrate and verapamil added to the patient's blood. Conduit selection reflected anatomical considerations, target vessel stenosis severity, and the surgeon’s preference. Completeness of revascularization was assessed using the index of completeness of revascularization, defined as the number of grafts performed divided by the number of graftable vessels with angiographically significant stenoses. All procedures were performed by 7 consultant surgeons during the study period. Perioperative calcium-channel antagonist use was not standardized and was inconsistently recorded across eras; therefore, it could not be reliably quantified or compared between groups. End Points The primary end point was long-term all-cause mortality. Secondary end points were in-hospital outcomes including 30-day mortality defined in the Appendix E1. Statistical Analysis Categorical variables were compared using the χ2 or Fisher exact tests, and continuous variables were compared using the Wilcoxon rank-sum test or Student t test as appropriate. Unmatched survival was analyzed with Kaplan-Meier and adjusted Cox models; matched survival used a Cox model stratified by matched pair. Year of surgery was included as a continuous covariate in the multivariable Cox model to account for potential era effects. To reduce treatment-selection bias, we constructed a 1:1 nearest-neighbor propensity score match with a caliper width of 0.20. Preoperative covariates included age, sex, diabetes, hypertension, hypercholesterolemia, renal impairment, pulmonary disease, neurologic disease, extracardiac arteriopathy, atrial fibrillation, extent of coronary artery disease, left main stem disease, left ventricular ejection fraction, smoking status, Canadian Cardiovascular Society class, New York Heart Association class, previous myocardial infarction, and previous cardiac surgery. Matching yielded 399 well-balanced pairs (n = 798). Covariate balance was assessed using standardized mean differences. Statistical analyses were performed using R, version 4.5.0 (R Foundation for Statistical Computing). Results Baseline Characteristics In the unmatched cohort (n = 1727), recipients of RAs had fewer comorbidities than recipients of SVGs—diabetes 14.8% versus 26.0%, hypertension 62.4% versus 77.7%, renal impairment 5.5% versus 25.2%, and pulmonary disease 4.5% versus 9.0% (all P ≤ .004), respectively. Recipients of RAs also had lower New York Heart Association and Canadian Cardiovascular Society classes and less left main disease (16.9% vs 26.6%, P < .001). After matching (n = 798; 399 pairs), covariates were well balanced; residual imbalances were small (Table 1).Table 1Preoperative demographicsVariableUnmatched (n = 1727)Matched (n = 798)BIMA+SVG(n = 1307)BIMA+RA(n = 420)SMDP valueBIMA+SVG(n = 399)BIMA+RA(n = 399)SMDP valueAge, y, mean ± SD60.60 ± 9.5959.53 ± 9.51−0.1119.04859.43 ± 9.0459.58 ± 9.490.0161.816Female128 (9.8)36 (8.6)−0.0437.51734 (8.5)34 (8.5)01Angina status CCS<.001.924 083 (6.4)32 (7.6)0.047829 (7.3)29 (7.3)0 174 (5.7)36 (8.6)0.103938 (9.5)31 (7.8)−0.0627 2476 (36.4)181 (43.1)0.1348169 (42.4)171 (42.9)0.0101 3450 (34.4)97 (23.1)−0.26996 (24.1)96 (24.1)0 4224 (17.1)74 (17.6)0.012667 (16.8)72 (18.0)0.0329NYHA<.001.922 I297 (22.7)165 (39.3)0.3391148 (37.1)146 (36.6)−0.0103 II682 (52.2)182 (43.3)−0.1785174 (43.6)181 (45.4)0.0354 III278 (21.3)59 (14.0)−0.207964 (16.0)58 (14.5)−0.0433 IV50 (3.8)14 (3.3)−0.027413 (3.3)14 (3.5)0.014Congestive HF.512.678 Never1247 (95.4)396 (94.3)−0.0484383 (96.0)378 (94.7)−0.054 Past38 (2.9)17 (4.0)0.057910 (2.5)14 (3.5)0.0509 Now22 (1.7)7 (1.7)−0.00136 (1.5)7 (1.8)0.0196Number of previous MI.041.867 0724 (55.4)261 (62.1)0.1391252 (63.2)250 (62.7)−0.0103 1525 (40.2)140 (33.3)−0.145132 (33.1)131 (32.8)−0.0053 258 (4.4)19 (4.5)0.004115 (3.8)18 (4.5)0.0362Previous cardiac surgery7 (0.5)5 (1.2)0.0604.2864 (1.0)4 (1.0)01Diabetes340 (26.0)62 (14.8)−0.3172<.00156 (14.0)61 (15.3)0.0353.689Smoking.016.587 Never577 (44.1)152 (36.2)−0.1656161 (40.4)147 (36.8)−0.073 Ex-smoker583 (44.6)216 (51.4)0.1365195 (48.9)205 (51.4)0.0501 Current147 (11.2)52 (12.4)0.034443 (10.8)47 (11.8)0.0304Hypertension1015 (77.7)262 (62.4)−0.3154<.001250 (62.7)257 (64.4)0.0362.659Hypercholesterolemia1105 (84.5)332 (79.0)−0.1351.011318 (79.7)317 (79.4)−0.00621Renal impairment∗330 (25.2)23 (5.5)−0.8691<.00126 (6.5)23 (5.8)−0.033.768Pulmonary disease118 (9.0)19 (4.5)−0.2167.0049 (2.3)19 (4.8)0.1206.083Neurologic disease57 (4.4)19 (4.5)0.0078.99615 (3.8)17 (4.3)0.0241.857Extracardiac arteriopathy96 (7.3)35 (8.3)0.0358.57627 (6.8)30 (7.5)0.0272.783AF24 (1.8)12 (2.9)0.0613.28112 (3.0)11 (2.8)−0.0151Extent of CAD.106.994 012 (0.9)7 (1.7)0.05856 (1.5)7 (1.8)0.0196 13 (0.2)2 (0.5)0.03582 (0.5)2 (0.5)0 2122 (9.3)26 (6.2)−0.130525 (6.3)25 (6.3)0 31170 (89.5)385 (91.7)0.0777366 (91.7)365 (91.5)−0.0091LMS disease348 (26.6)71 (16.9)−0.2594<.00170 (17.5)71 (17.8)0.00671LVEF.001.462 Good (>50%)1079 (82.6)311 (74.0)−0.1941314 (78.7)300 (75.2)−0.08 Moderate (30%-50%)198 (15.1)96 (22.9)0.183673 (18.3)87 (21.8)0.0836 Poor (<30%)30 (2.3)13 (3.1)0.046212 (3.0)12 (3.0)0EuroSCORE, median (IQR)1.54 (1.07-2.72)1.66 (0.88-3.49)0.0678.7341.51 (0.94-2.33)1.66 (0.88-3.49)0.0071.132Data n (%); mean ± SD; median (IQR). BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery; SMD, standardized mean difference; SD, standard deviation; CCS, Canadian Cardiovascular Society; NYHA, New York Heart Association; HF, heart failure; MI, myocardial infarction; AF, atrial fibrillation; CAD, coronary artery disease; LMS, left main stem; LVEF, left ventricular ejection fraction; EuroSCORE, European System for Cardiac Operative Risk Evaluation; IQR, interquartile range.∗Renal impairment defined as chronic kidney disease Kidney Disease: Improving Global Outcomes stage 2 or greater on the basis of estimated glomerular filtration rate. Table 1Preoperative demographicsVariableUnmatched (n = 1727)Matched (n = 798)BIMA+SVG(n = 1307)BIMA+RA(n = 420)SMDP valueBIMA+SVG(n = 399)BIMA+RA(n = 399)SMDP valueAge, y, mean ± SD60.60 ± 9.5959.53 ± 9.51−0.1119.04859.43 ± 9.0459.58 ± 9.490.0161.816Female128 (9.8)36 (8.6)−0.0437.51734 (8.5)34 (8.5)01Angina status CCS<.001.924 083 (6.4)32 (7.6)0.047829 (7.3)29 (7.3)0 174 (5.7)36 (8.6)0.103938 (9.5)31 (7.8)−0.0627 2476 (36.4)181 (43.1)0.1348169 (42.4)171 (42.9)0.0101 3450 (34.4)97 (23.1)−0.26996 (24.1)96 (24.1)0 4224 (17.1)74 (17.6)0.012667 (16.8)72 (18.0)0.0329NYHA<.001.922 I297 (22.7)165 (39.3)0.3391148 (37.1)146 (36.6)−0.0103 II682 (52.2)182 (43.3)−0.1785174 (43.6)181 (45.4)0.0354 III278 (21.3)59 (14.0)−0.207964 (16.0)58 (14.5)−0.0433 IV50 (3.8)14 (3.3)−0.027413 (3.3)14 (3.5)0.014Congestive HF.512.678 Never1247 (95.4)396 (94.3)−0.0484383 (96.0)378 (94.7)−0.054 Past38 (2.9)17 (4.0)0.057910 (2.5)14 (3.5)0.0509 Now22 (1.7)7 (1.7)−0.00136 (1.5)7 (1.8)0.0196Number of previous MI.041.867 0724 (55.4)261 (62.1)0.1391252 (63.2)250 (62.7)−0.0103 1525 (40.2)140 (33.3)−0.145132 (33.1)131 (32.8)−0.0053 258 (4.4)19 (4.5)0.004115 (3.8)18 (4.5)0.0362Previous cardiac surgery7 (0.5)5 (1.2)0.0604.2864 (1.0)4 (1.0)01Diabetes340 (26.0)62 (14.8)−0.3172<.00156 (14.0)61 (15.3)0.0353.689Smoking.016.587 Never577 (44.1)152 (36.2)−0.1656161 (40.4)147 (36.8)−0.073 Ex-smoker583 (44.6)216 (51.4)0.1365195 (48.9)205 (51.4)0.0501 Current147 (11.2)52 (12.4)0.034443 (10.8)47 (11.8)0.0304Hypertension1015 (77.7)262 (62.4)−0.3154<.001250 (62.7)257 (64.4)0.0362.659Hypercholesterolemia1105 (84.5)332 (79.0)−0.1351.011318 (79.7)317 (79.4)−0.00621Renal impairment∗330 (25.2)23 (5.5)−0.8691<.00126 (6.5)23 (5.8)−0.033.768Pulmonary disease118 (9.0)19 (4.5)−0.2167.0049 (2.3)19 (4.8)0.1206.083Neurologic disease57 (4.4)19 (4.5)0.0078.99615 (3.8)17 (4.3)0.0241.857Extracardiac arteriopathy96 (7.3)35 (8.3)0.0358.57627 (6.8)30 (7.5)0.0272.783AF24 (1.8)12 (2.9)0.0613.28112 (3.0)11 (2.8)−0.0151Extent of CAD.106.994 012 (0.9)7 (1.7)0.05856 (1.5)7 (1.8)0.0196 13 (0.2)2 (0.5)0.03582 (0.5)2 (0.5)0 2122 (9.3)26 (6.2)−0.130525 (6.3)25 (6.3)0 31170 (89.5)385 (91.7)0.0777366 (91.7)365 (91.5)−0.0091LMS disease348 (26.6)71 (16.9)−0.2594<.00170 (17.5)71 (17.8)0.00671LVEF.001.462 Good (>50%)1079 (82.6)311 (74.0)−0.1941314 (78.7)300 (75.2)−0.08 Moderate (30%-50%)198 (15.1)96 (22.9)0.183673 (18.3)87 (21.8)0.0836 Poor (<30%)30 (2.3)13 (3.1)0.046212 (3.0)12 (3.0)0EuroSCORE, median (IQR)1.54 (1.07-2.72)1.66 (0.88-3.49)0.0678.7341.51 (0.94-2.33)1.66 (0.88-3.49)0.0071.132Data n (%); mean ± SD; median (IQR). BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery; SMD, standardized mean difference; SD, standard deviation; CCS, Canadian Cardiovascular Society; NYHA, New York Heart Association; HF, heart failure; MI, myocardial infarction; AF, atrial fibrillation; CAD, coronary artery disease; LMS, left main stem; LVEF, left ventricular ejection fraction; EuroSCORE, European System for Cardiac Operative Risk Evaluation; IQR, interquartile range.∗Renal impairment defined as chronic kidney disease Kidney Disease: Improving Global Outcomes stage 2 or greater on the basis of estimated glomerular filtration rate. Preoperative demographics Data n (%); mean ± SD; median (IQR). BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery; SMD, standardized mean difference; SD, standard deviation; CCS, Canadian Cardiovascular Society; NYHA, New York Heart Association; HF, heart failure; MI, myocardial infarction; AF, atrial fibrillation; CAD, coronary artery disease; LMS, left main stem; LVEF, left ventricular ejection fraction; EuroSCORE, European System for Cardiac Operative Risk Evaluation; IQR, interquartile range. Renal impairment defined as chronic kidney disease Kidney Disease: Improving Global Outcomes stage 2 or greater on the basis of estimated glomerular filtration rate. Intraoperative Data Operative details for the unmatched and matched cohorts are summarized in Table 2. In the unmatched cohort, on-pump coronary artery bypass was performed in 37.0% of BIMA+SVG cases and in 41.0% of BIMA+RA cases. Among patients who underwent on-pump coronary artery bypass, median cardiopulmonary bypass times were similar between groups, whereas crossclamp time was shorter in the BIMA+RA cohort. The total number of grafts was slightly lower in the BIMA+RA group, reflecting less-frequent sequential grafting compared with SVG, and the number of arterial grafts was significantly greater in the BIMA+RA cohort.Table 2Intraoperative dataVariableUnmatched (n = 1727)Matched (n = 798)BIMA+SVG (n = 1307)BIMA+RA (n = 420)P valueBIMA+SVG (n = 399)BIMA+RA (n = 399)P valueONCAB483 (37.0%)172 (41.0%).158136 (34.1%)165 (41.4%).041CPB time92 (77-110)89 (74-108).30994 (78-116)88 (74-108).060Crossclamp time65 (55-78)57 (33-75)<.00166 (57-78)58 (33-75)<.001Total number of grafts3 (3-4)3 (3-3)<.0013 (3-4)3 (3-3)<.001Total number of arterial grafts2 (2-2)3 (3-3)<.0012 (2-2)3 (3-3)<.001ICOR1 (1-1.33)1 (1-1.33).3331 (1-1.33)1 (1.33).333Data n (%); median (interquartile range). BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery; ONCAB, on-pump coronary artery bypass; CPB, cardiopulmonary bypass; ICOR, index of completeness of revascularization. Table 2Intraoperative dataVariableUnmatched (n = 1727)Matched (n = 798)BIMA+SVG (n = 1307)BIMA+RA (n = 420)P valueBIMA+SVG (n = 399)BIMA+RA (n = 399)P valueONCAB483 (37.0%)172 (41.0%).158136 (34.1%)165 (41.4%).041CPB time92 (77-110)89 (74-108).30994 (78-116)88 (74-108).060Crossclamp time65 (55-78)57 (33-75)<.00166 (57-78)58 (33-75)<.001Total number of grafts3 (3-4)3 (3-3)<.0013 (3-4)3 (3-3)<.001Total number of arterial grafts2 (2-2)3 (3-3)<.0012 (2-2)3 (3-3)<.001ICOR1 (1-1.33)1 (1-1.33).3331 (1-1.33)1 (1.33).333Data n (%); median (interquartile range). BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery; ONCAB, on-pump coronary artery bypass; CPB, cardiopulmonary bypass; ICOR, index of completeness of revascularization. Intraoperative data Data n (%); median (interquartile range). BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery; ONCAB, on-pump coronary artery bypass; CPB, cardiopulmonary bypass; ICOR, index of completeness of revascularization. Additional RA-specific operative details—including TAR rates, target-territory distribution, and proximal and distal anastomotic configuration—are provided in Table E1. Use of the RA varied substantially over time, with annual trends shown in Figures E1 and E2. Temporal variation in RA use likely reflects evolving surgical practice rather than a single identifiable factor. Early adoption was driven by emerging evidence supporting arterial graft durability, whereas subsequent fluctuation may relate to variability in surgeon preference, technical considerations (including susceptibility to vasospasm), and the need for careful target vessel selection to minimize competitive flow. In addition, inconsistent use of perioperative antispasm strategies and the absence of routine graft patency surveillance may have influenced conduit choice over time. No systematic signal of increased arm complications or inferior radial artery performance was identified in our dataset, although such outcomes were not comprehensively captured. In-Hospital Outcomes In the matched cohort, postoperative renal complications occurred more frequently in the RA group (2.5% vs 0.5%; P = .042). Other major in-hospital outcomes—including mortality, arrhythmias, pulmonary and infectious complications, and neurologic events—were broadly comparable between groups, with no significant differences observed (Table 3). Postoperative renal complications were examined in detail, with findings summarized in Table E2, Table E3, Table E4. Events were rare, heterogeneous in nature, and not attributable to baseline renal function, surgical era, or conduit choice.Table 3In-hospital outcomesVariableUnmatched (n = 1727)Matched (n = 798)BIMA+SVG(n = 1307)BIMA+RA(n = 420)P valueBIMA+SVG(n = 399)BIMA+RA(n = 399)P valueArrhythmias205 (15.7)61 (14.5).6270 (17.5)57 (14.3).246Return to operating theater63 (4.8)22 (5.2).8316 (4.0)19 (4.8).73Pulmonary complications235 (18.0)72 (17.1).75171 (17.8)69 (17.3).926Neurologic complications17 (1.3)9 (2.1).3163 (0.8)9 (2.3).146Infectious complications96 (7.3)39 (9.3).23630 (7.5)37 (9.3).444Deep sternal wound infection17 (1.3)2 (0.5).2549 (2.3)2 (0.5).069Postoperative renal complications27 (2.1)10 (2.4).8462 (0.5)10 (2.5).042GI complications17 (1.3)6 (1.4)17 (1.8)6 (1.5)1Pleural effusions40 (3.1)6 (1.4).1036 (1.5)6 (1.5)1Multisystem failure13 (1.0)4 (1.0)12 (0.5)3 (0.8)1In-hospital and 30-d mortality14 (1.1)4 (1.0)14 (1.0)3 (0.8)1Postoperative stay, d, mean ± SD8.00 ± 8.758.80 ± 8.48.18.21 ± 10.908.85 ± 8.62.358Data n (%); mean ± SD. BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery; GI, gastrointestinal; SD, standard deviation. Table 3In-hospital outcomesVariableUnmatched (n = 1727)Matched (n = 798)BIMA+SVG(n = 1307)BIMA+RA(n = 420)P valueBIMA+SVG(n = 399)BIMA+RA(n = 399)P valueArrhythmias205 (15.7)61 (14.5).6270 (17.5)57 (14.3).246Return to operating theater63 (4.8)22 (5.2).8316 (4.0)19 (4.8).73Pulmonary complications235 (18.0)72 (17.1).75171 (17.8)69 (17.3).926Neurologic complications17 (1.3)9 (2.1).3163 (0.8)9 (2.3).146Infectious complications96 (7.3)39 (9.3).23630 (7.5)37 (9.3).444Deep sternal wound infection17 (1.3)2 (0.5).2549 (2.3)2 (0.5).069Postoperative renal complications27 (2.1)10 (2.4).8462 (0.5)10 (2.5).042GI complications17 (1.3)6 (1.4)17 (1.8)6 (1.5)1Pleural effusions40 (3.1)6 (1.4).1036 (1.5)6 (1.5)1Multisystem failure13 (1.0)4 (1.0)12 (0.5)3 (0.8)1In-hospital and 30-d mortality14 (1.1)4 (1.0)14 (1.0)3 (0.8)1Postoperative stay, d, mean ± SD8.00 ± 8.758.80 ± 8.48.18.21 ± 10.908.85 ± 8.62.358Data n (%); mean ± SD. BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery; GI, gastrointestinal; SD, standard deviation. In-hospital outcomes Data n (%); mean ± SD. BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery; GI, gastrointestinal; SD, standard deviation. Long-Term Survival Median follow-up was 10.52 years (interquartile range [IQR], 5.99-14.29) for BIMA+SVG and 21.58 years (IQR, 16.48-22.64) for BIMA+RA in the unmatched cohort. In the matched cohort, median follow-up was 11.22 years (IQR, 6.62-15.63) and 21.53 years (IQR, 15.90-22.60), respectively, reflecting the earlier era of predominant RA use compared with ongoing SVG use throughout the study period. In the unmatched cohort, Kaplan-Meier analysis demonstrated superior survival with BIMA+RA (log-rank χ2 = 5, P = .03), with 20-year survival of 96.6% versus 93.7% for BIMA + SVG (Figure 1, Tables E5 and E6). After propensity matching, survival differences attenuated and were not statistically significant (log-rank χ2 = 2.7, P = .10) (Figure 2, Tables E7 and E8). In multivariable Cox regression of the unmatched cohort, BIMA+RA remained an independent predictor of improved survival (hazard ratio [HR], 0.345; 95% CI, 0.170-0.698; P = .003). Increasing age (HR, 1.063 per year; 95% CI, 1.034-1.093), ≥2 previous myocardial infarctions (HR, 2.276; 95% CI, 1.008-5.136), and preoperative atrial fibrillation (HR, 2.570; 95% CI, 1.010-6.540) were associated with greater mortality risk (Figure 3 and Table E9).Figure 1Kaplan-Meier survival curves stratified by grafting strategy in the unmatched cohort. Shaded areas represent the 95% CIs. BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery.Figure 2Kaplan-Meier survival curves stratified by grafting strategy in the matched cohort. Shaded areas represent the 95% CIs. BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery.Figure 3Forest plot displaying multivariable Cox regression. BIMA, Bilateral internal mammary artery; RA, radial artery; HR, hazard ratio; NYHA, New York Heart Association; HF, heart failure; MI, myocardial infarction; AF, atrial fibrillation; LVEF, left ventricular ejection fraction. Kaplan-Meier survival curves stratified by grafting strategy in the unmatched cohort. Shaded areas represent the 95% CIs. BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery. Kaplan-Meier survival curves stratified by grafting strategy in the matched cohort. Shaded areas represent the 95% CIs. BIMA, Bilateral internal mammary artery; SVG, saphenous vein graft; RA, radial artery. Forest plot displaying multivariable Cox regression. BIMA, Bilateral internal mammary artery; RA, radial artery; HR, hazard ratio; NYHA, New York Heart Association; HF, heart failure; MI, myocardial infarction; AF, atrial fibrillation; LVEF, left ventricular ejection fraction. Discussion The principal findings are 3-fold. First, crude survival in the unmatched cohort favored the BIMA+RA strategy, but this apparent advantage diminished after propensity matching, indicating that baseline differences between groups contributed substantially to the observed survival separation. Second, early postoperative outcomes were largely comparable between groups, with the exception of a greater incidence of postoperative renal complications in the matched RA cohort. Third, despite the attenuation of survival differences after matching, RA use remained independently associated with improved long-term survival in multivariable analysis, suggesting a potential late benefit that may not be fully captured by matched comparisons. The high 10-year survival rates observed in both cohorts reflect the favorable baseline risk profile of patients selected for BIMA grafting—typically younger individuals with preserved ventricular function and limited comorbidity—and the outcomes expected in a high-volume arterial-grafting center with consistent perioperative care. These findings align with, but also refine, the existing literature on multiarterial grafting. Meta-analytic evidence consistently demonstrates that the addition of a third arterial conduit improves long-term survival without increasing perioperative risk.9, 10, 11 Randomized data further support the biological advantages of RA grafting, showing superior angiographic patency and reduced adverse cardiac events compared with SVG at midterm follow-up, although without a demonstrated mortality benefit at approximately 5 years.8 Large observational studies of TAR, particularly those combining bilateral internal thoracic arteries with RA, report improved unadjusted and propensity-matched survival.12 However, studies specifically comparing BIMA+RA with BIMA+SVG have produced more heterogeneous results. Several propensity-matched analyses have shown no significant survival difference within the first decade, with divergence emerging only in later follow-up.13, 14, 15, 16 Yoshida and colleagues17 similarly found no matched survival or patency advantage for RA when used to revascularize the right coronary system in the context of BIMA grafting to the left coronary circulation. The present study's attenuation of survival benefit after matching, coupled with a renal risk signal, is consistent with this complex and sometimes conflicting body of evidence. Mechanistically, the potential long-term advantage of RA grafting is grounded in the superior durability of arterial conduits. Arterial grafts exhibit enhanced resistance to atherosclerosis and maintain long-term patency more effectively than SVGs, which are prone to progressive intimal hyperplasia and atherosclerotic degeneration.2,5 However, arterial graft performance is highly sensitive to competitive flow. Optimal outcomes are achieved when the target vessel exhibits severe stenosis (≥80%-90%), minimizing competitive native flow that can precipitate graft dysfunction or the “string sign.”6,7 Territory selection also matters: RA grafts may perform differently when directed to the right coronary artery versus the circumflex system, reflecting variations in flow dynamics and myocardial demand.8 In addition, RA physiology—particularly its propensity for vasospasm—may influence early graft behavior. Perioperative antispasm strategies, including papaverine, nitrates, and calcium channel blockers, are therefore integral to optimizing RA performance.2 The greater rate of postoperative renal complications in the BIMA+RA group reflects a composite of stage 2 to 3 acute kidney injury and requirement for standard renal-replacement therapy (dialysis or hemofiltration), rather than dialysis-dependent renal failure. As shown in Table E2, Table E3, Table E4, events were rare, not era-specific, unrelated to baseline renal function, and often associated with perioperative hemodynamic instability. Several limitations must be acknowledged. The retrospective, single-center design introduces inherent risks of selection bias and residual confounding, even with robust propensity matching. Multiple secondary end points were evaluated, and these analyses were not adjusted for multiplicity, which may increase the risk of type I error. Although follow-up was complete and survival analyses were methodologically sound, the 27-year study period spans substantial changes in surgical practice, perioperative care, and conduit-harvesting techniques, raising the possibility of era effects. The absence of routine graft-patency imaging limits mechanistic interpretation of survival differences, particularly regarding competitive flow and conduit-specific failure patterns. Although matching achieved excellent covariate balance, unmeasured factors—such as target-vessel stenosis severity, conduit quality, or surgeon-specific technical variations—may still have influenced conduit selection and outcomes. The relatively small number of postoperative dialysis events also limits the precision of estimates related to renal risk. In addition, only 46 patients who received BIMA + SVG remained at risk at 20 years, which constrains interpretation of the extreme tail of the survival curves. Perioperative calcium-channel antagonist administration was variably documented, limiting our ability to assess its influence on postoperative outcomes. Finally, although multivariable Cox regression identified RA use as an independent predictor of improved survival, this finding should be interpreted cautiously, given the potential for residual confounding. Despite these limitations, the study has important implications for clinical practice and policy. The findings reinforce the principle that arterial conduits offer durable long-term benefits and should be considered in patients with reasonable life expectancy, particularly younger individuals with favorable anatomy. This aligns with contemporary guidelines emphasizing arterial grafting strategies.1,2 However, the renal signal associated with RA use highlights the need for careful patient selection, particularly in individuals with borderline renal function or heightened susceptibility to perioperative hemodynamic instability. For greater-risk patients—such as those with advanced age, obesity, diabetes, or significant pulmonary disease—SVG in combination with internal thoracic arteries remains a pragmatic and effective strategy for achieving complete revascularization.18 These results underscore the importance of individualized conduit selection that integrates anatomical, physiological, and comorbidity considerations rather than relying on a uniform arterial-first approach. Several unanswered questions remain. The mechanisms underlying the observed renal vulnerability in RA recipients require targeted investigation, ideally through prospective studies incorporating granular intraoperative hemodynamic data, vasoactive medication profiles, and renal biomarkers. Future research should also use multicenter datasets with detailed mapping of conduit-to-target relationships, enabling more precise evaluation of competitive flow and territory-specific outcomes. Randomized trials comparing RA versus SVG as the third conduit in the context of BIMA grafting—adequately powered for long-term survival and renal end points—would provide definitive evidence to guide practice. In addition, the role of contemporary RA harvesting techniques, antispasm protocols, and postoperative medical therapy warrants further exploration. In conclusion, the apparent survival advantage of using the RA as a third conduit in BIMA CABG disappeared after propensity matching, suggesting selection bias influenced initial findings. Early outcomes were similar overall, although RA use was linked to more postoperative renal complications. These results underscore the trade-off between the long-term durability of arterial grafts and their perioperative risks. Tailored conduit selection—guided by anatomy, comorbidities, and individual risk profiles—remains central to optimizing multiarterial CABG outcomes. Webcast You can watch a Webcast of this AATS meeting presentation by going to: https://www.aats.org/resources/167-does-the-third-conduit-mat-13317. Conflict of Interest Statement The authors reported no conflicts of interest. The Journal policy requires editors and reviewers to disclose conflicts of interest and to decline handling or reviewing manuscripts for which they may have a conflict of interest. The editors and reviewers of this article have no conflicts of interest.