Survival of multiple arterial grafting in diabetic populations: a 20-year national experience

Abstract OBJECTIVES Diabetics may have diminished survival after coronary artery bypass grafting even with multiple arterial revascularization. We compared multi-arterial versus single-arterial grafting (SAG) survival in diabetic and non-diabetic patients undergoing primary isolated bypass surgery. METHODS This is a retrospective analysis of the Australian and New Zealand Society of Cardiac-Thoracic Surgical Database from June 2001 to January 2020. Patients were classified as having either single or multiple arterial grafting irrespective of the number of venous grafts. The end points were long-term all-cause mortality and 30-day clinical outcomes, which was compared in 1:1 propensity score-matched patients. Cox regression model was used to assess interactions between diabetes and the treatment effect of multi-arterial grafting, reported as hazard ratios (HRs) and confidence intervals (CIs). Short-term outcomes were compared with McNemar’s paired t-test. RESULTS From 69 624 patients, matching generated 17 474 non-diabetic and 10 989 diabetic patient pairs. At a median [interquartile range] of 5.9 [3.2–9.6] years postoperative, mortality was significantly lower after multi-arterial grafting for both diabetic (HR, 0.83; 95% CI, 0.76–0.90, P < 0.001) and non-diabetic (HR, 0.88; 95% CI, 0.82–0.95; P < 0.001) cohorts than SAG. The incidence of 30-day myocardial infarction was significantly higher in single than multiple arterial grafting for both cohorts (diabetic, P = 0.029; non-diabetic, P < 0.001). The interaction analysis suggested an insignificant effect of diabetes (P = 0.55) on the observed survival advantage. Further stratification by diabetic management generated consistent results. CONCLUSIONS Multi-arterial grafting was associated with improved overall survival compared to SAG for both non-diabetic and diabetic patients.


INTRODUCTION
Diabetes mellitus (DM) has become increasingly prevalent [1] as one of the major risk factors for coronary artery disease, which increases the anatomic disease severity and postoperative mortality [2] in patients undergoing coronary revascularization. Coronary artery bypass grafting (CABG) is shown to have superior outcomes compared with percutaneous coronary intervention for diabetic patients with multivessel coronary artery disease [3][4][5].
Conventional technique involves the left internal mammary artery and supplementary saphenous vein grafts (SVG). However, multiple arterial revascularization has gained popularity due to the common late progressive atherosclerosis of SVG. Recent international guidelines recommend the use of a second arterial conduit in patients without a contraindication [6]. Strong evidence has reported significant relative reduction of mortality and myocardial infarction (MI) associated with multiple arterial grafting (MAG) [7,8], but uncertainty exists over whether the benefits of MAG outweigh the operative risks in the diabetic populations.
We therefore sought to compare the long-term survival of patients undergoing MAG versus single-arterial grafting (SAG) in diabetic and non-diabetic cohorts in Australian & New Zealand Society of Cardiothoracic Surgeons (ANZSCTS) National Database.

Ethics statement
The Melbourne Health Institutional Review Board (# 2011.164) has approved this study with an official waiver for individual patient consent.

Study design
This is a retrospective observational cohort study that identified all adult patients who underwent primary isolated CABG with > _2 grafts at 59 cardiac centres between June 2001 and January 2020 from the ANZSCTS registry. This national database prospectively records patient demographic, risk factors and procedural details checked regularly and validated through multiple administrative linkages, including to the National Death Index. Patients who had reoperations, concomitant or previous cardiac surgery procedures, single-graft and no arterial grafting were excluded. Complete case analyses were conducted.

Patient selection
MAG was defined as receiving at least 2 arterial grafts with or without supplementary SVG, whereas SAG patients had only 1 arterial graft and at least 1 supplementary SVG. No further distinctions were made with regards to preoperative and intraoperative characteristics. The choice of graft configurations was by surgeon preference. All procedures and medical management were at the discretion of each centre but no considerable change in practice was reported throughout the study period.
A total of 36 preoperative and intraoperative baseline covariates were collected for both groups, including age, sex, body mass index, creatinine level, hypertension, hypercholesterolaemia, DM, smoking history, dialysis, arrhythmia, cerebrovascular event, peripheral vascular disease, chronic obstructive lung disease, MI, left ventricular ejection fraction, congestive heart failure, New York Heart Association classification, left main disease, number of grafts, number of diseased vessels, operative status, on-pump and minimally invasive surgery.

Study outcomes
The primary end point was all-cause mortality reported as timeto-event measured from the date of the index operation to the longest follow-up. Administrative linkage had population coverage for death from a Federal Government agency with completeness of follow-up measured by follow-up index [9]. Secondary end points included 30-day mortality and 30-day readmission to the hospital due to a composite of MI, deep sternal wound infection (DSWI), arrhythmia, congestive heart failure or recurrent angina. Each component was then individually analysed. Readmission excluded admission to emergency, short-stay wards or planned transfer to rehabilitation facility.
For further evaluation, the diabetic population was stratified by degree of disease severity. The 3 subgroups were no medical intervention (no treatment or dietary control only), oral medication and insulin dependent patients, in increasing order of severity. Classifications were determined by the most aggressive diabetes control therapy at the time of surgery. Patients with insulin usage may receive other treatments simultaneously.

Statistical analysis
A complete case retrospective analysis was performed with categorical variables reported as frequency counts and percentages, and continuous variables reported as median [interquartile range] or mean ± standard deviations (SDs) after assessment of normality.
Propensity score matching (PSM) was used to minimize potential selection bias by controlling for imbalanced baseline characteristics. Using generalized linear regression, the propensity score estimation algorithm incorporated all 36 variables from Table 1 to balance covariates between MAG and SAG cohorts. Nonreplacement one-to-one greedy matching with calliper width of 0.2 SD of the logit of the propensity score was performed separately for the diabetic and non-diabetic cohorts. Absolute standardized mean differences were evaluated before and after PSM for balance diagnostics; a difference value of <0.1 generally indicated balance between patient groups [10].
A Cox proportional hazard model was applied to the time-toevent analysis estimating hazard ratios (HRs) and 95% confidence intervals (CIs) for long-term mortality in the matched diabetic and non-diabetic cohorts separately, with stratification on patient pairs. The influence of diabetes on the survival were assessed by using diabetes as an interaction term embedded in the Cox regression model. The proportionality assumption was verified with Schoenfeld residuals. Graphical visualization on the rate of death over time was achieved by Kaplan-Meier (KM) survival functions for the matched groups. The secondary short-term end points were compared via McNemar's paired t-test.
Sensitivity analysis was conducted through (i) introducing an exact-match constraint on the number of grafts and (ii) additionally adjusting for the year of operation to ensure robustness of our conclusion.
The presence of significance was defined by a two-tailed P-value of <0.05. P-values were not adjusted for multiplicity, and

RESULTS
This study identified 69 624 eligible patients in total, which included 39 478 (56.7%) patients undergoing multiple arterial revascularizations and 25 751 (37.0%) patients with DM (Fig. 1). Among those with diabetes, 483 (1.9%) were not on any treatment, 3236 (12.6%) had dietary control, 14 349 (55.8%) were on oral treatment and 7683 (29.9%) were insulin-dependent. The proportional hazards assumption for the Cox regression model was verified with weighted residuals for all subsequent analyses. The median follow-up duration was 5.9 [interquartile range 3.2-9.6] years postoperatively. The mean follow-up index was 0.999 with an SD of 0.030. In the unmatched cohort, compared to SAG, the MAG group was younger, more likely to be male and to have hypercholesterolaemia, but less likely to have other preoperative comorbidities including hypertension, DM, smoking history, dialysis, arrhythmia, MI and cerebrovascular event. Compared to non-diabetic patients, the diabetic group was younger and had higher female proportion and incidence of comorbidities with more severe coronary disease (Table 1).

Primary outcome
Non-diabetics. The non-diabetic cohort generated 17 474 matched pairs, where the MAG group was associated with significantly improved long-term survival than the SAG group (HR, 0.88; 95% CI, 0.82-0.95; P < 0.001). The KM curve is shown in Fig. 2. Diabetics. In the diabetic population of 10 989 matched pairs, the MAG group again had superior survival than the SAG group (HR, 0.83; 95% CI, 0.76-0.90, P < 0.001). Interaction term analysis found insignificant subgroup interaction effect of diabetes on the survival difference (P = 0.55). The KM curve is shown in Fig. 3. Matching has reduced all standardized mean differences to below 10% except for the number of grafts used (SMD 0.11-0.12) ( Table 2). The propensity score distribution plot confirmed high-quality matchings (Supplementary Material,  Fig. S1).

Secondary outcomes
The incidence of 30-day MI was significantly higher after SAG than MAG in both diabetic (P = 0.029) and non-diabetic (P < 0.001) patients. Other short-term clinical outcomes remained similar between 2 procedures (Table 3).

DISCUSSION
This nation-wide study of 25 751 diabetic and 43 873 nondiabetic patients is the largest investigation in the literature to date comparing the impact of multiple arterial revascularization on long-term survival in diabetic and non-diabetic patients. The diabetic cohort had greater perioperative comorbidities and aggressive coronary disease than patients without diabetes. Nevertheless, MAG was associated with reduced all-cause mortality for both diabetic and non-diabetic patients. The interaction term analysis found insignificant subgroup effect from diabetes, but mortality rates numerically increased with diabetic severity. For insulin-dependent diabetes, the treatment effect remained similar to other diabetic treatment groups, but only with borderline significance. The incidence of 30-day MI was significantly higher in patients undergoing SAG regardless of diabetic status.
A novel finding of this study was that the survival benefit provided by MAG appears to be time-sensitive, which appears maximal between 6-9 years and declines after 9 years postoperatively. The mechanism remains unknown but could relate to the late graft failure of saphenous veins in MAG patients which results in an increased cumulative mortality effect despite a reduced number of SVG being at risk of failure compared to SAG.
The FREEDOM trial documented a strong survival benefit as well as a reduction in major cardiac and cerebrovascular events following CABG compared to drug-eluting stents [11]. Similarly, the SYNTAX Extended Survival Study with 12-year follow-up, demonstrated improved outcomes for MAG in diabetic patients. The updated ACC/AHA/SCAI guideline for bypass conduits proposed Class I recommendation for the use of radial artery as the supplementary conduit in preference to SVG for the second most important target [12], based on ease of use, superior durability and reduced adverse cardiac events and mortality [13][14][15]. However, MAG usage remains low in the world practice, and especially in diabetics [16].
Intense debate relating to the use of MAG in diabetic population continues because of discordant findings reported by prior investigations. Kunihara et al. [17] suggested indifferent long-term survival between LIMA-RA-T grafts and a combination of LIMA and sequential SVG. However, Yamaguchi et al. [18] reported significantly higher 12-year overall and major cardiac and cerebrovascular events-free survival in both diabetic (64.9% vs 58.8%; P = 0.041) and non-diabetic (71.4% vs 63.8%; P = 0.014) groups treated with MAG rather than SAG. The cumulative survival rates are consistent with our findings. A recent post hoc analysis of the Arterial Revascularization Trial [19] which documented significantly lower all-cause mortality at 10 years after MAG in patients with (HR, 0.68; 95% CI, 0.51-0.91); and without diabetes (HR, 0.83; 95% CI, 0.69-1.00). Importantly, most relevant studies only focused on a subset of MAG procedures (e.g. bilateral internal mammary technique) with limited generalizability and underpowered statistical inferences, which should be interpreted cautiously in the context of MAG.  MAG has a perceived increased risk of DSWI if BIMA is used. A meta-analysis of observational studies found that BIMA grafting increased DSWI by 62% [20]. The BIMA arm from the Arterial Revascularization Trial [21] also demonstrated higher incidence of DSWI (3.5% vs 1.9%; P = 0.005) and sternal reconstruction procedures (1.9% vs 0.6%; P = 0.002). In addition, its post hoc MAG  analysis [19] on diabetic patients reported incrementally higher DSWI rates (7.9% vs 4.78%; P = 0.24). The influence of this complication on the long-term postoperative life expectancy after CABG is however uncertain. A recent meta-analysis of 24 studies at a mean follow-up duration of 3.5 years [22] reported increased overall mortality in patients with DSWI, whereas another 10-year study [23] suggested insignificant influence on risk of death in a general cardiac-surgical patient population. In this series, we stratified diabetic population by the type of medical management to comprehensively evaluate the subgroup interaction between DM and treatment effect of MAG versus SAG. The survival benefit of MAG appeared similar in all 3 diabetic cohorts. Significance was only detected in patients with oral medication but other groups consistently demonstrated a trend towards better MAG survival. In addition, our paired t-test did not detect any significant increase of DSWI in the MAG groups at least within 30-day follow-up. The use of MAG also appeared to have a protective effect against early MIs, but the mechanism would require further investigation.

Limitations
The following limitations have to be considered during the interpretation of results. Although extensive PSM has been used to minimize selection bias inherent in retrospective observational studies, unmeasured confounders may limit analysis. Our database did not document harvest techniques, quality of conduits, preoperative coronary stenosis, target vessels or postoperative medical management. The number of grafts could not be perfectly matched (1-2% in excess to the 10% SMD recommendation) due to rigorous adjustment of 36 baseline covariates. The MAG groups had slightly higher average number of grafts (mean difference: 0.1 graft per patient) in both diabetic and nondiabetic populations, potentially introducing some bias. Nevertheless, a sensitivity analysis imposing exact matching on the number of grafts generated consistent results with the primary analysis.

CONCLUSIONS
MAG was associated with significantly lower overall mortality compared to SAG in CABG patients with, or without, diabetes. A survival treatment effect was present in diabetic patients for MAG, though the insulin-dependent cohort failed to reach statistical significance.

SUPPLEMENTARY MATERIAL
Supplementary material is available at EJCTS online.