The effect of diabetes on long-term outcomes in patients following coronary artery bypass grafting

Abstract

OBJECTIVES

Diabetes mellitus is a risk factor for coronary artery disease, but its role following coronary artery bypass grafting (CABG) is still unclear and few data on long-term outcomes are available. This study aimed to evaluate the impact of diabetes on long-term outcomes after CABG.

METHODS

The PRIORITY (PRedictIng long-term Outcomes afteR Isolated coronary arTery bypass surgerY) project is an observational cohort study merging 2 prospective multicentre studies on isolated CABG. Follow-up information was obtained through administrative databases and was truncated 10 years after the intervention. Baseline differences between patients with and without diabetes were balanced with inverse probability of treatment weighting.

RESULTS

The cohort consisted of 10 989 patients with complete follow-up information who underwent isolated CABG (diabetes 32.3%). Diabetes did not affect short-term mortality [odds ratio (OR) 0.90, 95% confidence interval (CI) 0.73–1.10] and repeat revascularization (OR 0.79, 95% CI 0.42–1.49), while it is related to lower incidence of 30-day major adverse cardiac and cerebrovascular events (OR 0.67, 95% CI 0.60–0.76), acute myocardial infarction (OR 0.60, 95% CI 0.51–0.70) and stroke (OR 0.47, 95% CI 0.28–0.77). Diabetic patients had a higher long-term risk for major adverse cardiac and cerebrovascular event [weighted hazard ratio (HR) 1.31, 95% CI 1.26–1.37], mortality (HR 1.45, 95% CI 1.37–1.53), as well as stroke (HR 1.38, 95% CI 1.25–1.53) and myocardial infarction (HR 1.39, 95% CI 1.26–1. 53). Diabetes had not been associated with an increased incidence of repeated revascularization up to 10 years (HR 1.04, 95% CI 0.96–1.12).

CONCLUSIONS

Diabetic patients had worse long-term outcomes. Diabetes may have a greater negative impact on micro-vasculopathy than grafts, as evidenced by the increased long-term incidence of myocardial infarction without affecting myocardial revascularization.

Open in new tabDownload slide

INTRODUCTION

The relationship between diabetes and ischaemic heart disease (IHD) is well-documented [1]. Diabetes exacerbates the progression of atherosclerosis, leading to plaque formation and narrowing of the coronary arteries. Individuals with diabetes have a higher likelihood of developing IHD due to both microvascular and macrovascular complications [2]. The combination of diabetes and IHD poses a significant clinical challenge, often requiring comprehensive management strategies to mitigate the risk of adverse cardiovascular events [3].

Coronary artery bypass grafting (CABG) is the treatment of choice for multivessel coronary artery disease in diabetic patients, with a class IA of recommendation in the 2018 ESC/European Society of Cardiothoracic Surgery (EACTS) guidelines on myocardial revascularization [4]. However, the effect of diabetes on outcomes after CABG is still debated, with contrasting evidence [5]. There are few studies analysing how diabetes affects the long-term outcome of patients undergoing surgical myocardial revascularization and most of them are performed on limited cohorts.

The PRIORITY (PRedictIng long term Outcomes afteR Isolated coronary arTery bypass surgerY) project was developed through a collaboration between the National Centre for Epidemiology, Surveillance and Health Promotion of the Italian National Institute of Health and the Italian Society of Cardiac Surgery (SICCH) to assess the long-term outcomes of isolated CABG in Italy [6]. The PRIORITY project includes information on the diabetic status of patients, and it could be employed to assess the role of diabetes on outcomes up to 10 years. Therefore, we aimed to analyse the role of diabetes on long-term outcomes after isolated CABG.

PATIENTS AND METHODS

Data sources and study population

The PRIORITY project was designed to assess the long-term follow-up of an observational cohort derived from merging 2 multicentre prospective studies conducted in 2002–2004 (Italian-CABG project, IT-CABG; Supplementary Material, Appendix S1) and 2007–2008 (Italian-CABG project 2, IT-CABG II; Supplementary Material, Appendix S2) [7, 8]. Data from the 2 studies were collected prospectively and shared the same variables coded according to the EuroSCORE (European System for Cardiac Operative Risk Evaluation) criteria [8, 9]. The data collected and processed data in the period 2002–2008 fell within the Current Research Activities envisaged Italian regions, the National Health Institute and Universities (art. 5, Legislative Decree 282/99 of Italian Government) [7] did not require the written consent of the participants.

Follow-up information was obtained by linking the cohort with the National Hospital Discharged Records database and the Tax Registry Information System, provided by the Italian Ministry of Health through a collaboration with the Italian National Program for Outcome Evaluation (AGENAS). Methods employed for record linkage among databases have been previously described [7–9]. Outcome data were available from administrative records and registries since 2004; therefore, patients enrolled in the IT-CABG study before 2004 were excluded from the present analysis. Follow-up was truncated at 10 years after the intervention.

The study groups were identified according to the definition of diabetes in the 2 original multicentre IT-CABG e IT-CABG II, classified as diabetes on insulin or oral hypoglycaemic medications (without identifying the type of drug).

Outcomes

The primary outcome of this study was a composite of the following major adverse cardiac and cerebrovascular events (MACCEs) up to 10 years: all-cause mortality, acute myocardial infarction, repeat revascularization with percutaneous coronary intervention (PCI) or CABG, and stroke. Secondary outcomes were long-term all-cause mortality, acute myocardial infarction, repeat revascularization with PCI or CABG, and stroke. All outcomes were identified through specific administrative codes of rehospitalization.

Data analysis

The distributions of continuous variables were evaluated using the one-sample Kolmogorov–Smirnov test. Continuous variables are presented as mean and standard deviation for data with a normal distribution or as median and percentile for data with a non-Gaussian distribution. Categorical variables are expressed as numbers and proportions.

The potential selection bias related to the observational nature of the study was mitigated with a propensity score approach [10]. The probability of a diabetic patient undergoing isolated CABG was generated by a non-parsimonious logistic regression model, including 23 variables listed in Table 1. Propensity scores were employed for calculating the inverse probability of treatment weight (IPTW), defined as the inverse of the probability of receiving the treatment that the patient received, and estimates the average treatment effect on the treated. The balance between the weighted groups was evaluated with standardized mean differences [10]. An absolute standardized difference < 0.1 was considered an acceptable balance between covariates. Propensity score weighting gains the advantage of using all the subjects in the 2 treatment groups for the outcome analysis compared to other balancing methods based on propensity score, such as propensity score matching.

The study outcomes were then evaluated with IPT-weighted multivariable analyses. Thirty-day end-points were analysed with grouped frailty multivariable logistic regression models, accounting for heterogeneity among centres with a random-intercept parameter. The variables selection for the logistic models was performed by a back-and-forward stepwise regression (probability of stay = 0.10, probability of entry = 0.05), considering all factors included in Table 1 [11, 12]. Long-term outcomes were analysed with IPT-weighted time-to-event non-parametric (Kaplan–Meier estimators) and semi-parametric methods [11, 12]. Time-to-event distributions were separately analysed according to the primary event type. Time-to-death and time-to-MACCE were modelled with a grouped frailty semi-parametric Cox model, accounting for heterogeneity among centres with a random-intercept parameter. Grouped frailty semi-parametric Fine and Gray models were used in competing risk analysis for time to rehospitalization for myocardial infarction, repeat revascularization and/or stroke with death as a competing risk, accounting for heterogeneity among centres with a random-intercept parameter [11, 12]. The variables selection for the Cox models was performed by a back-and-forward stepwise regression (probability of stay = 0.10, probability of entry = 0.05), considering all factors included in Table 1. The variables selection for the Fine and Gray models was performed by a back-and-forward stepwise regression with the Akaike information criterion as selection criteria, considering all factors included in Table 1. Hazards proportionality and time-dependent effects were checked with the analysis of Schoenfeld residuals, Kolmogorov–Smirnov test and Cramer von Mises test. All regression models were adjusted for all the variables included in Table 1.

Missing values occurred for variables ‘left ventricular ejection fraction’ (2.1%), ‘neurological dysfunction disease’ (0.8%), haemodynamic instability (0.3%), cardiogenic shock (02%), unstable angina (0.1%), ‘recent myocardial infarction’ (0.1%), ‘chronic pulmonary disease’ (0.1%), ‘extracardiac arteriopathy’ (0.1%).

Missing values were substituted by single conditional mean imputation, as described, to reduce bias and increase statistical power [12].

Two-sided statistics were performed with a significance level of 0.05. Analyses were performed with R language [R 4.3.3; R Development Core Team (2023). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3.900051-07-0. http://www.R-project.org/].

RESULTS

The cohort consisted of 10 989 patients with complete follow-up information who underwent isolated CABG (32.3% with diabetes). The follow-up time was 86005.52 patient-years (median 7.9 years; 26845.19 patient-years in diabetic patients and 59051.62 patient-years in non-diabetic patients). The baseline characteristics before and after IPTW were different between the 2 groups (Table 1). Groups were not homogeneous before IPTW, while the distribution of baseline characteristics was well balanced after IPTW, as demonstrated by the mean standardized differences (Table 1 and Supplementary Material, Fig. S1). It is worth noting that the overall number of patients in the 2 groups is not integer because of the IPTW. The analysis of Propensity Score distribution did not reveal significant outliers requesting sensitivity analysis (Supplementary Material, Fig. S2).

Diabetes was associated with a higher long-term incidence of MACCE in the unweighted [hazard ratio (HR) 1.32, 95% confidence interval (CI) 1.24–1.40; P-value < 0.001; Supplementary Material, Fig. S3 and Table S1) and IPTW multivariate (HR 1.22, 95% CI 1.17–1.27; P-value < 0.001) models. Heterogeneity among centres was significant (random parameter θ = 0.023, P-value < 0.001). The proportional hazards assumption was not respected (proportional hazard assumption test P-value < 0.001). Schoenfeld residuals suggested a different effect of diabetes in the perioperative period, and hence, we landmarked at 30 days. As shown in Table 2 and Fig. 1, the relationship between diabetes and MACCE is confirmed to be significant, also excluding the perioperative period (IPTW HR 1.31, 95% CI 1.26–1.37, P-value < 0.001). Heterogeneity among centres is confirmed (random parameter θ = 0.02, P-value < 0.001), while the proportionality hazards assumption was held excluding perioperative events (P-value = 0.28), confirming that this relationship is constant over time after 30 days.

Analysing single outcomes, diabetes was associated with worse long-term survival in both the unadjusted and weighted adjusted models (HR 1.56, 95% CI 1.45–1.68, P < 0.001 and HR 1.39, 95% CI 1.32–1.47, P < 0.001; Supplementary Material, Fig. S4 and Table S1). The estimate of random parameter for heterogeneity was significant (θ = 0.038, P-value < 0.001). The analysis of the relationship excluding the first 30 days confirmed this association (IPT-weighted HR 1.45, 95% CI 1.37–1.53, P-value < 0.001; random parameter θ = 0.033 P-value < 0.001; Table 2 and Fig. 2) and the proportional hazard ratio assumption (proportional hazard assumption test P-value = 0.36).

Diabetes was related to a higher cumulative incidence of stroke (IPTW HR 1.33, 95% CI 1.21–1.47, P < 0.001; Supplementary Material, Fig. S5 and Table S1). Random parameter θ = 0.33, P-value < 0.001). The proportional hazards assumption was held (P-value = 0.98). However, we also tested the association between stroke and diabetes 30 days after surgery (HR 1.38, 95% CI 1.25–1.53, P < 0.001; Random parameter θ = 0.33, P-value < 0.001), as shown in Fig. 3 and Table 2.

Diabetes was a risk factor for myocardial infarction up to 10 years (IPTW HR 1.16, 95% CI 1.07–1.25, P < 0.001, Fig. S6 and Table S1. Random parameter θ  =  0.40, P-value < 0.001), but the proportionality of the hazards was not confirmed (P-value <0.001). The sub-analysis excluding perioperative events demonstrated that the relationship between diabetes and acute myocardial infarction (AMI) is constant over time after 30 days (proportional hazard assumption test P-value =0.99) with an HR of 1.39, 95% CI 1.26–1.53 (Table 2 and Fig. 4).

Despite diabetes being a risk factor for cardiac and cerebrovascular events up to 10 years, it did not affect the incidence of repeat revascularization (IPTW HR 1.03, 95% CI 0.95–1.12; P-value 0.41; Supplementary Material, Fig. S7 and Table S1). Proportional hazard assumption was confirmed (test P-value =0.16). These data were also concordant excluding 30-day outcomes (Table 2 and Fig. 5).

The relationship between 30-day outcomes and diabetes is reported in Table 3. Diabetes did not affect mortality (IPTW OR 0.90, 95% CI 0.73–1.10, P-value = 0.31) and repeat revascularization (IPTW OR 0.79, 95% CI 0.42–1.49, P-value = 0.46), while it is related to lower rates of MACCE (IPTW OR 0.67, 95% CI 0.60–0.76, P-value < 0.001), AMI (IPTW OR 0.60, 95% CI 0.51–0.70, P-value < 0.001) and stroke (IPTW OR 0.47, 95% CI 0.28–0.77, P-value 0.003).

DISCUSSION

This study of the PRIORITY Project examined the impact of diabetes on long-term outcomes following isolated CABG. The main findings of this study were as follows: (i) the significant relationship between diabetes and worse outcomes up to 10 years after isolated CABG; (ii) the discordant effect of diabetes on long-term incidence of AMI and repeat revascularization; (iii) in our cohort, diabetes did not influence 30-day mortality or repeat revascularization and was associated with lower rates of 30-day MACCE, stroke, and AMI.

The relationship between diabetes and outcomes after CABG is multifaceted. Diabetes raises the risk of cardiovascular complications, such as wound infections and sternal dehiscence, which can prolong hospital stays and increase the number of hospital readmissions for heart-related reasons [13, 14]. Furthermore, multiorgan structural and functional changes linked to an elevated risk of IHD can result from chronic diabetes macro- and microvascular changes [15]. However, there is ongoing debate and inconsistent data about the effect of diabetes on long-term results following CABG. Diabetes has been related to lower long-term survival [5, 16], but this finding is not supported by other studies [17]. The multifarious variety of consequences is also more noticeable for MACCE, even though diabetic multiorgan chronic changes are a long-term risk factor for neurological, cardiac and renal complications in the general population [5]. The heterogeneity of study designs and different methodologies used to examine the association might be responsible for these discordant results. Although insulin-treated diabetes has been correlated to outcomes that are opposite to those of non-insulin therapy [16], it is impossible to differentiate between treatment types in several cohorts, as the PRIORITY one. Some authors have linked the greater incidence of all-cause death to factors other than cardiac mortality [5, 16]. Once more, most of the studies are retrospective, and the absence of balancing techniques may be an additional bias influencing the findings [18]. Aside from the drawback of using a more precise definition of diabetes, our results provide strong evidence for the long-term detrimental effects of diabetes on MACCE.

Interestingly, there is no discernible rise in long-term repeat revascularization in the diabetic cohort despite a higher long-term incidence of myocardial infarction. Graft deterioration may be caused by diabetic micro and macroangiopathy, raising the risk of recurrent revascularization [19, 20]. However, as our results also suggest, these experimental ultrastructural damages of grafts do not appear to be associated with an increased risk of macroscopic graft failure in diabetic patients undergoing CABG [21, 22]. The discrepancy between the absence of macroscopic graft failure and ultrastructural damage draws attention to native coronary microangiopathy, which can support the finding of a higher long-term incidence of AMI without requiring further revascularization [23, 24]. Endothelial dysfunction, diabetes-related hypercoagulability, and the progression of native vessel disease may more specifically target the coronary microvasculature, which is not addressed by surgical revascularization [9]. The decreased myocardial flow reserve in angiographically normal coronary arteries is related to glycaemic control but not to insulin resistance or lipid fractions, confirming that coronary microangiopathy is most likely associated with chronic glycaemic control in diabetic patients [9, 15].

In our cohort, perioperative outcomes did not align with the long-term detrimental effect of diabetes on MACCE. Confirming earlier findings, diabetes did not appear to be a risk factor for 30-day repeat revascularization or 30-day all-cause death [5, 17]. The lower incidence of 30-day outcomes in diabetic patients is somewhat difficult to interpret. The broad definition of diabetes, the administrative nature of follow-up data, and the potential risk of confounding by indication are some of the study’s inherent shortcomings that may be reflected in this. The original prospective studies’ definition of diabetes does not allow for assessing differences in treatment by disease severity or subtype, introducing the risk of “confounding by severity”. Additionally, even though the baseline risk profile is worse in the diabetic cohort, there may be confounding by indication due to the lack of data on the extension of coronary disease, target coronaries and revascularization strategies because diabetes may shift indication from PCI to surgery in more favourable anatomical features with intrinsic perioperative lower risk. The lower incidence of perioperative outcomes in the diabetic cohort might prompt a discussion on ischaemic preconditioning in diabetic patients. According to recent data, diabetic patients may experience this protective cardiac phenomenon more effectively than non-diabetic patients [25, 26], potentially offering a benefit in terms of myocardial protection during the perioperative phase.

Limitations

Our study is limited by its observational nature. The risk of selection bias could still be present for regressors that were not available in the databases. No comprehensive data on clinical history, coronary artery disease, surgical strategies, intraoperative details, postoperative complications or the expertise of surgeons or centres was gathered for the IT-CABG and IT-CABG II, which were intended for a nationwide quality assessment of healthcare providers based on administrative data and clinical comorbidities primarily defined by EuroSCORE.

Moreover, even if the analyses have been IPT-weighted and adjusted, the risk of bias is inflated because the 2 datasets were collected over separate periods, and the aged data cannot represent more recent cohorts and practices. Therefore, it is necessary to consider the hypotheses generated from this discussion in the framework of an observational study. However, the study was designed to focus on a ‘real-world’ countrywide experience and the relationship between diabetes and long-term outcomes after CABG. This study's strength is its access to a sizeable dataset that includes over 10 000 patients with full administrative follow-up. Lastly, long-term data were collected in a ‘blind fashion’, given its independence from participating centres and clinicians.

CONCLUSIONS

As expected, we confirmed that diabetes is a risk factor for cardiac and cerebrovascular events up to 10 years after CABG. The long-term risk of MACCE, death and stroke was greater in the diabetic population who received CABG. Diabetes may have a greater detrimental impact on microvasculopathy than grafts, as evidenced by the difference between its effects on myocardial revascularization and long-term myocardial infarction.

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

FUNDING

The PRIORITY study was supported by a grant from the Italian Ministry of Health for the finalized research (grant no: WFR GR-2013–02359264).

Conflict of interest: Fabio Barili reports receiving consulting fees from Abbott, outside the present work.

DATA AVAILABILITY

Data available on request. The data underlying this article will be shared on reasonable request to the corresponding author.

Author contributions

Fabio Barili: Conceptualization; Formal analysis; Funding acquisition; Methodology; Writing—review & editing. Nicolò Vitale: Conceptualization; Investigation; Methodology; Writing—original draft. Paola D'Errigo: Conceptualization; Formal analysis; Funding acquisition; Methodology; Writing—review & editing. Francesco Porcedda: Data curation; Formal analysis; Methodology; Writing—original draft. Francesco Pollari: Investigation; Resources; Supervision; Writing—review & editing. Giovanni Baglio: Methodology; Supervision; Writing—original draft; Writing—review & editing. Andrea Daprati: Validation; Visualization; Writing—review & editing. Gabriella Badoni: Supervision; Writing—review & editing. Giorgia Duranti: Formal analysis; Methodology; Writing—original draft; Writing—review & editing. Francesco Donatelli: Resources; Supervision; Writing—original draft; Writing—review & editing. Alessandro Parolari: Methodology; Project administration; Supervision; Writing—original draft; Writing—review & editing. Stefano Rosato: Data curation; Formal analysis; Investigation; Methodology; Writing—review & editing

Reviewer information

European Journal of Cardio-Thoracic Surgery thanks the anonymous reviewers for their contribution to the peer review process of this article.

REFERENCES

ABBREVIATIONS

ABBREVIATIONS
 
• AGENAS

  Italian National Program for Outcome Evaluation

 
• AMI

  Acute myocardial infarction

 
• BITA

  Bilateral internal thoracic arteries

 
• CABG

  Coronary artery bypass grafting

 
• EACTS

  European Society of Cardiothoracic Surgery

 
• HR

  Hazard ratio

 
• IHD

  Ischaemic heart disease

 
• IPTW

  Inverse probability of treatment weight

 
• MACCEs

  Major adverse cardiac and cerebrovascular events

 
• PCI

  Percutaneous coronary intervention

 
• PRIORITY

  PRedictIng long-term Outcomes afteR Isolated coronary arTery bypass surgerY

 
• SICCH

  Italian Society of Cardiac Surgery

Author notes