STRENGTHS AND LIMITATIONS OF THIS STUDY
-
This study was conducted at the largest national-level cardiology centre in China, with a focus on the unique cohort of ischaemic heart failure patients with concurrent diabetes.
-
A limitation of this research is its retrospective cohort design, which may introduce selection bias and recall bias.
-
Another limitation was the presence of approximately 9% loss to follow-up within the cohort.
Introduction
It is estimated that more than 415 million adults have diabetes mellitus (DM) globally, which might rise to 642 million by 2040.1 Therefore, the prevalence of DM and its complications pose a major global health threat. Coronary heart disease (CHD) is the most common complication and the primary cause of mortality.1 And multivessel disease, diffuse lesions and segment obstructive lesions were observed more frequently in diabetic CHD.2 Thereby, diabetic CHD had more chance to progress to ischaemic heart failure (HF).3–5
The ischaemic HF patient concomitant with DM has worse outcomes compared with the ischaemic HF without DM. The percutaneous coronary intervention (PCI) remained an effective approach to alleviate the symptoms and prove the prognosis. Patients with complex coronary artery lesions and left ventricular ejection fraction (LVEF)-reduced HF undergoing PCI may reverse left ventricular remodelling and improve clinical prognosis.6 7 Patients with ischaemic HF subjected to revascularisation reduced the rate of cardiac mortality by 61% in the 5 years.7 However, few reliable markers or scoring systems predict the major cardiovascular events (MACE) after PCI in patients with ischaemic HF and DM.
D-dimer to albumin ratio (DAR) has emerged to be a prognostic marker with the outbreak of COVID-19. Recently, its utility in cardiovascular disorders is increasingly being recognised.8–12 For instance, Yuan et al found a significant association between serum albumin, D-dimer levels and non-valvular atrial fibrillation, with D-dimer levels being inversely correlated with albumin levels.13 Two retrospective studies investigated the diagnostic value of albumin and D-dimer in pulmonary embolism and found that lower plasma albumin and higher D-dimer were observed in patients with pulmonary embolism.12 14 Moreover, the age-D-dimer-albumin score has been used to predict acute myocardial infarction (MI) patients.8 However, the association between DAR and ischaemic HF concomitant with DM has never been investigated before.
Ischaemic HF combined with DM affects a substantial proportion of patients and has a worse prognosis than ischaemic HF patients without DM. Early intervention is essential for managing this medical issue. However, currently, reliable indicators to predict the prognosis of ischaemic HF patients with DM after PCI are lacking. Therefore, it is imperative to identify a reliable predictor. One such potential predictor is DAR, which has not been explored for its predicting ability for MACE after PCI in HF patients with DM. This knowledge gap needs to be addressed. Therefore, our study aims to investigate DAR as a predictor of MACE after PCI in patients with ischaemic HF concomitant with DM.
Methods
Study population
This is a single-centre observational retrospective study. The patient inclusion flow chart is represented in figure 1. We screened a total of 3707 patients with ischaemic HF undergoing elective PCI between June 2017 and June 2019. The diagnosis criteria of ischaemic HF were the following15: (1) congestive HF, left ventricular failure, cardiac insufficiency, diastolic HF or HF, unspecified according to the International Classification of Diseases 10th revision (I50.001, I50.106, I50.90, I50.919, I50.905 or I50.911) and (2) concomitant multiple vessel disease (coronary artery stenosis >50% in ≥2 vessels or left main (LM)). The diagnosis criteria of DM followed the 1999 criteria of the WHO. Exclusion criteria were as follows: (a) LVEF≥50% (n=254); (b) history of coronary artery bypass graft (n=56); (c) acute MI (n=239); (d) miss data on D-dimer and albumin (n=36); (e) pulmonary thromboembolism and deep venous thrombosis (n=11); (f) exclude patients with malnutrition, hepatocirrhosis, tuberculosis, hyperthyroidism, any kind of cancer, acute inflammatory states (pathogen infections, autoimmune inflammatory diseases) (n=47) and (g) the individuals lost to follow-up (n=334). After these exclusions, a total of 1021 patients were available at inclusion.
Flow chart of inclusion and exclusion. CABG, coronary artery bypass graft; DM, diabetes mellitus; LVEF, left ventricular ejection fraction; MACE, major adverse cardiovascular events; MVD, multiple vessels disease; PCI, percutaneous coronary intervention.
Definition and data collection
The hospital information system provided the baseline data of demographics, vital signs, body mass index, comorbidities, medical history, New York Heart Association (NYHA) class, laboratory parameters, echocardiography, angiographic lesion characteristics, procedural results and medication use. Among them, the comorbidity of renal insufficiency was defined as the estimated glomerular filtration rate (Modification of Diet in Renal Disease Equation) <60 mL/min×1.73 m2. D-dimer was analysed by the Mindray C3100 Auto Coagulation Analyzer with immunoturbidimetry, while albumin levels were detected using the Roche cobas c701 automated chemistry analyzer with the standard Methyl bromide green method. We collected blood samples within 48 hours after in-patient admission. Angiographic lesion characteristics included LM disease, three-vessel disease, chronic total occlusion (CTO), diffuse lesion and Synergy between PCI with Taxus and Cardiac Surgery (SYNTAX) (www.syntaxscore.com).16 The lesion characteristics were defined as follows: (1) LM disease: an angiographically estimated stenosis >50% or a fractional flow reserve <0.80 in the LM coronary artery ostium, mid-shaft or distal bifurcation. (2) Three-vessel disease: more than two main coronary branches (vessel diameter ≥2 mm) with the extent of stenosis ≥50%. (3) CTO lesion: lesion with complete obstruction that thrombolysis in MI flow grade 0 lasting longer than 3 months. It was determined by the medical history or coronary angiogram results. The duration of the occlusion was determined by the interval from (1) the last occurrence of acute coronary syndromes consistent with the location of the occlusion; (2) the first episode of effort angina or (3) the earliest findings of previous coronary angiographic confirmed occlusion and (4) diffuse lesion: a single stenotic lesion with a length of ≥20 mm. PCI indications and strategies were determined and performed by at least two senior cardiologists and followed the current practice guidelines in China.17 The angiographic data were recorded in a standardised operation recording system. All these baseline data were extracted automatically by technical support engineers in the Department of Information.
Grouping
We first divided subjects into tertiles according to DAR. All patients were sent into three groups by DAR tertiles: tertile 1, under 1.10; tertile 2, 1.10–1.28 and tertile 3, higher than 1.28. The DAR was defined as ln[D-dimer(ng/mL)]/albumin(g/dL). After PCI, the record lists were transferred to the outpatient follow-up centre for the start of follow-up.
Main outcome measurement
The MACE was the primary measured outcome. Moreover, the secondary endpoint was any of the components of the defined MACE, including all-cause mortality, non-fatal MI and any revascularisation. The trained physicians followed up with the participants at 3, 6, 9, 12 and 24, by using telephone questionnaires or tracking the rehospitalisation records until the occurrence of MACE or the end of the study period (June 2022). The most severe adverse outcome (all-cause mortality>non-fatal MI>any revascularisation) was selected for analysis in patients with multiple adverse outcomes during follow-up. Only the first occurrence was analysed when a single event occurred more than once.
Data analysis
Statistical analyses were performed with Stata software (StataCorp, V.15.0) and R software (R-project, Vienna, Austria, V.4.2.1). The cases with missing data of albumin and D-dimer were removed from the dataset. For the numerous variables that followed a normal distribution, the means were used to impute any missing values. For numerous variables that exhibited an abnormal distribution, medians were used to fill in the missing values. And there were no missing data in the categorical variables. Continuous variables presented as mean±SD for normally distributed data and median (25th percentile, 75th percentile) for abnormally distributed one. Categorical variables were expressed as numbers and percentages. Numerical variables comparisons among three tertiles were performed by analysis of variance or the Kruskal-Wallis test. χ2 tests were applied for comparisons of categorical data among different DAR tertiles. The significance level was 0.05.
The DAR tertiles’ differential incidence of MACE (all-cause mortality, non-fatal MI and any revascularisation) were determined by the cumulative hazards plots and log-rank test. We used backward stepwise regression analysis to identify the potential confounders that have an impact on MACEs. Multivariable Cox proportional-hazard models were used to evaluate the independent predictors of 36-month primary and secondary endpoints after adjusting for potential confounders identified by backward stepwise regression. Restricted cubic spline models were used to identify non-linear correlations between DAR and risk of MACE based on the Cox proportional hazards model with four knots (5th, 35th, 65th and 95th). Receiver operating characteristic (ROC) curves and corresponding area under the curve (AUC) of DAR were generated for MACE and its components. Stratified analyses were designed to evaluate the independency of DAR in subgroups of sex, age (≤65 years and >65 years), obesity, hypertension, hypercholesterolaemia, renal insufficiency status, angiographic lesion characteristics and target vessels.
Results
Patient characteristics
From June 2017 to June 2019, 3707 patients with HF and multiple vessel disease undergoing PCI were screened. After exclusion, the data of 1021 patients (ischaemic HF concomitant with DM) were available for analysis (figure 1). The baseline characteristics classified by the tertiles of DAR are summarised in table 1. The baseline average age was 61.3 years. A total of 24.5% of participants were women. The average LVEF was 41.1%±6.1%. The blood glucose was 9.03±3.48 on average. Tertiles 2 and 3 exhibited a higher D-dimer level and a lower albumin level than in tertile 1 (p<0.001). Among the three groups stratified by DAR tertiles, the incidence of MACE was 26.7%, 40.6% and 51.5% in tertiles 1, 2 and 3, respectively (p for trend <0.001). The rate of MACE, all-cause mortality, non-fatal MI and any revascularisation, among the three groups, was demonstrated in online supplemental table 1.
Supplemental material
Baseline characteristics of the ischaemic heart failure patient combined with diabetes undergoing PCI
Cumulative hazards curve and log-rank test
The cumulative hazard curves among tertiles for MACE, all-cause mortality, non-fatal MI, and any revascularisation are shown in figure 2. The incidences of MACE in tertile 2 and 3 were significantly higher than in tertile 1 (log-rank test, p<0.0001) (figure 2A). Cumulative hazards curve curves also showed that the participants with a higher DAR have a significantly higher all-cause mortality probability and any revascularisation probability (log-rank test, both p<0.0001) (figure 2B and D). However, there was no significant difference in non-fatal MI probability among the three tertiles according to the log-rank test (p=0.075) (figure 2C).
Cumulative hazards curves. The cumulative hazards curves for MACE (A), all-cause mortality (B), non-fatal MI (C) and any revascularisation (D) according to DAR tertiles. DAR, D-dimer to albumin ratio; MACE, major adverse cardiovascular events; MI, myocardial infarction.
Cox proportional hazard models
The crude model (model 1) demonstrates that DAR tertiles 2 and 3 are associated with a significantly increased incidence of MACE by 1.73-fold (95% CI 1.33 to 2.25) and 2.30-fold (95% CI 1.79 to 2.96), respectively, when compared with tertile 1 (both p<0.001). Additionally, DAR also proves to be an independent predictor of all-cause mortality and any revascularisation in model 1. In model 2, after adjusting for age and sex, DAR remains an independent factor correlated with the occurrence of MACE in tertiles 2 and 3 (HR 1.70; 95% CI 1.30 to 2.22, HR 2.28; 95% CI 1.75 to 2.98, respectively, both p<0.001). Moreover, DAR remains a significant independent risk factor for all-cause mortality (tertile 2: HR 1.96, 95% CI 1.32 to 2.89; tertile 3: HR 2.46, 95% CI 1.66 to 3.65, both p<0.01) and any revascularisation (tertile 2: HR 1.55, 95% CI 1.03 to 2.35; tertile 3: HR 2.09, 95% CI 1.39 to 3.15, both p<0.05). Additionally, in model 3, even after further adjustments for age, sex, NYHA class, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, heart rate, total cholesterol, total bilirubin, direct bilirubin, glucose, highly sensitive C reactive protein (hs-CRP), brain natriuretic peptide, LVEF, left ventricular end-diastolic dimension, three-vessel disease, diffuse lesion, CTO lesion, SYNTAX, target vessel of a right coronary artery (RCA), ACE inhibitor (ACEI), angiotensin receptor blocker (ARB), sacubitril/valsartan, ezetimibe, DDP4 depressor and sodium-glucose cotransporter 2 (SGLT2) inhibitor, all three DAR tertile groups continue to exhibit a significant independent increase in MACE risk. The calculated MACE HRs with 95% CIs for tertiles 2 and 3 are 1.82 (1.37 to 2.42) and 1.74 (1.28 to 2.36), respectively. Furthermore, DAR remains an independent predictor of all-cause mortality model 3. The calculated all-cause mortality HRs with 95% CIs for tertiles 2 and 3 are 2.04 (1.35 to 3.11) and 1.89 (1.20 to 2.98), respectively (table 2). The HRs of each component of DAR were also assessed using Cox regression. Both in model 1 and model 3, DAR demonstrates the highest HRs for MACE or all-cause mortality when compared with D-dimer or albumin alone (online supplemental table 2). We also assess the efficacy of continuous DAR in comparison with other well-established risk factors in model 3. Regarding these risk factors, NYHA class (HR 1.17, 95% CI 1.0095 to 1.36), LVEF (HR 1.015, 95% CI 1.011 to 1.025), hs-CRP (HR 1.018, 95% CI 1.0051 to 1.032), total cholesterol (HR 1.42, 95% CI 1.27 to 1.609) and glucose (HR 1.058, 95% CI 1.030 to 1.087) identified as independent risk factors. However, their HR values are lower than that of DAR (1.84). Therefore, in comparison to these conventional risk factors, a higher DAR may be more indicative of MACE. DAR potentially holds predictive value in patients with ischaemic HF and concomitant diabetes (online supplemental table 3).
Supplemental material
Supplemental material
The association between DAR
The optimal cut-off and ROC curves of DAR
The restricted cubic spline curve was used to visualise the correlation of DAR with MACE based on Cox proportional hazard model 3. DAR showed an overall non-linear positive relationship with the risk of MACE. The curve increased rapidly with a steep slope between 1.2 and 1.5 of DAR. The curve outside of the interval was relatively flat. It indicated that DAR might have higher discrimination for risk of MACE when DAR was 1.2–1.5. The optimal clinical cut-off for DAR is 1.2 because a DAR higher than 1.2 usually means a rapidly increasing risk of MACE (online supplemental figure 1). We have ROC curves of DAR for MACE and its components (all-cause mortality, non-fatal MI and any revascularisation). The results revealed the AUC for primary and secondary outcome events: 0.729 for MACE, 0.735 for all-cause mortality, 0.557 for non-fatal MI and 0.616 for any revascularisation. The results indicate that the DAR showed a relatively better performance in the outcomes of MACE and all-cause mortality (online supplemental figure 2).
Supplemental material
Supplemental material
Stratified analysis for DAR and MACE
As shown in figure 3 and online supplemental table 4, crude (model 1) and adjusted (model 3) stratified analysis was performed for subgroups of patients. The multivariable regression analysis that was adjusted for heart rate, NYHA class, hs-CRP, glucose, total bilirubin, direct bilirubin, LVEF, left ventricular end-diastolic dimension, ezetimibe, ACEI, ARB, sacubitril/valsartan, SGLT2 inhibitor and DPP4 depressor. Fully adjusted stratified analyses verified DAR as an independent predictor of MACE regardless of sex, age, the status of hypertension and hypercholesterolaemia. However, DAR remained an independent predictive factor only for those without obesity (HR 3.55, 95% CI 2.11 to 5.97) and those with coexisting renal insufficiency (HR 3.41, 95% CI 2.01 to 5.81). For the angiographic lesion characteristic, multivariable stratified analyses also verified the association of DAR with increased MACE risk in patients with (HR 3.11, 95% CI 1.25 to 7.76) or without (HR 2.49, 95% CI 1.56 to 3.96) LM disease, with (HR 2.47, 95% CI 1.51 to 4.04) or without(HR 2.30, 95% CI 1.07 to 4.94) three-vessels disease. For the results of the PCI procedures, higher DAR still showed an independent increase in the risk of MACE regardless of whether the participants had total revascularisation or which interfered target coronary vessel (left descending artery, left circumflex artery and RCA).
Supplemental material
Forest plots. The forest plots of crude stratified model based on univariable Cox regression (A) and adjusted stratified model based on multivariable Cox regression (B) for subgroups of patients. CTO, chronic total occlusion; F, female; LAD, left anterior descending artery; LCX, left circumflex artery; LM, left main; M, male; N, no; RCA, right coronary artery; Y, yes.
Discussion
Ischaemic HF accounts for the most significant proportion of newly diagnosed cases of HF each year. The patients with both ischaemic HF and DM are a large group and have become an unneglectable social burden that calls for more attention.15 16 The key findings of our study are that higher DAR was independently associated with MACE, especially with all-cause mortality, in the ischaemic HF patients with DM undergoing PCI. Besides, sex, age, hypertension, hypercholesterolaemia, total revascularisation and any interfered vessel did not affect the correlation. The strengths of our study are that we first applied the DAR predictive value to the ischaemic HF and DM cohort. Furthermore, we present novel data suggesting DAR was an independent predictor of long-term MACE after PCI in ischaemic HF patients with DM. DAR showed a good predictive performance in predicting MACE. There are some limitations to our study. This study is a single-centre study and there is about 9% loss-to-follow-up rate in this cohort, which was caused by population mobile and custom factors. It raises potential selection bias. And this is also a retrospective study, the incidence of MACE might be influenced by recall bias. Further data from additional centres are necessary in the future to enhance the robustness and applicability of our findings.
The patients with both HF and DM have worse clinical outcomes than HF patients without DM. DM independently increases the risk of death in patients with HF.17 Notably, ischaemic HF with DM has a poor outcome than non-ischaemic HF combined with DM,18 including mortality, the incidence of first hospitalisation,19 20 more occurrence of rehospitalisation21 and worse health-related quality of life.22 23 There are some prognosis factors of HF combined with DM. hs-CRP is one of the outcome predictors. Compared with none rehospitalisation cohort, hs-CRP was higher in patients with rehospitalisation for HF.21 Besides, some clinical parameters, such as red cell distribution width and heart rate related to the prognosis.24–26 Growth differentiation factor-15 and soluble ST were also biomarkers of adverse outcomes for diabetes and cardiovascular diseases.27–29 However, their application has been limited for now because of their difficulty obtainment in clinical practice. Nonetheless, a reliable marker was barely developed for predicting the risk of MACE after PCI therapy in ischaemic HF and DM comorbidity patients. According to the early literature, DAR was considered a novel diagnostic and prognostic biomarker in cancer and organ torsion.10 Since the COVID-19 pandemic, the DAR has been getting increasing attention for its prognostic value on mortality and thrombosis.8 9 30 DAR has recently shown the latent capacity to apply in cardiovascular disease. Several studies demonstrated that the lower albumin level in the HF patient is associated with advanced HF and MACE.31 32 On the opposite, the higher level of D-dimer had a poor outcome, including in-hospital mortality, long-term mortality and cardiovascular death.33–35 Yuan et al found that albumin and D-dimer were negatively related in the same cohort. In addition, both were significantly associated with non-valvular atrial fibrillation.13 It also found that in the same pulmonary embolism cohort, the median albumin levels were lower while D-dimer was higher than in patients without pulmonary embolism. Both of them have a diagnostic value.11
In our study, DAR showed a good performance in predicting MACE and mortality after PCI in ischaemic HF combined with patients with DM. We inferred that DAR provides risk assessment of both underlying disease (ischaemic HF and DM) and PCI procedures combined in this population via inflammation mechanisms. D-dimer and albumin are both related to systematic inflammation. Current evidence suggests that diabetes, multivessel disease, HF and PCI procedures are associated with the inflammatory process. First, inflammation plays an important role in diabetes-associated cardiovascular events.36 Synergistic effects of diabetes and inflammation promote atherothrombosis, cardiovascular morbidity and mortality.37 Elevated glucose contributes to enhanced inflammation, atherosclerosis and cardiovascular events.38 Second, it has long been observed that HF is associated with systemic inflammation. Numerous studies have validated that inflammation is involved in the pathogenesis of HF and associated with HF progression and prognosis.39–41 Last but not least, vascular inflammation is an underlying central mechanism determining the response to PCI or stenting. A cohort study published in the European Heart Journal indicated that persistent high inflammation status is observed frequently in patients undergoing PCI. In these patients, significantly higher all-cause mortality and MI rates are observed at 1-year follow-up.42 Besides, most anti-inflammatory pharmacotherapies are effective in improving outcomes for PCI.43
Apart from inflammation, there are some other potential mechanisms: (1) D-dimer can be elevated in HF because of endothelial dysfunction, leading to microthrombus formation and resolution within the circulation system. D-dimer was related to thrombosis, which contributes to the adverse events of HF.33–35 44 (2) The lower albumin in HF aggravates congestion, leading to lower intravascular colloid osmotic pressures, an increase of systematic oxidant stress and the tendency to infections.32 45 46 Therefore, albumin is a summation of several risk factors of poor outcomes. The DAR leverages early laboratory observations that the D-dimer is positively associated with MACE, while the albumin is negatively associated. The ratio amplified its MACE predictive potential.
Stepwise regression was used to identify and adjust for potential confounders. Diffuse lesion, three-vessels disease and CTO lesion remained the most decisive confounder increasing the risk of MACE. These findings were consistent with previous studies47–49 and might cause the predication deviation in the subgroup of the diffuse lesion and CTO lesion in the stratified analysis. Furthermore, some medications used lowered the incidence of MACE, such as sacubitril/valsartan was the most vital protective factor and lowered 37% the risk of MACE, which was compatible with the previous study in the HF and DM cohort.50 DAR remained an independent predictor in the multivariable Cox regression model after adjusting these confounders.
Conclusion
Higher DAR was independently associated with MACE and all-cause mortality after PCI in the ischaemic HF patients with DM.
Data availability statement
Data are available on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and the research ethics committee of the Beijing Anzhen Hospital in China approved this study and all the patients signed the informed consent (registry number, 2022235X). Participants gave informed consent to participate in the study before taking part.
Acknowledgments
The authors are grateful to the all participants in this study.
This post was originally published on https://bmjopen.bmj.com