Introduction
Acute coronary syndrome (ACS) is the leading cause of death worldwide. Although the prognosis of ACS patients has been improved in the past 18 years, there is a lot of room for further improvement, as the ACS patients clearly have a higher risk for a recurrent cardiovascular events compared with the corresponding general population [
1]. Biomarkers are good for prognosis evaluation and risk stratification of ACS patients. Cardiac troponins, CRP and NT- BNP are the main predictors for ACS patients. However, they still have some drawbacks as BNP levels are also found to be significantly influenced by age, fluid loading and physical exercise [
2], and CRP is influenced by inflammation. Cardiac troponins are affected by reperfusion modalities [
3]. There is a need for more reliable biomarkers for the prognosis of ACS patients.
Hyperglycemia and newly-diagnosed diabetes mellitus (DM) are found in a large number of ACS patients and a strong predictor for the poor prognosis of these patients [
4‐
9]. HbA1c, which reflects average blood glucose concentrations over the previous 8–12 weeks, was shown to be a better predictor of prognosis following ACS than fasting and admission glucose [
10]. Recent studies proved that HbA1C is an independent predictor of mortality [
11], while other studies got opposite results [
12]. Furthermore, intensive glycemic control using insulin injection did not improve both short-term and long-term mortality in patients with ACS and DM [
13,
14]. In spite of current uncertainty, no previous systematic review has summarized the prognostic role of HbA1c playing in ACS patients. We therefore conducted the meta-analysis of prospective and retrospective cohort studies to evaluate if high levels of HbA1c are associated with increased in-hospital and short-term mortality in ACS patients.
Methods
Search strategy
To find potentially relevant studies, we searched articles in English referenced in Medline, Central and Embase from database inception until July, 2019. The following free texts or MeSH terms were used in search: Acute coronary syndrome, unstable angina, ST-elevation, ST-segment, STEMI, NSTEMI or myocardial infarction combined with hemoglobin a1c, HbA1c, glycated hemoglobin, glycosylated hemoglobin, haemoglobin a1c, glycated haemoglobin, glycosylated haemoglobin, glycohemoglobin a or glycohaemoglobin a. The reference lists of retrieved articles were reviewed for additional eligible studies. Eligibility and study selection
Two authors (Wenjun, Haining) independently reviewed the titles and abstracts of all articles retrieved, and selected those meeting the following criteria: (1) Based on prospective and retrospective cohort studies. (2) Conducted in adults aged 18 years or older. (3) Reporting HbA1c levels as exposure. (4) Reporting the in-hospital and short-term (less than 1 year) mortality. (5) Reporting the effect of the exposure in the form of a relative risk [risk ratio (RR), odds ratio (OR) or hazard ratio (HR)] and their 95% CI or the dead number and total number of each group. Cross-sectional and case–control studies were excluded.
Data extraction and quality assessment
Two investigators (Wenjun and Baotao) independently extracted the following information from eligible studies: first authors’ name, publication year, study name, country, age range and/or mean age (years), number of participants, exposure levels, RRs and their 95% CI as well as the alive and dead numbers for each exposure category. We selected the RRs from the model with the most comprehensive covariate adjustment, and used the Newcastle–Ottawa Scale to assess the quality of the studies included [
15]. Any discrepancies were resolved through discussion under supervision of a senior author (Minzhou).
Statistical analysis
We calculated the average RRs for the top compared with the bottom quantile of HbA1c levels using a fixed-effects model. We calculated the unadjusted RRs using original data published in the studies if RRs and their 95% CI were not available in the papers. The I2 statistic and Cochran’s Q test were used to evaluate heterogeneity. If there is significant heterogeneity (I2 > 50%, p < 0.1) among studies, we conducted sensitivity analysis or subgroup analysis to explore the sources of heterogeneity. Publication bias were assessed using funnel plots’ asymmetry and tested with Egger’s asymmetry test and Begg’s test (p < 0.05) when there were at least 10 studies. All analysis was conducted with Stata 13.0.
Discussion
25 clinical trials were included in this study involving 304,253 subjects. We found: (1) higher HbA1c levels contributed to increased in-hospital mortality in ACS patients; (2) in general, HbA1c was significantly associated with increased short-term mortality in ACS patients; (3) for ACS patients without DM history and without DM, higher HbA1c had an positive relationship with increased short-term mortality; (4) for ACS patients with DM and no specified metabolic status, HbA1c was failed to show positive prediction for short-term mortality.
Consistent to the review by Capes et al. [
42], we found positive relationship between admission hyperglycemia and in-hospital mortality in ACS patients for studies considering HbA1c as a categorial valuable. Although there was significant heterogeneity among studies which came from the one with the highest RR value, the association between HbA1c levels and in-hospital mortality remained significant after excluding it. As compared to other studies, the in-hospital mortality (14%) was higher in the top quantile of HbA1c, which might cause the great heterogeneity. For studies considering HbA1c as a continuous valuable, in fixed-effects model we found a positive relationship between HbA1c and in-hospital mortality but failed to prove it in the random-effects model. Besides the limited number of studies, the diverse valuables for adjusted in the multivariable analysis should also be taken into consideration. As reported by several independent groups, hyperglycemia is associated with impaired microvascular function and decrease coronary flow velocity even induce microvascular obstruction in ACS patients [
43‐
45]. No reflow after reperfusion therapy due to chronic hyperglycemia could give rise to in-hospital death in ACS patients. We did pooled analysis of all the studies of short-term mortality and found that HbA1c contributes to increased short-term mortality in ACS patients. However, there was significant heterogeneity among these studies, which was resolved through subgroup analysis according to metabolic status. We found that for ACS patients without known DM, HbA1c significantly increases short-term mortality. For these included studies, besides patients with normal glucose metabolic status, there was high incidence of patients with newly diagnosed DM which have not been recognized and treated [
46]. Therefore, this subgroup reflects the effect of HbA1c on short-term mortality independent of previous anti-diabetic drugs. However, consistent with Østergaard et al. [
47] and She et al. [
48], we found for patients with DM, it showed no significant effect of HbA1c on short-term mortality in ACS patients. Several causes should be taken into consideration. Firstly, there might be several confounding factors interfering the results, such as anti-diabetic drugs, modified lifestyle and better compliance to the doctors’ advice before and after heart attack. Secondly, as reported by Li et al. [
49], higher HbA1c were associated with less risk of myocardial injury following PCI in diabetic patients because of better energy supply. As a large number of ACS patients included in our meta-analysis received PCI therapy, this might account for the confounding factor in short-term prognosis. Last but not least, a limited number of patients with a relatively short follow-up should be taken into consideration, as HbA1c has been commonly recognized as a long-term predictor of mortality in ACS patients with DM [
50,
51]. Consistent with randomized controlled trials [
13,
14], these patients did not benefit from intensive glycemia control after ACS. The intriguing thing we found in this systematic review is that for ACS patients with definitely no DM, HbA1c is also showed to be a predictor for short-term mortality, which is consistent with the finding by Geng et al. [
52]. Overall, HbA1c is a predictor for in-hospital and short-term death in ACS patients. For DM and not specified patients, due to complicate confounding factors which might interfere the results, we found no positive relationship between HbA1c levels and short-term mortality. More studies regarding HbA1c as a continuous variable are needed to prove its effect on in-hospital mortality.
A number of biological mechanisms have been proposed to explain a potential causal association between HbA1c and increased mortality in ACS patients. Firstly, basic experimental study indicates that hyperglycemia inhibits cardiac stem cells from cardiac repair and angiogenesis in streptozotocin-induced diabetic mice undergoing AMI surgery [
53], which might account for the worse prognosis for ACS patients with elevated HbA1c levels. Secondly, higher HbA1c levels reflect over-glycosylation of intracellular protein in cardiac myocytes like CaMKII [
54]. This may give rise to sudden cardiac death due to fatal arrhythmia. Furthermore, as reviewed by Shah et al. [
55], chronic hyperglycemia causes decreased mitochondrial biogenesis, increased reactive oxygen species (ROS) production and post-translational protein modifications, which eventually leads to malignant cardiac remodeling. Last but not least, recent experiment found that baseline plasma fibrinogen is associated HbA1c levels in ACS patients [
56]. Fibrinogen is an important part of coronary thrombosis, which is consistent with that higher HbA1c levels were significantly related to increased refraction in ACS patients [
25,
27,
41].
The present meta-analysis has some inevitable limitations. On one hand, for the categorial variables, we only took the highest and the lowest quantiles into consideration, which might exaggerate the effect of HbA1c. On the other hand, while we extracted the adjusted RRs if they are available, as we cannot get individual participant data, the confounding factors such as baseline characterization and therapy after admission might still interfere the results. Thus, the result should be interpreted with caution in clinical practice.
While acknowledging the inherent limitations of original data, we insist there are merits and clinical significance of our meta-analysis. To our knowledge, this is the first systematic review regarding the predictive value of HbA1c for short-term mortality in ACS patients. Although there is a similar review regarding STEMI patients [
57], it did not explain the heterogeneity explicitly. As unstable angina and NSTEMI share the same pathological process as STEMI, we took all the ACS patients into account in this systematic review. Due to the increased in-hospital mortality in patients with elevated HbA1c, people at high risk of ACS should take plasma glucose and HbA1C under control. According to recent reports, there are substantial ACS patients with unknown DM at admission [
34,
41], which accounts for the high risk of short-term mortality in ACS patients without known DM. It is intriguing to note that elevated HbA1c is a predictor of short-term mortality in ACS patients without DM, which instructs ACS patients especially pre-DM patients to control the plasma glucose and HbA1c levels strictly [
13,
14]. While there are randomized controlled trials indicating that ACS patients with DM did not benefit from the intensive glycemia control, we recommend people at high risk of ACS without (known) DM to monitor and control HbA1c levels who might benefit from it according to our meta-analysis.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.