The data analyzed in this study were from the Acute Coronary Syndrome Quality Improvement in Kerala (ACS-QUIK) Study, which was available on the National Heart, Lung and Blood Institute website with reasonable application (https://biolincc.nhlbi.nih.gov/studies/acs_quik/). The rationale and main result of the ACS-QUIK Study have been published previously [13, 14]. In brief, the ACS-QUIK Study was a cluster-randomized, stepped-wedged clinical trial conducted in 63 hospitals in Kerala, India, from November 10, 2014 to November 9, 2016. The aim of this study was to assess whether a locally adapted quality improvement tool kit could improve the process of care measures and clinical outcomes for patients with acute myocardial infarction. The ACS-QUIK Study was approved by the ethics committees of local, national and international agencies and approved by the Indian Health Ministry Screening Committee. All participants or their representatives provided written informed consent to participate in the trial. Among 21,374 patients with acute myocardial infarction enrolled in this trial, the locally adopted quality improvement kits did not reduce the incidence of 30-day MACEs compared with conventional care.

This analysis was to evaluate the effect of the glucose level on admission on the incidence of 30-day MACEs in patients with acute myocardial infarction at ACS-QUIK Study and to assess the difference in outcome between STEMI and NSTEMI patients. After 7976 participants without glucose on admission were excluded, we finally included 13,398 participants in this analysis and divided them according to the glucose quintiles [Q1 ($\le$5.3), Q2 (5.4–6.3), Q3 (6.4–7.8), Q4 (7.9–10.6), Q5 ($\ge$10.7), with Q2 (5.4–6.3) as reference]. The flowchart of analysis of this study is shown in Supplementary Fig. 1.

Other baseline parameters included demographic data (age, gender, weight, smoking or tobacco), examination at admission (systolic blood pressure (SBP), heart rate, high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), triglycerides), left ventricular ejection fraction (LVEF) and LVEF category, prior comorbidities [hypertension, peripheral artery disease (PAD), diabetes, prior transient ischemic attack (TIA) or stroke], type of myocardial infarction (NSTEMI, STEMI), Killip class at admission and cardiac status, medication at admission (Beta Blocker, antiplatelet), symptom onset to arrival time, and percutaneous coronary intervention (PCI).

The study outcome was 30-day MACEs, including morality, stroke and reinfarction. The diagnostic criteria for reinfarction were defined according to the Third Universal Definition of Myocardial Infarction [15].

The categorial variables were described statistically by frequency and percentage, and the continuous variables were described statistically by mean $\pm$ standard deviation (for normal distribution) or median (P25, P75) (for skewness distribution), respectively. Analysis of variance (ANOVA) or nonparametric test was used for testing inter-group differences of continuous variables, and chi-square test or Fisher test was used for categorical variables. The association of 30-day MACEs with the baseline glucose quintiles was assessed by three logistic regression analysis models. All covariates that might influence short-term outcomes in patients with acute MI were included in the analysis. The diagnosis of multicollinearity in the covariates included in the logistic model were measured by the variance inflation factor (VIF) (Supplementary Table 1). If VIF $\ge$5, multicollinearity existed among the covariates and the corresponding variables were eliminated. Covariates with a p value of the regression coefficient less than 0.1 were adjusted in the full model (Supplementary Table 2). According to the STROBE statement [16], model 1 was unadjusted model, model 2 was minimally adjusted for intervention, age, sex and SBP, and model 3 was fully adjusted for intervention, age, sex, SBP, weight, heart rate, HDL-C, LDL-C, triglyceride (TG), smoking or tobacco, diabetes, hypertension, PAD, MI Type, time from symptom onset to arrival, prior TIA or stroke, cardiac arrest at admission, LVEF category, PCI, antiplatelet, beta blocker. The same analysis procedures were used to assess the relationship between MI type and 30-day MACEs, adjusting for all the covariates mentioned above, including glucose in model 3. We also used a generalized additive model (GAM) to visualize the dose-response relationship between glucose levels at admission and the risk of 30-day MACEs and stratified the dose-response relationship by MI type (covariates in model 3 were adjusted). We also analyzed the interaction between glucose levels at admission and prespecified subgroups on the risk of 30-day MACEs. All analyses were performed using the statistical software package R version 4.0.0 (The R Foundation; http://www.R-project.org). Statistical significance was set at p 0.05.

Table 1 shows the baseline characteristics and crude outcomes across the glucose quintiles. As expected, the higher glucose quartile was associated with a higher incidence of 30-day MACE and a higher prevalence of diabetes. Participants in the higher glucose quartile had a higher proportion of females, a higher SBP and heart rate at admission, higher Killip class (II~IV), a lower rate of smoking or tobacco and a higher prevalence of hypertension, PAD and stroke, as well as a lower LVEF and a higher incidence of cardiac arrest than participants in the lower glucose quartile. There were significant inter-group differences in weight, HDL-C, triglyceride MI type and PCI therapy, while there were no inter-group differences in medication (beta blocker, antiplatelet) at admission and time from symptom onset to arrival.

Supplementary Table 1 presented baseline characteristics and crude outcomes according to MI type (NSTEMI vs. STEMI). There were significant inter-group differences in all variables except HDL-C. Patients with STEMI had a higher risk of 30-day MACEs in the fully adjusted model [OR (95% CI): 1.41 (1.11, 1.80), p = 0.005] (Supplementary Table 2).

As shown in Table 2, comparing with the reference (Q2, 5.4–6.3), participants with the highest quintile of glucose level were associated with increased risk of 30-day MACEs in the fully adjusted logistic regression model 3 [OR (95% CI): 1.82 (1.33, 2.50), p 0.001]. Participants with a glucose reading $\ge$7.9 to $\le$10.6 (Q4) had an increased risk of 30-day MACEs in unadjusted [OR (95% CI): 1.35 (1.02, 1.80), p = 0.035] and minimally adjusted [OR (95% CI): 1.35 (1.01, 1.80), p = 0.040] models, while the risk was not significant in the fully adjusted model [OR (95% CI): 1.22 (0.89, 1.69), p = 0.219]. Participants with glucose $\ge$6.4 to $\le$7.8 (Q3) or $\le$5.3 (Q1) had a higher but nonsignificant risk of 30-day MACEs, as compared with the reference Q2.

GAM was used to visualize the dose-response relationship between glucose level on admission and the risk of 30-day MACEs. As shown in Fig. 1, the risk of 30-day MACEs had a linear trend of increase with the increase of glucose. For each 1 mmol/L increase in blood glucose, the risk of 30-day MACES increased by 5% [OR (95% CI): 1.05 (1.03, 1.07)]. The parameters of other covariates in the Generalized additive model are shown in Supplementary Table 3. This trend was more significant in STEMI patients [OR (95% CI): 1.07 (1.04, 1.10)], with the risk of 30-day MACEs increasing with blood glucose levels more significant than in NSTEMI patients [OR (95% CI): 1.02 (0.98, 1.06)] (Fig. 2). The dose-response relationship between glucose level and the risk of 30-day MACEs was nearly flat trend (Fig. 2).

Fig. 1.

Estimated risk of MACEs in different blood glucose levels for total participants. The solid red line was the estimated risk, and the dashed lines above and below were the upper and lower limits of 95% CI, respectively. All covariates in model 3 were adjusted. MACEs, major adverse cardiovascular events; OR, odds ratio; CI, confidence interval.

Fig. 2.

Estimated risk of MACEs in different blood glucose levels stratified by MI type. All covariates in model 3 were adjusted. MACEs, major adverse cardiovascular events; STEMI, ST-segment elevation myocardial infarction; NSTEMI, non-ST-segment elevation myocardial infarction; MI, myocardial infarction; OR, odds ratio; CI, confidence interval.

The relationship between glucose level on admission and the risk of 30-day MACEs was still robust across the following subgroups (Table 3): age group (65 vs. $\ge$65; p for interaction = 0.221), sex (female vs. male; p for interaction = 0.894), hypertension (no vs. yes; p for interaction = 0.421), diabetes (no vs. yes; p for interaction = 0.109).

But, there was a significant interaction between glucose level and MI type in fully adjusted model 3 (NSTEMI vs. STEMI; p for interaction = 0.036). For example (Table 4), in STEMI patients, the highest quintile of glucose level (Q5) was significantly associated with an increased risk of 30-day MACEs after full adjustment [OR (95% CI): 2.23 (1.48, 3.35), p 0.001]. But, this association was not significant in NSTEMI patients [OR (95% CI): 1.40 (0.84, 2.34), p = 0.191].

A few limitations of our study need to be addressed. First, given the nature of the retrospective study, it was possible that there was some residual confounding factors that were not measured. In addition, our study was limited to glucose levels on admission. Since data on glucose levels during hospitalization and follow-up information was not available, we were not able to assess whether the glucose levels were persistent during hospitalization. Above all, the proportion of AMI patients receiving standardized coronary reperfusion and drug therapy was limited in this study, which may affect the evaluation of the relationship between admission blood glucose and Mace. Future studies are warranted into the appropriate management of patients with AMI with high glucose levels on admission to hospital.