Between January 2010 and February 2020, we enrolled STEMI patients undergoing PCI at the Guangdong Provincial People's Hospital. The definition of STEMI was determined in accordance with the guidelines of the European Society of Cardiology (2, 17). We excluded any patients: (1) who had a diagnosis of non-ST-elevation myocardial infarction (NSTEMI); (2) who did not undergo PCI treatment; (3) for whom no admission laboratory parameters were available, or for whom variables needed to calculate the ratio were missing; or (4) who had active carcinoma or acute/chronic inflammatory diseases. This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human subjects/patients were approved by the research ethics committee of Guangdong Provincial People's Hospital (No. GDREC2016378H). Written informed consent was obtained from all participants.

All patients were treated in accordance with the hospital protocol and the European Society of Cardiology guidelines for STEMI patients (2, 17). Within 24 h of admission, before PCI, blood samples were obtained to measure the relevant laboratory parameters, including admission blood glucose and serum albumin levels. Before PCI, all eligible patients were prescribed aspirin (300 mg) and clopidogrel (300 or 600 mg) or ticagrelor (180 mg). Following the recommendations for catheter procedures (18), coronary angiography and PCI were performed by experienced interventional cardiologists, who made all decisions relating to the procedures, including the choice of vessel access, the selection of stents, the volume of contrast, and so on. After PCI, patients received dual antiplatelet therapy, including aspirin (100 mg/day) combined with clopidogrel (75 mg/day) or ticagrelor (90 mg twice/day). Other medications were prescribed by the patients’ physicians according to standard regimens. Clinical information on demographic characteristics and other relevant data on the enrolled patients were extracted from the hospital electronic database by trained researchers. Follow-up data were collected by trained nurses through telephone tracking or outpatient visits.

In-hospital all-cause mortality was the primary outcome. The secondary outcomes were in-hospital major adverse cardiac events (MACEs), as well as all-cause mortality and MACEs during follow-up. In-hospital MACEs were defined as the combined endpoint of all-cause mortality, stroke, target vessel revascularization, and recurrent myocardial infarction during hospitalization. MACEs during follow-up were defined as the combined endpoint of follow-up all-cause mortality or stroke.

AAR was calculated as [admission blood glucose (mmol/L) × 18/albumin (g/L)] or [admission blood glucose (mg/dl)/albumin (g/L)] using the same blood samples obtained at admission. The value of AAR was calculated for every patient according to the formula. AAR values were stratified into tertiles, according to which patients were divided into three groups: T1 (n = 1,074; AAR < 3.35), T2 (n = 1,076; 3.35 ≤ AAR < 4.56), and T3 (n = 1,074; AAR ≥ 4.56). Categorical variables are reported in the form of numbers (percentages). In order to compare the distributions of categorical variables, Chi-square tests or Fisher's exact tests were employed. For continuous variables, normally distributed data are reported in the form of means ± standard deviation (SD), while non-normally distributed data are reported in the form of medians (Q25–Q75). Normally distributed continuous variables were compared using one-way ANOVA; otherwise, the Kruskal–Wallis test was performed. To calculate odds ratios (ORs) and 95% confidence intervals (CIs) for in-hospital mortality and MACEs according to AAR, we performed univariate and multivariate logistic regression analyses. Model 1 was adjusted for, age, sex, smoking status, prior stroke, chronic obstructive pulmonary disease, diabetes mellitus (DM), hypertension, prior myocardial infarction, prior PCI, anemia, Killip class ≥II, estimated glomerular filtration rate (eGFR), aspirin, glycoprotein IIb/IIIa inhibitors, multiple lesions, and transradial access. Model 2 was adjusted for alanine aminotransferase (ALT) in addition to the covariates adjusted for in Model 1; Model 3 was adjusted for insulin in addition to the covariates adjusted for in Model 1; and Model 4 was adjusted for heart rate and prior cardiac arrest in addition to the covariates adjusted for in Model 1. These models were used to perform univariate and multivariate logistic regression analyses. To determine how well AAR predicts in-hospital mortality and MACEs, receiver operating characteristic (ROC) curves were constructed and the area under the ROC curve (AUC) was subsequently calculated. The AUC value can be used to evaluate the efficacy of the predictor (AUC <0.6, poor discrimination; 0.6–0.75, good discrimination; >0.75, excellent discrimination) (19). The optimal cutoff value was determined by the maximum Youden index (Youden = sensitivity + specificity–1). To assess the prognostic value of AAR in patients with and without DM, ROC curves were plotted to determine the optimal cutoff value for AAR as a predictor of MACEs in each of the two groups. In addition, nomograms for the predictive value of in-hospital all-cause mortality and MACEs were constructed. To assess for non-linearity in the relationship between AAR as a continuous variable and in-hospital death or MACEs, we fitted restricted cubic spline models to identify a non-linear relationship. Kaplan–Meier analysis was conducted to visualize several survival curves illustrating follow-up risks of mortality and MACEs in different AAR groups. The effect of AAR on long-term mortality and MACEs was assessed using univariate and multivariate Cox proportional hazards regression analyses in Models 1–4. Log-rank tests were used to compare survival curves between groups. A P-value of 0.05 or below was considered to represent statistical significance. SAS version 9.4 (SAS Institute, Cary, NC, USA) was used to perform the statistical analyses.

A flow chart of the study is presented in Figure 1. Patients were divided into three groups according to AAR tertile. The mean AAR value across all patients was 4.58 ± 2.38, and the average age was 62.10 ± 12.24 years. The baseline clinical characteristics of the three groups are presented in Table 1. Compared to the other groups, patients in the T3 group were older, had a higher Killip class, stayed in hospital for longer, had lower systolic and diastolic blood pressure, and were more likely to have been treated with prior PCI and to have been diagnosed with hypertension, anemia, DM, and stroke. In addition, a higher AAR was associated with higher hemoglobin A1c (HbA1c) and serum creatinine levels, as well as lower low-density lipoprotein cholesterol, total cholesterol, hemoglobin, left ventricular ejection fraction, and eGFR.

In total, 130 patients (3.9%) died in hospital and 181 (5.4%) experienced in-hospital MACEs. Moreover, significantly higher rates of in-hospital all-cause mortality and MACEs were reported in the T3 group compared to the T1 group (8.0% vs. 1.5% and 9.9% vs. 3.0%, respectively; P < 0.001).

Among these STEMI patients after PCI, the continuous variable of AAR was positively associated with risk of in-hospital all-cause mortality (OR: 1.22, 95% CI: 1.16–1.27, P = 0.000) and MACEs (OR: 1.19, 95% CI: 1.14–1.24, P = 0.000) in univariate logistic regression analyses. After adjusting for confounders in the multivariate logistic regression analysis, the results showed that a higher risk of in-hospital all-cause mortality (adjusted OR: 1.14, 95% CI: 1.07–1.22, P = 0.000) and MACEs (adjusted OR: 1.11, 95% CI: 1.05–1.18, P = 0.001) (Supplementary Table S1, Model 1) was independently predicted by AAR, indicating that AAR was an independent predictor of these outcomes. When AAR was examined as a categorical variable, univariate logistic regression analysis revealed that the T3 group was at an elevated risk of unfavorable outcomes in terms of in-hospital all-cause mortality (OR: 5.75, 95% CI: 3.35–9.87, P = 0.000) and MACEs (OR: 3.57, 95% CI: 2.38–5.35, P = 0.000), but not the T2 group (OR: 1.77, 95% CI: 0.95–3.28, P = 0.072; OR: 1.36, 95% CI: 0.85–2.16, P = 0.201, respectively). Furthermore, the results of the multivariate logistic regression analysis revealed that AAR level was substantially associated with the adjusted risk of in-hospital all-cause mortality, with higher adjusted risk in the T3 group (adjusted OR: 2.72, 95% CI: 1.47–5.03, P = 0.001) and MACEs (adjusted OR: 1.91, 95% CI: 1.18–3.10, P = 0.009) as compared to the T1 group (Supplementary Table S2, Model 1). Furthermore, when considering the effects of factors potentially associated with albumin or blood glucose and several important determinants of prognosis, these findings of the logistic regression analyses were not materially affected in additional models that further adjusted for ALT (Supplementary Tables S1 and S2, Model 2), insulin (Supplementary Tables S1 and S2, Model 3), heart rate, and prior cardiac arrest (Supplementary Tables S1 and S2, Model 4).

ROC curve analysis showed that AAR could predict in-hospital all-cause mortality (AUC = 0.718, 95% CI: 0.675–0.761) (Figure 2A) as well as MACEs (AUC = 0.672, 95% CI: 0.631–0.712) (Figure 2B). The optimal cutoff value for AAR was 4.429, as calculated on the basis of the Youden index. Interestingly, according to ROC curve analysis in the DM subgroup, the predictive value of AAR for in-hospital all-cause mortality was different for patients with and without DM (with-DM group: AUC = 0.653, 95% CI: 0.587–0.718; without-DM group: AUC = 0.669, 95% CI: 0.615–0.723; interaction P = 0.001) (Figure 2C). Furthermore, it is necessary to account for HbA1c, one of the indicators of DM and long-term glycemic control, in evaluating the prognostic value of AAR. However, additional ROC analyses for HbA1c subgroups revealed that there was no significant difference between HbA1c subgroups for either in-hospital all-cause mortality (HbA1c ≤6.1 subgroup: AUC = 0.723, 95% CI: 0.662–0.783; HbA1c >6.1 subgroup: AUC = 0.749, 95% CI: 0.691–0.808; P for comparison = 0.535) or in-hospital MACEs (HbA1c ≤6.1 subgroup: AUC = 0.666, 95% CI: 0.609–0.722; HbA1c >6.1 subgroup: AUC = 0.704, 95% CI: 0.648–0.761; P for comparison = 0.343).

Several other factors should be considered in the evaluation of the predictive value of AAR. In order to compare AAR with two individual indices (admission blood glucose and albumin), ROC curve analyses for these three indicators were conducted (Supplementary Figure S1). These analyses showed that, for both in-hospital all-cause mortality and in-hospital MACEs, the predictive value of AAR was significantly greater than that of admission blood glucose (AUC for mortality = 0.664, 95% CI: 0.671–0.712, P for comparison with AAR <0.0001; AUC for MACEs = 0.633, 95% CI: 0.591–0.676, P for comparison with AAR <0.0001), while the AUC for AAR was marginally higher than the AUC for albumin (AUC for mortality = 0.701, 95% CI: 0.655–0.747, P for comparison with AAR = 0.508; AUC for MACEs = 0.655, 95% CI: 0.614–0.696, P for comparison with AAR = 0.466). In addition, time heterogeneity should be considered. Specifically, given the fluctuations in glucose level that occur over the 24 h of each day (20, 21), we wondered whether the effect of postprandial blood glucose had an impact on the prognostic value of AAR. To test this, we divided patients into three groups according to the time period in which the admission measurements were taken: 06:00–12:00 (morning), 12:00–18:00 (afternoon), and 18:00–06:00 (nighttime); these groups consisted of 380 (11.8%), 1,358 (42.1%), and 1,486 (46.1%) STEMI patients who underwent PCI, respectively. ROC curve analyses for each of the three groups (Supplementary Figure S2) showed that the AUC values for prediction of clinical outcomes were largest for the nighttime group (18:00–6:00) (AUC for mortality = 0.734, 95% CI: 0.677–0.791; AUC for MACEs = 0.707, 95% CI: 0.655–0.759); however, comparisons among the three groups indicated no statistically significant differences in the AUC values for in-hospital all-cause mortality (P = 0.614) or MACEs (P = 0.186). Finally, given that the 10-year span of the data collection period may have been a source of major heterogeneity, we performed ROC curve analyses with patients categorized into two 5-year subgroups, namely a 2010–2015 subgroup and a 2016–2020 subgroup (Supplementary Figure S3). These analyses revealed that the 10-year time heterogeneity was small, both in the case of in-hospital all-cause mortality and in the case of in-hospital MACEs (P for comparisons between AUCs of two subgroups = 0.583 and 0.446, respectively).

According to the restricted cubic spline model, as AAR increased, ORs for in-hospital all-cause mortality and MACEs increased (Supplementary Figure S4). Moreover, in order to confirm the predictive value of AAR for in-hospital all-cause mortality and MACEs, nomograms were constructed (Supplementary Figure S5). In summary, AAR provides satisfactory prognostic value in the prediction of adverse events during hospitalization.

Over the median follow-up duration of 1.66 years, 339 cases of all-cause mortality and 376 MACEs occurred: by group, there were 57 (5.3%), 100 (9.3%), and 182 (16.9%) cases of all-cause mortality and 70 (6.5%), 108 (10.0%), and 198 (18.4%) MACEs in the T1 group (AAR < 3.35), the T2 group (3.35 ≤ AAR < 4.56), and the T3 group (AAR ≥ 4.56), respectively. Based on Kaplan–Meier curves, patients in the T3 group experienced higher rates of all-cause mortality and MACEs during follow-up than those in the T1 group (P < 0.0001) (Figure 3). According to the results of multivariable Cox regression analysis, AAR was found to be an independent risk factor for long-term all-cause mortality and MACEs, regardless of whether AAR was treated as a continuous variable or a categorical variable (Table 2 and Supplementary Table S3). In particular, when AAR was treated as a categorical variable, the T3 group had a higher risk of all-cause mortality [adjusted hazard ratio (HR): 1.64, 95% CI: 1.19–2.28, P = 0.003] as well as MACEs (adjusted HR: 1.58, 95% CI: 1.16–2.14, P = 0.003) compared with the T1 group (Table 2, Model 1). Similarly, the HRs for all-cause mortality and MACEs in different models were comparable when further adjusting for ALT (Table 2 and Supplementary Table S3, Model 2), insulin (Table 2 and Supplementary Table S3, Model 3), heart rate, and prior cardiac arrest (Table 2 and Supplementary Table S3, Model 4).