The data for this study were derived from the coronary artery disease (CAD) specialized database within the Tianjin Health and Medical Big Data Super Platform (referred to as the “platform”). Tianjin Health and Medical Big Data Co., Ltd. serves as the authorized data provider responsible for the collection, management, and application of data on this platform. The platform aggregates clinical diagnosis and treatment information from 43 tertiary and 39 secondary hospitals in the Tianjin region, as well as data from the public health system.

After normalization and de-identification on the platform, the data were transformed into a specialized database for scientific research on the CAD database, which includes patients who were hospitalized at least once between January 1, 2010, and June 30, 2023, with discharge diagnoses of CAD. Comprehensive healthcare information for these patients was collected, encompassing demographic characteristics, disease diagnoses, medication and non-medication prescriptions, examination and laboratory test results, surgical information, cost details, community medication and health examination data, as well as public health mortality information.

This study employed the International Classification of Diseases, 10th Revision (ICD-10) diagnostic codes to identify patients discharged with a diagnosis of STEMI. Following STEMI diagnosis, patients were prescribed aspirin as a baseline medication, in addition to either ticagrelor or clopidogrel. Patients were required to have an initial admission platelet count $\ge$350 $\times$ 10⁹/L, a threshold established by large-scale East Asian registry studies, including the TIMI trials [10] and the Korea Acute Myocardial Infarction Registry-National Institutes of Health [17], both of which demonstrated that platelet counts at or above this level are independently associated with significantly increased short- and long-term adverse outcomes in patients with STEMI [10, 17]. This cutoff corresponds to the uppermost range ($>$Q3 + 1.5 $\times$ IQR) in the platelet count distribution of contemporary East Asian AMI cohorts, thereby enriching for a subgroup with markedly heightened platelet reactivity and thrombotic risk. ICD-10 codes were utilized to extract data on complications and comorbidities, including ventricular fibrillation, ventricular tachycardia, third-degree atrioventricular block, diabetes, hypertension, and atrial fibrillation (Table 1). The overall flowchart is illustrated in Fig. 1. Among 150,530 patients with AMI, 1534 were included in the study. These patients were categorized into two groups based on their use of ticagrelor (n = 1073) or clopidogrel (n = 461) during hospitalization. Propensity score matching (PSM) at a 1:1 ratio was performed between the two groups, resulting in 461 patients being analyzed for 1-year follow-up survival. The extracted database population solely comprised patients on aspirin who were treated exclusively with either ticagrelor or clopidogrel, and all patients maintained consistent antiplatelet therapy without any in-hospital changes. Dual antiplatelet therapy was initiated at the discretion of the physician and in accordance with the European Society of Cardiology (ESC) guidelines [18]. In non-PCI patients, tertiary centers frequently preferred ticagrelor for its intensified antithrombotic effect, particularly in cases of thrombocytosis. This study was conducted in accordance with the guiding principles of the Declaration of Helsinki and received approval from the Clinical Research Ethics Committee of the Second Hospital of Tianjin Medical University (KY2023052-01), which waived the requirement for informed patient consent.

Fig. 1.

Study flowchart. MACCE, major adverse cardiac and cerebrovascular event; NACE, net adverse clinical event; AMI, acute myocardial infarction; PSM, propensity score matching; BARC, Bleeding Academic Research Consortium.

This study recorded basic information and clinical data, including sex, age, Killip grade, and comorbidities (hypertension, hyperlipidemia, diabetes, atrial fibrillation, stroke, cerebral hemorrhage, renal insufficiency, and peripheral vascular disease). Hospitalization and discharge medications were also documented, including aspirin, oral anticoagulants, statins, nitrates, beta-blockers, renin–angiotensin system inhibitors (RASIs; angiotensin-converting enzyme inhibitors (ACEIs), angiotensin II receptor blockers (ARBs), or angiotensin receptor-neprilysin inhibitors (ARNIs)), calcium channel antagonists, diuretics, levosimendan, and amiodarone. Laboratory tests included D-dimer and fibrinogen assays. Data collection was conducted by trained physicians using post-discharge electronic health records. Patients with AMI were treated according to established guidelines—unless contraindicated or explicitly rejected by family members—with dual antiplatelet therapy, coronary artery reperfusion therapy, and subsequent ventricular anti-remodeling. Successful PCI was defined as achieving post-procedure TIMI grade 3 flow with 30% residual stenosis. PCI success rates were comparable between groups after matching (96.8% versus 95.7%, p = 0.480).

In this study, MACCE constituted the primary outcome, while secondary outcomes included net adverse clinical events (NACEs), all-cause mortality, cardiac mortality, recurrent non-fatal myocardial infarction, coronary revascularization, cerebral infarction, and bleeding events (Bleeding Academic Research Consortium (BARC) type 3–5). MACCE were specifically defined as cardiac mortality, recurrent non-fatal myocardial infarction, and cerebral infarction, while NACE encompassed cardiac mortality, recurrent non-fatal myocardial infarction, cerebral infarction, and bleeding events (BARC type 3–5).

The study outcomes included both in-hospital adverse events and 1-year adverse outcomes. Primary outcome measures comprised in-hospital and 1-year cardiac mortalities. Secondary in-hospital adverse outcomes included malignant arrhythmias not excluded from the study cohort, specifically ventricular tachycardia, ventricular fibrillation, and ventricular flutter. Secondary 1-year adverse outcomes included all-cause mortality, follow-up myocardial infarction, revascularization events, stroke events, and bleeding events classified as BARC type 3–5. Follow-up myocardial infarctions were defined as subsequent hospital admissions for AMI recorded in the database after the baseline event. Stroke was defined as new hospital admissions within 1 year owing to either new cerebral infarction or hemorrhage. Revascularization events were defined as any revascularization procedure performed on any segment of the coronary artery, including branch vessels, after discharge from the baseline AMI. “BARC type 3–5” events were defined as bleeding incidents that satisfy the BARC criteria for types 3 to 5, as recorded in the database. The breakdown is as follows: type 3a (overt bleeding with a hemoglobin decline of 3–5 g/dL or transfusion of 1–2 units), type 3b (hemoglobin decline $\ge$5 g/dL, transfusion $\ge$3 units, or intervention), type 3c (intracranial bleeding), type 4 (coronary artery bypass grafting-related major bleeding), and type 5 (fatal: 5a probable, 5b confirmed). Incidence values are presented in Table 1. Follow-up was censored at 12 months or the last known vital status.

We conducted a complete-case analysis, ensuring no missing data or imputation of missing values. All statistical analyses were performed using R software (version 4.3.2; R Foundation for Statistical Computing, Vienna, Austria). Variable selection for the exploratory multivariable Cox proportional hazards regression model employed a combined clinical and statistical approach. Clinically relevant covariates identified in prior literature (for example, age, sex, Killip class, hypertension, diabetes mellitus, prior stroke, renal insufficiency, and in-hospital PCI) were included in the model irrespective of their univariate significance. Additional variables with a p value 0.10 in univariate analyses were also considered for inclusion. All statistical tests were two-sided, with statistical significance set at p 0.05.

This analysis adhered to a predefined statistical analysis plan established before data extraction from the Tianjin Health and Medical Big Data Platform. The study protocol and analytic framework received approval from the Clinical Research Ethics Committee of the Second Hospital of Tianjin Medical University (KY2023052-01). Although the study was not prospectively registered in a public trial registry, all primary and secondary endpoints, covariate definitions, and model specifications were prespecified before statistical analysis commenced.

PSM was performed using a 1:1 nearest-neighbor matching algorithm without replacement, with caliper width set to 0.2 times the SD of the propensity score logit. 1:1 propensity score matching was performed using a greedy nearest-neighbor algorithm, maintaining a 1:1 ratio and a caliper width of 0.02 times the SD of the propensity score logit. Matched variables included sex, age, Killip classification, receipt of PCI during hospitalization, prior history of diabetes, hypertension, stroke, or hyperlipidemia, and the use of calcium channel blockers, beta-blockers, RASIs, statins, nitrates, or oral anticoagulants. Following the matching process, inter-group balance was evaluated using SMDs, with values less than 0.1 indicating adequate balance.

We compared baseline characteristics between the ticagrelor and clopidogrel groups pre- and post-PSM. Continuous variables are presented as the mean $\pm$ SD or median IQR, and they were compared using independent-sample t-tests or the Wilcoxon rank-sum test, as appropriate. Categorical variables are reported as frequencies (percentages), and they were compared using the Chi-square or Fisher’s exact test.

To achieve doubly robust estimation, inverse probability treatment weighting (IPTW) was applied to the full cohort, incorporating reperfusion variables (primary PCI (PPCI), elective PCI). Sensitivity analyses included Fine–Gray competing risk models for non-fatal endpoints (with death as a competing event) and landmark Cox models for timing-specific outcomes (in-hospital, 0–30 days, and 31–365 days). Subgroup analyses, based on age, sex, PPCI, renal function, and GPI use, employed interaction tests (p for interaction).

Survival analyses utilized Kaplan–Meier curves to estimate the cumulative incidence of adverse events, with the log-rank test employed to compare survival differences between treatment groups. Exploratory multivariable Cox proportional hazards regression was applied to evaluate the independent association between ticagrelor use and study outcomes. Additionally, due to the limited number of events (62 all-cause deaths and 50 cardiac deaths), the full multivariable Cox models that adjusted for nine prespecified covariates yielded a low events-per-variable ratio (approximately 6–7). These multivariable-adjusted hazard ratios should therefore be regarded as exploratory and interpreted with caution. The primary evidence for the benefits of ticagrelor derives from the more robust propensity score-matched intention-to-treat comparison and doubly robust IPTW analyses.

As depicted in Tables 1,2, significant pre-PSM differences were noted between the clopidogrel and ticagrelor groups. The ticagrelor group exhibited a greater proportion of men (64.4% versus 56.7%, p = 0.005) and a lower mean age (56.3 versus 62.0 years, p 0.001). The distribution of Killip classes also differed significantly (p 0.001), with more patients falling under Killip Class I in the ticagrelor group. Additionally, the ticagrelor group displayed higher rates of PCI (74.4% versus 54.5%, p 0.001) and PPCI (69.6% versus 44.4%, p 0.001). Certain complications and comorbidities, such as ventricular fibrillation, diabetes, and hypertension, demonstrated no significant pre-PSM differences between the two groups.

Post-PSM, the differences were largely mitigated, enhancing comparability between the groups. The proportions of men and the mean ages were balanced (p = 1.000 and p = 0.970, respectively), and the distribution of Killip classes was similar (p = 0.872). PCI usage varied insignificantly (p = 0.761), although the PPCI rate remained slightly higher in the ticagrelor group (69.6% versus 63.1%, p = 0.043), representing potential residual confounding. Medication usage, including RASIs, $\beta$-blockers, nitrates, and diuretics, exhibited no significant post-PSM differences. Laboratory parameters, such as fibrinogen and D-dimer levels, were also comparable between the groups. This matching procedure established a more robust basis for comparing clinical outcomes between the two treatment groups. The C-statistic for the logistic regression model used in propensity score estimation was 0.709, indicating acceptable discriminative ability.

The clinical outcomes of the study, as detailed in Table 3 and Fig. 2, demonstrate significant differences in the effectiveness of ticagrelor compared with that of clopidogrel in patients with STEMI presenting with elevated platelet counts.

Fig. 2.

Kaplan–Meier survival curves illustrating cumulative incidences of various 1-year clinical outcomes in propensity score-matched patients ((A) Cardiac death; (B) All-cause death; (C) MACCE; and (D) NACE). Footnotes: The curves depict 1 minus the Kaplan–Meier survival probability, representing the cumulative incidence function. The step functions are constructed based on event times within specified intervals, with risk tables displaying at-risk numbers and censoring counts positioned below the x-axis at designated time points (days: 0, 30, 60, 90, 120, 150, 180, 210, 240, 270, 300, and 360). The y-axes are standardized to a 0%–25% scale. p values were obtained through a log-rank test comparing clopidogrel to ticagrelor within a propensity score-matched cohort (n = 461 per group at baseline). The analysis employed an intention-to-treat approach, with follow-up censored at 360 days or upon event/death. Abbreviations: MACCE, major adverse cardiac and cerebrovascular event (cardiac death/non-fatal myocardial infarction/cerebral infarction); NACE, net adverse clinical event (MACCE plus BARC type 3–5 bleeding).

The primary outcome, MACCE, was significantly less pronounced in the ticagrelor group (6.9%) than in the clopidogrel group (12.1%, p = 0.008). Similarly, NACE rates were significantly lower in the ticagrelor group (8.2%) than in the clopidogrel group (13.0%, p = 0.021). Furthermore, the ticagrelor group demonstrated a substantial reduction in all-cause mortality (3.9%) relative to the clopidogrel group (9.5%, p 0.001). Cardiac deaths occurred less frequently in the ticagrelor group (3.5%) than in the clopidogrel group (7.4%, p = 0.0096).

Nonetheless, other secondary outcomes—recurrent myocardial infarction, revascularization, new stroke, and “BARC type 3–5” bleeding events—revealed no significant differences between the two groups. Recurrent myocardial infarction incidence rates were 3.3% and 2.4% in the clopidogrel and ticagrelor groups, respectively (p = 0.40). The revascularization rate was 3.5% in the clopidogrel group and 4.1% in the ticagrelor group (p = 0.64). New strokes occurred in 2.0% and 1.3% of the clopidogrel and ticagrelor groups, respectively (p = 0.41). “BARC type 3–5” bleeding events were reported at 0.9% in the clopidogrel group and 1.3% in the ticagrelor group (p = 0.52). BARC type 3 or higher bleeding events refer to those classified by the BARC as types 3 to 5, as documented in the database.

Fig. 3 corroborates these conclusions by presenting aHRs and 95% CIs. The aHR for MACCE was 0.59 (95% CI, 0.37–0.93; p = 0.02), indicating a significantly reduced risk with ticagrelor. Regarding NACE, the aHR was 0.64 (95% CI, 0.42–0.98; p = 0.04). All-cause mortality yielded an aHR of 0.47 (95% CI, 0.26–0.83; p = 0.01), demonstrating a significant survival benefit associated with ticagrelor. Nevertheless, no statistically significant differences were noted for cardiac deaths, recurrent MI, revascularization, new stroke, or “BARC type 3–5” bleeding events (p $>$ 0.05).

Fig. 3.

Adjusted forest plot of clinical outcomes with median follow-up time. Footnotes: The forest plot displays exploratory multivariable-adjusted HRs with 95% CIs derived from Cox proportional hazards regression models within the propensity score-matched cohort (N = 461 per group). The vertical line at HR = 1.0 indicates no difference between groups; points to the left of 1.0 favor ticagrelor, whereas those to the right favor clopidogrel. Models were adjusted for all baseline characteristics listed in Table 1 (including age, sex, Killip class, comorbidities, and complications), as well as in-hospital treatments that were significantly imbalanced after matching (primary PCI). The proportional hazards assumption was verified using Schoenfeld residuals (global p $>$ 0.05 for all outcomes). Follow-up was censored at 12 months or at the last known vital status. MACCE: cardiac mortality, recurrent non-fatal myocardial infarction, and cerebral infarction; NACE: MACCE plus BARC type 3–5 bleeding. BARC type 3–5 refers to major bleeding events classified by the BARC.

IPTW was applied to the entire pre-matching cohort for doubly robust estimation. After IPTW, baseline characteristics were well balanced (all SMD 0.14; Table 4). IPTW-adjusted Cox proportional hazards models confirmed the robustness of the main findings, demonstrating a significant reduction in 1-year MACCE (HR 0.605, 95% CI, 0.378–0.969, p = 0.036) and consistent trends toward lower all-cause mortality (HR 0.557, 95% CI, 0.299–1.036, p = 0.064) and NACE (HR 0.654, 95% CI, 0.425–1.008, p = 0.054) with ticagrelor (Table 5). Fine–Gray competing-risk models accounting for the competing risk of death produced similar results for non-fatal endpoints (Supplementary Table 1).

Sensitivity analyses stratified by treatment era (2010–2018 versus 2019–2023) showed consistent benefit of ticagrelor in the more recent period (2019–2023), with significant reductions in 1-year all-cause death (HR 0.412, 95% CI, 0.182–0.934), MACCE (HR 0.451, 95% CI, 0.244–0.834), and NACE (HR 0.503, 95% CI, 0.278–0.908), whereas no significant differences were observed in the earlier period (Table 6).

Pre-specified subgroup analyses (age, sex, primary PCI, glycoprotein IIb/IIIa inhibitor (GPI) use) showed no significant heterogeneity of treatment effect for most outcomes, although numerical trends favoured greater benefit of ticagrelor in patients without primary PCI and without GPI use (all p_interaction $>$ 0.05 except marginal trends; Supplementary Table 2).

Landmark analysis showed that the mortality benefit of ticagrelor was particularly pronounced beyond the first 30 days, with a significant reduction in all-cause death from day 31 to 365 (0.9% versus 4.6%; p = 0.001). The reduction in all-cause death within the first 30 days did not reach statistical significance (3.0% versus 5.0%; p = 0.179). For cardiovascular death, no statistically significant difference was noted in the first 30 days (3.0% versus 5.0%; p = 0.179), while a significant reduction was identified during the 31–365 day period (0.4% versus 2.4%; p = 0.025) (Table 7).

This study leveraged a vast dataset from 82 secondary and tertiary hospitals in Tianjin City, covering the 2010–2023 period and encompassing a sizable number of patients with AMI. This extensive sample size enhances statistical power and overall reliability. Furthermore, the analysis assessed multiple primary and secondary clinical outcomes, including MACCE, NACE, all-cause mortality, cardiac mortality, recurrent non-fatal myocardial infarction, coronary artery revascularization, stroke, and bleeding events (BARC type 3–5). Such a comprehensive evaluation enabled an in-depth comparison of the efficacy of ticagrelor against that of clopidogrel in patients with STEMI and thrombocytosis.

The observed benefits of ticagrelor—characterized by a significantly lower rate of cardiovascular events than that associated with clopidogrel, with a more pronounced advantage than in previous trials—are attributable to the high-risk patient population, optimized treatment protocols, and rigorous follow-up strategies. Additionally, the high prevalence of CYP2C19 intermediate metabolizers (approximately 40%–45%) within the Han population [29], driven by loss-of-function allele carrier rates of 38.6% for CYP2C192 and 5.2% for CYP2C193, likely enhanced the superiority of ticagrelor, as it circumvents the metabolic limitations of clopidogrel. Although pharmacogenetic data were not directly incorporated, this ethnic-specific context further strengthens the relevance of the findings for East Asian cohorts.

Despite the significantly lower incidence of cardiovascular events associated with ticagrelor than with clopidogrel—potentially attributable to the high-risk thrombocytosis subgroup, optimized protocols, and rigorous follow-up—several limitations warrant consideration.

First, as an administrative database study, endpoint classifications (for example, cardiac death and all-cause mortality) relied solely on ICD codes without independent adjudication of pathological causes, potentially introducing misclassification bias. Furthermore, the database did not capture detailed causes of cardiovascular death (for example, pump failure, reinfarction-related death, stent thrombosis, arrhythmic death, procedure-related death, or unknown cause), limiting the ability to perform sub-classifications and direct comparisons of incidence rates between groups. This constraint hinders a more comprehensive exploration of the factors driving survival benefits, suggesting that future studies should enhance detailed death adjudication.

Second, several non-fatal secondary endpoints exhibited low absolute event counts (for example, recurrent myocardial infarction: 11 versus 15; BARC type 3–5 bleeding: 6 versus 4), thus limiting the statistical power to detect meaningful differences. Additionally, post-PSM sample sizes (n = 461 per group) were modest, potentially yielding insufficient power for rarer outcomes and increasing the risk of false negatives.

Third, all data were sourced from hospitals in Tianjin, where regional variations in medical standards, treatment practices, and population characteristics limit the generalizability of the findings to broader Chinese or international populations. Caution is thus warranted when extrapolating results on a nationwide or global scale.

Fourth, despite stringent inclusion and exclusion criteria, residual selection bias might have persisted owing to patient heterogeneity, variability in treatment strategies, and inconsistencies in follow-up duration. The Killip classification, derived from discharge documentation, may not accurately reflect admission severity; although adjusted for shock proxies (vasopressors, intubation, and intra-aortic balloon pump/extracorporeal membrane oxygenation), residual misclassification remains possible and requires prospective validation. While temporal adjustments increased robustness, residual secular trends are acknowledged as limitations. Moreover, post-discharge medication adherence was unmeasured, and reactive thrombocytosis was not adjusted for owing to the unavailability of markers—factors warranting future exploration via claims data integration and marker assessment. Despite PSM and IPTW, residual confounding may persist, including the slight post-PSM imbalance in primary PCI rate (p = 0.043), which could influence ischemic outcomes favoring ticagrelor. However, the consistency of benefits in IPTW-adjusted models (which included reperfusion variables) and sensitivity analyses mitigates this concern.

Fifth, the database lacks longitudinal outpatient medication dispensing or refill data. As a result, post-discharge adherence to the initially prescribed P2Y12 inhibitor, rates of treatment switching, and discontinuation could not be evaluated. The intention-to-treat analysis assuming persistent exposure to the discharge medication may therefore underestimate or overestimate the true on-treatment effect of ticagrelor versus clopidogrel. Importantly, this limitation most likely introduces a conservative bias that underestimates the true magnitude of ticagrelor’s benefit. In an ITT framework, patients initially prescribed ticagrelor who subsequently discontinued therapy or switched to clopidogrel (e.g., due to dyspnea, cost, or physician preference) would be analyzed in the ticagrelor arm despite receiving reduced or no exposure to the drug. The fact that highly significant reductions in MACCE, all-cause mortality, and cardiac mortality were still observed despite this probable dilution of treatment effect strongly supports a robust biological advantage of ticagrelor in this high-thrombotic-risk population and validates the observed benefits. Future studies incorporating pharmacy claims linkage or prospective follow-up are needed to address adherence and persistence in this high-risk population. Future research should prioritize the assessment of bleeding risks across diverse platelet functional states and implement targeted preventive measures.

Additionally, due to the limited number of events (62 all-cause deaths and 50 cardiac deaths), the full multivariable Cox models that adjusted for nine prespecified covariates yielded a low events-per-variable ratio (approximately 6–7). These analyses are exploratory, and the multivariable-adjusted hazard ratios should therefore be regarded as exploratory and interpreted with caution. The primary evidence for the benefits of ticagrelor derives from the more robust propensity score-matched intention-to-treat comparison and doubly robust IPTW analyses, which are not subject to the same events-per-variable constraint.

Although the subgroup analyses were prespecified and adjusted for multiple comparisons where appropriate, the nominal interactions observed for MACCE (p_interaction = 0.039) and NACE (p_interaction = 0.047) in patients with versus without primary PCI should be interpreted with considerable caution. These interactions are driven by low event counts in several subgroups and composite endpoints that include non-fatal events of lesser clinical severity, rendering them underpowered for detecting true heterogeneity of treatment effect. Accordingly, no definitive claims of subgroup-specific differences are made, and the overall findings remain consistent across the majority of tested subgroups.

To address these limitations, subsequent studies should adopt prospective designs with comprehensive data collection (for instance, detailed death adjudication, genotype profiling, and precise timing of platelet measurements), minimize biases through multicenter recruitment, and enhance generalizability via nationwide cohorts.