Study Design and Population
Our analysis was based on the CCC-ACS project, which is a nationwide registry jointly initiated by the American Heart Association and the Chinese Society of Cardiology from 2014. The CCC-ACS project was approved by the institutional review board of Beijing Anzhen Hospital, Capital Medical University, with a waiver for informed consent. This study is registered at the following URL: https://clinicaltrial.gov (unique identifier: NCT02306616).
A total of 104,516 ACS patients were enrolled in the CCC-ACS project from November 2014 to July 2019. As shown in Figure 1, we included 5,896 patients (3,322 GPI users and 2,574 non-GPI users) for analysis after excluding the following groups: those admitted with a diagnosis of non-ST-elevation myocardial infarction; those who were not treated with PCI; those with a missing value for body weight; those who were not treated by thrombus aspiration therapy; and those who received GPI after the occurrence of an ischemic event during hospitalization. GPIs used in the CCC-ACS project included tirofiban, eptifibatide, abciximab, or others at any time during the indexed hospitalization.
Study Covariates
The following variables were treated as covariates for multivariable adjustment and propensity score matching: demographics (age, sex, and body weight); previous history (diabetes, hypertension, dyslipidemia, smoking, MI, PCI, coronary artery bypass grafting, atrial fibrillation, heart failure, renal failure, ischemic stroke, hemorrhagic stroke, peripheral vascular disease, chronic obstructive pulmonary disease); on-admission clinical features [Killip class, peak levels of creatine kinase-MB (CK-MB) isoform, serum levels of low density lipoprotein cholesterol (LDL-C), high density lipoprotein cholesterol (HDL-C) and triglycerides (TG), levels of systolic and diastolic blood pressure (SBP and DBP), heart rate, estimated glomerular filtration rate (eGFR) and baseline hemoglobin)]; prehospital medications (prehospital thrombolysis, aspirin, P2Y12 inhibitors, statins, β-blockers, angiotensin converting enzyme inhibitors (ACEIs) or angiotensin receptor blockers (ARBs), aldosterone antagonists and oral anticoagulants); in-hospital medications [DAPT status, P2Y12 inhibitors (clopidogrel or ticagrelor), statins, β-blockers, ACEIs/ARBs, aldosterone antagonists, oral anticoagulants, and perioperative anticoagulants (unfractionated heparin, low molecular weight heparin (LMWH) and others)]; and PCI-related characteristics [PCI types (primary PCI < 12 hours after symptom onset, primary PCI ≥ 12 hours after symptom onset, rescue PCI, and elective PCI) and radial route for PCI or not]. Estimated glomerular filtration rate (eGFR) was calculated according to the equation by chronic kidney disease.14 DAPT status within the first 24 hours was defined by one of the following three categories: non-loading DAPT (DAPT not in loading dose), single-loading DAPT, and both-loading DAPT (DAPT both in loading dose). The loading dose of aspirin was defined as ≥ 150 mg. The loading dose of the P2Y12 receptor inhibitor was defined as ≥ 300 mg for clopidogrel and ≥ 180 mg for ticagrelor. The definitions of the abovementioned study variables are listed in Supplemental Table 1.
Study Outcomes
The primary study outcome concerned major in-hospital bleeds, defined by any of the following three major bleeding definitions that occurred during hospitalization: (a) Bleeding Academic Research Consortium (BARC) type 3b (defined as a hemoglobin drop of ≥ 5 g/dL or cardiac tamponade or bleeding requiring surgical intervention or bleeding requiring intravenous vasoactive agents), 3c (intracranial hemorrhage) and type 5 (fatal bleeding); (b) Thrombolysis in Myocardial Infarction (TIMI) major bleeding (defined as intracranial hemorrhage or clinically overt bleeding associated with a hemoglobin drop of ≥ 5 g/dL, or fatal bleeding); and (c) PLATelet inhibition and patient Outcomes (PLATO) life-threatening bleeding (defined as fatal bleeding, intracranial bleeding, intraoperative bleeding with cardiac tamponade, severe hypotension, hypovolemic shock because of bleeding and requiring either vasopressor or surgery, a hemoglobin drop of ≥ 5 g/dL, or the need for transfusion > 4 U of whole blood or packed red blood cells). Coronary artery bypass-grafting-related bleeding was excluded. Other study outcomes included in-hospital mortality and less severe but clinically significant in-hospital bleeds (defined as a hemoglobin drop of 3 to 5 g/dL). The associations between GPI use and ischemic events and all-cause in-hospital mortality were also examined. We defined ischemic events as the occurrence of reinfarction, ischemic stroke, non-bleeding related fatal events, and in-stent thrombosis. The data of study outcomes above for this study were collected from their medical records.
Statistical Analysis
Results obtained from continuous data with normal distribution are presented as means and standard deviations. Those from non-normal continuous data are presented as medians with 25th and 75th percentiles, and those from categorical data are presented as numbers and percentages. The absolute standardized difference (ASD), which is superior to rank-sum tests or t-tests because it is independent of sample size, was used for between-group comparisons. The between-group imbalances were considered ideal if the ASD was less than 10% (Stata command “stddiff”). We used propensity score matching to balance the differences in patient demographics, medical history, and pre-admission and in-hospital management strategies between GPI users and non-GPI users. We developed a non-parsimonious multivariable logistic regression model to estimate a propensity score for GPI status (yes or no) as the dependent variable. Then, a propensity score matching of a maximal ratio of 1-to-1, without replacement, with a caliper width of 0.02 was performed (Stata command “calipmatch”). The risk of in-hospital bleeding, ischemic events, and mortality in the matched groups was assessed using a logistic regression model on the matched pairs.
We performed the following interaction tests and subgroup analyses based on matched population, including age (< 65 and ≥ 65 years), sex, eGFR, Killip class (Class I vs. > Class I), DAPT status (full loading or not), and Low Molecular Weight Heparin LWMH (use or not).
Finally, we performed the following sensitivity analyses based on the matching cohort that excluded the following: (a) patients who died within 48 hours of admission; (b) patients with Killip Class IV; (c) patients receiving ticagrelor; (d) patients receiving DAPT with both in loading dose; (e) femoral PCI; and (f) patients receiving unfractionated heparin (nonoperative, unfractionated heparin use). Additionally, an inverse probability weighting based on multivariate logistic regression (Stata command “teffects ipw”) was used as a sensitivity analysis to validate the primary findings.
We imputed data for variables with missing values using the sequential regression multiple imputation method by IVEware (version 0.2; Survey Research Center, University of Michigan, Ann Arbor, MI), as previously described.15 The missing rates of the study variables are shown in Supplemental Table 2. We used Stata version 15.1 (StataCorp, College Station, TX) for analysis. A two-tailed p < 0.05 was considered statistically significant.