2.1. Data source
The National Health Insurance (NHI) program in Taiwan is since 1995 and covers the medical needs of 99.8% of 23 million people in the country. All standardized information and data of in this healthcare services are prospectively recorded by the NHIRD and contains inpatient and outpatient data, including date of birth, sex, diagnosis codes (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes), drug prescriptions, surgical procedures, admission dates, hospitalizations, discharge dates, and expenditure which has been validated by the previous researches [27-31]. The ethics approval and protocol of this study were approved by the Ethics Institutional Review Board of Linkou Chang Gung Memorial Hospital.
2.2. Identification of study cohort
The current study was a nationwide population-based, observational and open cohort using retrospective-collected data from the NHIRD between January 1, 2007 and December 31, 2013. We identified patients with diagnoses of T2DM (excluding type 1 diabetes mellitus) between January 1, 2007 and December 31, 2013. Only 121,760 T2DM patients admitted with a principal diagnosis of ACS or AIS were included for analysis. After relevant exclusion, 6,959 patients with T2DM aged ≥40 years who were admitted due to ACS or AIS were eligible for analyses and after propensity score matching (PSM) in a 1:1 ratio, a total of 3,372 subjects (1,686 subjects in the ATOR 40 group and 1,686 subjects in the EZ-SIM 20 group) were included into the final analysis (Figure 1). The definition of the index date was the date of discharge. The follow-up period was based on the index date to the date of death or December 31, 2013.
2.3. Exposure of study statin
During the study period, patients received these two kinds of statin-containing drugs according to the lipid-lowering guidelines in Taiwan. T2DM patients with elevated level of LDL-C ≥130 mg/dL received NHI-paid LDL-C lowering agents to reach the therapeutic goal of LDL-C. The treatment goal of T2DM patients with established CV diseases by NHI in Taiwan was less than 100 mg/dL during the study period. Eligible patients who received study drugs within 30 days of the index date were divided into two groups according to the prescribed LDL-C lowering agents, the ATOR 40 group and the EZ-SIM 20 group, respectively. This definition of exposure of study drugs was reported in our previous publications [32-34].
2.4. Ascertainment of T2DM, ACS and AIS
The diagnosis of T2DM is validated according to ICD-9-CM codes where at least four visits of outpatient corresponded to an accuracy of 95.7% and with prescriptions of oral anti-diabetic agents corresponded to an accuracy of 99% [35]. We identified patients with T2DM based on diagnosis code and anti-diabetic agents simultaneously. The inclusions of ACS and AIS were requiring a principal diagnosis of admission. The diagnosis codes of ACS and AIS have been validated in previous NHIRD studies that have obtained high positive predictive values with ≥95% [27,29,30].
2.5. Covariates measurements
Comorbidities and histories of event at baseline were identified based on ICD-9-CM diagnosis codes (Supplemental Table 1). Comorbidities were defined as at least two outpatient visits or anyone inpatient diagnosis in the previous year of the index date. Histories of event was detected using anyone inpatient diagnosis before the index date which can be tracked to year 1997. The baseline medication was defined as the medications prescribed in the outpatient visits or the refill in the pharmacy within 30 days of the index date.
2.6. Ascertainment of primary, secondary and safety outcomes
The primary composite outcome was an endpoint of CV death, non-fatal MI, and non-fatal stroke. The occurrences of ACS and AIS were defined as a principal inpatient diagnosis. The definition of CV death is the criteria of the Standardized Definitions for Cardiovascular and Stroke Endpoint Events in ClinicalTrials by the FDA in the United States.
The secondary composite CV outcome was defined as hospitalization for unstable angina (HUA), percutaneous coronary intervention (PCI), and coronary artery bypass grafting (CABG). The occurrence of HUA was defined as a principal inpatient diagnosis. Information of PCI and CABG was extracted using the Taiwan NHI reimbursement codes of inpatient claims. Other secondary outcomes included hospitalization for heart failure (HHF), and all-cause mortality [36].
The safety outcomes were defined as hemorrhagic stroke, acute hepatitis, rhabdomyolysis, newly diagnosed dementia and newly diagnosed cancer during the period of follow-up.
2.7. Statistical analysis
To decrease bias due to confounding when comparing treatment effects between the ATOR 40 and EZ-SIM 20 groups, we conducted a propensity score matching analysis. The propensity score was the predicted probability of being the in one group given the values of covariates in the logistic regression. The selected covariates to calculate propensity score were listed in Table 1 where the follow up year was replaced with the index date. We adopted a greedy nearest neighbor algorithm with a caliper of 0.2 times the standard deviation of the logit of propensity score, with random matching order and without replacement. Matching quality was assessed using the absolute value of the standardized difference (STD) between the groups after matching, where a value lower than 0.1 represented negligible difference between the groups.
We compared the risks of all-cause mortality between the groups by using the Cox proportional hazards model. The risks of other time-to-event outcomes in the two groups were compared using the Fine and Gray subdistribution hazard model that considered death a competing risk. The within-pair clustering of outcomes after propensity score matching was accounted for by using a robust standard error (the marginal model) [37]. The unadjusted cumulative incidence function of the outcomes was calculated and plotted under a subdistribution hazard model. The study group (ATOR 40 versus EZ-SIM 20) was the only explanatory variable in survival analyses.
A post-hoc subgroup analysis was performed to determine whether the subdistribution hazard ratios of secondary composite CV outcomes for the ATOR 40 and EZ-SIM 20 groups were similar in the subgroups. A two-sided P value of less than 0.05 was considered statistically significant and no adjustment for multiple testing (multiplicity) was made in this study. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA), including the procedures of ‘psmatch’ for PSM, ‘phreg’ for survival analyses, and the macro of ‘%cif’ for the cumulative incidence function.