WSO October 2023 – Issue 1 - 950

950
International Journal of Stroke 18(8)
Covariates
Sociodemographic factors (sex and age), medical resource
utilization, comorbidities, and stroke-related concomitant
prescription medication at baseline were assessed at the
index day. Medical resource utilization includes designated
hospital level for medical insurance and the number of outpatient
and hospital visits in the 12 months prior to baseline.
Comorbidities were identified from inpatient and outpatient
records from 1 January 2008 to baseline by ICD-10
coding and text diagnosis. We established Charlson
comorbidity indexes for patients based on their multisystem
comorbidities to reflect the baseline health status, as
detailed in the construction method described in a previous
study.23 In addition, baseline records of concomitant prescription
drugs, including antihypertensive, anticoagulant
(heparin, warfarin, rivaroxaban, and other thrombin inhibitors),
antiplatelet (aspirin, clopidogrel, etc.), nonsteroidal
anti-inflammatory drugs (NSAIDs), and thrombolytic
drugs (Alteplase, streptokinase, urokinase, etc.), were also
included in the analysis.
Statistical analysis
The respective distributions of baseline characteristics
compared among patients receiving any versus no statin
and high, low/moderate-intensity statin versus no statin
were shown as frequencies with percentages for categorical
variables and means with standard deviations for continuous
variables. Variance between groups was evaluated with
χ2 tests for categorical variables and Kruskal-Wallis tests
for continuous variables.
Unadjusted Kaplan-Meier incidence curves and logrank
tests were used to evaluate the ICH incidence in different
intensity groups and types of statin adherence groups
after the index date (high, low/moderate intensity versus
none and rosuvastatin, simvastatin versus atorvastatin).
Cox regression models were used to estimate the adjusted
hazard ratios (HRs) for the risk of ICH in statin users compared
with nonusers (high, low/moderate-intensity statin
compared with no statin, rosuvastatin, simvastatin compared
with atorvastatin adherence), adjusting for gender,
age, comorbidities, hospital level, medical resource utilization,
and concomitant medication prescription. In addition,
adjusted HRs and 95% confidence intervals (CIs) were calculated
for different statin treatment patterns (different
intensities and types), stratified according to the time interval
between baseline IS and ICH readmission. The proportionality
hazard hypothesis was tested using Schoenfeld
residuals and no proportionality bias was observed. The
Fine-Gray model of competitive risk regression was used to
further account for the competing risk of mortality, and the
covariates were adjusted as above. Since differences in
patient characteristics may affect the outcomes of statin
International Journal of Stroke, 18(8)
therapy, a series of subgroup analyses were constructed,
including age (<60 versus ⩾60 years), sex (male versus
female), hospital rank (primary/secondary versus tertiary),
comorbidity index (0-1 versus ⩾2), number of visits (0-2
versus >2), history of hypertension, use of anticoagulants,
and thrombolytic therapy.
To account for the effect of any statin use throughout the
follow-up period after IS and to make the basic characteristics
of the treated and nontreated groups more consistent, we
defined any statin use after stroke as the treatment group. To
optimize the comparability of users and nonusers in terms of
confounding factors, they were matched 1:1 within a caliper
of 0.01 times the standard error of the logit propensity score
for age, sex, first stroke date, type of stroke, hospital rank,
medical resource utilization, comorbidities, and concomitant
medication prescription. The standardized mean difference
(SMD) of acceptable matching was less than 0.1. Other
sensitivity analyses were conducted by (1) excluding participants
with atrial fibrillation (AF), (2) excluding subarachnoid
hemorrhage as an outcome that could be attributed
to ruptured cerebral aneurysms or arteriovenous malformations
rather than statin-mediated changes in serum cholesterol,
(3) excluding patients who received thrombolytic
therapy to rule out thrombolysis-related ICH, and (4) using
a competing-risk regression model to further examine the
competing risks of other cardiovascular diseases including
various heart diseases and valvular diseases, hypertension,
coronary heart disease, heart failure, and atherosclerosis
(ICD-10: I01-I11, I13, I20-I51, I70).
A two-sided p value of <0.05 was considered to be statistically
significant. All statistical analyses were performed
by using Stata (version 14.0, StataCorp).
Results
Baseline characteristics
A total of 62,252 remaining patients with newly diagnosed
IS were enrolled in this study, of whom 43,434 (69.8%) initiated
statin treatment at baseline and 18,818 (30.2%) were
nonusers (Table 1). Differences in the distribution of basic
characteristics between the two groups were mainly indicated
by younger age, fewer women, fewer hospital visits,
more tertiary hospitals, fewer comorbidities (except for
hypertension), and more thrombolytic therapy among statintreated
patients than nontreated patients (all p value < 0.05).
For the intensity of statin therapy, 34,301 (55.1%) patients
received low/moderate-intensity statin therapy, and 9133
(14.7%) received high-intensity statin therapy. Patients with
high-intensity statin therapy had more comorbidities (especially
hypertension and ischemic heart disease), hospital
visits and medication use and fewer tertiary designated hospitals
than patients with low/moderate-intensity statin therapy
(all p value < 0.05).
WSO October 2023 – Issue 1

Table of Contents for the Digital Edition of WSO October 2023 – Issue 1
