Study design
This is a retrospective cohort study. We used insurance claims data provided by the Taiwanese Bureau of National Health Insurance (BNHI) from January 2003 to December 2016. The BNHI stores complete follow-up information on major interventions and medications as well as admission, outpatient clinic, and emergency department visit records of the Taiwanese population. Statin users were defined as those patients who received at least one statin prescription for more than 180 days during the study period. In contrast, non-stain users were designated as those patients who did not receive a statin prescription throughout the whole study period.
Study population
The study population comprised patients with RA (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code: 714.0) who were admitted to the hospital or outpatients who were newly diagnosed with hyperlipidemia (ICD-9-CM code272.X) and were statin users between 2003 and 2016. At least one of the following enrollment criteria had to be met for inclusion in this study: (1) two or more outpatient visits within a six-month period or (2) all statin prescriptions were continuously administered to the patients for more than 6 months within a 14-year follow-up period. Comorbidities related to dementia were as follows: coronary heart disease (ICD-9-CM code 410–415), hypertension (ICD-9-CM code 401–405), diabetes mellitus (ICD-9-CM code 250), and cerebrovascular attack (ICD-9-CM code 430–438). Patients who fulfilled any of the following criteria were excluded from the study: (1) had a prior history of dementia before January 1, 2003 and (2) patients aged less than 40 years. Given the differences in baseline characteristics and dementia risk between the statin users and non-statin users, we applied age, sex, RA duration, and propensity score matching at a ratio of 1:1 for patients with RA with and without statin use. Finally, the study group comprised 25,764 participants with RA who were statin users, and the control group included 25,764 randomly selected participants with RA who were non-statin users (Figure 1).
Main outcome measures
The primary endpoint was the development of dementia, which was defined by the time a code of dementia (ICD-9-CM codes 290) first appeared in the inpatient or outpatient claim records.
Statistical analysis
Initially, the propensity score matching (PSMATCH ) was conducted between two groups by using the Statistical Analysis System (SAS) software. The PSMATCH Procedure was used to perform the greedy nearest neighbor matching. First, the logistic regression was performed to estimate the logit probability of exposure to statin for each individual. The details of predictors including RA duration, sex, age, comorbidities, and concurrent medication taken by the patients are listed in Table 1. The algorithm used was the greedy nearest neighbor matching with the caliper of 0.01.
Differences in demographic data and clinical characteristics between statin users and non-statin users were examined using t-test for continuous variables, whereas chi-square tests were used for categorical variables. The dementia-free survival rates in the two groups were calculated using the Kaplan–Meier method and the log-rank test. The Cox proportional hazard regression model was used to compare the development risk of dementia between statin users and non-statin users. Adjusted HRs and 95%CIs were calculated. Their values were adjusted for important risk factors, such as sex, concurrent medication, and comorbidities, for the development of dementia. P values <0.05 were considered to be statistically significant for this study. All the statistical calculations were performed using statistical analysis software, version 9.3 (SAS Institute, Inc., Cary, NC, USA).