Methods
This case-control study, nested into population of Finland, was based on The Finnish Parkinson’s disease study (FINPARK), which includes 22,189 community-dwelling persons who received clinically verified PD diagnosis during 1996–2015. Special Reimbursement Register maintained by the Social Insurance Institution of Finland was used to identify persons with PD diagnosis. PD diagnosis was based on United Kingdom Parkinson’s Disease Society Brain Bank’s criteria [15]. The FINPARK study and exclusion diagnoses have been described in detail previously [16].
Identification of persons with diabetes within the FINPARK cohort was done using Special Reimbursement Register (code 103, since 1972) and Prescription Register (Anatomical Therapeutic Chemical classification (ATC) [17]; ATC code A10 excluding guar gum, since 1995). Prescription Register contains information on all reimbursed prescription drug dispensing for community-dwelling persons as it does not include drug use during hospital or public nursing home care. Cases who received diabetes diagnosis after PD diagnosis (n=927) or less than three years before PD diagnosis date (index date) (n=639) were excluded (Figure 1). Up to four controls were matched for every PD case by age (+/- 2 years), sex, same or adjacent university hospital district, and time from diabetes diagnosis (+/- 2 years). Controls were persons without PD using the same exclusion criteria that was applied to PD cases. In addition, controls who had dementia in Parkinson’s disease (ICD-10 code F02.3) within two years of index date were excluded. Furthermore, 16 PD cases were excluded as no controls could be matched to them.
The final study population for our main analysis included 2,017 cases and 7,934 matched controls. To address healthy user bias and confounding by indication, we performed a secondary analysis restricted to statin users only (1,045 cases and 2,817 controls after excluding cases and controls who were unmatched after the exclusion of nonusers).
The first purchases of statins among PD cases and controls since 1995 were collected from the Prescription Register. Statin use was defined using ATC codes and included simvastatin, lovastatin, pravastatin, fluvastatin, atorvastatin, cerivastatin, and rosuvastatin (supplementary table 1). While pitavastatin is authorized for sale in Finland, no purchases were present in the study data. A three-year lag time was used to exclude statin purchases that were made within a three-year period before PD diagnosis. Statin use was categorized as non-use, statin use initiated during lag time and use before three-year lag. Cumulative statin exposure was defined as a continuous variable by defined daily dose (DDD) [17] which was then divided into tertiles.
Identification of comorbidities (any cardiovascular disease, coronary artery disease, history of cancer, asthma or chronic obstructive pulmonary disease (COPD), stroke, history of traumatic brain injury, and rheumatoid arthritis and connective tissue diseases) is described in detail in Supplementary Table 2. Care Register for Health Care (International Classification of Diseases 8th, 9th, and 10th revision), Special Reimbursement Register (reimbursement codes) and Cancer Register (International Association of Cancer Registries Tools coding) were used to identify these conditions before the three-year lag.
T-test and Mann-Whitney U test were used to compare the distribution of continuous variables and chi square test to compare the distribution of categorical variables between cases and controls. Conditional logistic regression was used to calculate odds ratios (OR) and 95% confidence intervals (CI) for the risk of PD. Results were adjusted for comorbidities. Persons with no statin purchases before the three-year lag were the reference group in the main analysis, and the lowest statin exposure tertile in the secondary analysis. All analyses were performed using SAS 9.4 software (SAS Institute, Cary, NC).
Collected data was de-identified thus no ethics committee approval was required according to Finnish legislation.