Study design

This repeated cohort study used data from the French health insurance database sample ( Échantillon généraliste des bénéficiaires [EGB]). The inclusion period for the 2013 cohort was from 1 February 2012 to 31 January 2013 and the follow-up period was from 1 February 2013 to 31 October 2013. To provide baseline information on trends in statin use in the French population, two reference cohorts corresponding to the two years previous to the media event (2011 and 2012) were constituted using the same design and the same definition of inclusion and follow-up periods ( Fig. 1 ).

Study periods of the 2013 cohort and the two reference cohorts. For each cohort, the inclusion period corresponds to the year preceding the 31 January and the follow-up period was from 1 February to 31 October (9 months).

Data source

The data were extracted from the EGB, a 1/97th representative sample of the population covered by the French national healthcare insurance system . This database contains individual anonymous information on:

• •
  

  sex, year of birth, area of residence;

  
  
• •
  

  outpatient reimbursed healthcare expenditures, such as drugs with date of dispensing, drug strength and quantity dispensed;

  
  
• •
  

  registration for a list of 30 long-term disease groups (LTD) with date and ICD-10 (International Classification of Diseases, 10th revision) code, which concern chronic conditions for which all healthcare expenses are fully reimbursed;

  
  
• •
  

  hospital-discharge summaries (from the Programme de médicalisation des systèmes d’information ), with ICD-10 codes for main, related and associated diagnosis, the date and duration of hospitalization, medical procedures;

  
  
• •
  

  date but not cause of death, from the national death registry.

Study population

In each of the three cohorts, patients were included if they were regular users of statins during the year before inclusion (31 January) and were alive at the start of the follow-up period (1 February).

Regular users of statins (Anatomical Therapeutic Chemical ATC codes: C10AA and C10B) were defined as patients with a proportion of days covered (PDC) > 80% during the year before inclusion. The PDC corresponds to the number of days in possession of statins divided by the number of days of follow-up. The number of days in possession of statins was calculated assuming use of one tablet per day. In the case of two overlapping exposure periods, the days of overlap were cumulated only if the two dispensations concerned the same statin. Owing to a potential lack of data concerning statin exposure during hospitalization, all periods of hospital stay were considered as exposed if the patient was in possession of drugs at the date of hospital admission .

Regular statin users were classified into three groups according to their cardiovascular risk during inclusion periods. The high cardiovascular risk patients had coronary heart disease (LTD No. 13 or hospital admission with main diagnosis ICD-10 codes I20.0, I21 or I24 ), or ischaemic stroke (LTD No. 1 – stroke – associated with ICD-10 code I63 or hospital admission with main diagnosis ICD-10 code I63). Moderate cardiovascular risk patients had at least one of the following: diabetes (LTD No. 8 or hospitalization with main diagnosis ICD-10 codes: E10, E11, E13 or E14), dispensation of antidiabetic drugs, antihypertensive drugs, anticoagulants (vitamin K antagonists or direct-acting anticoagulants) or antiplatelet agents. Low-risk patients included all other regular statin users not included in the first two groups.

Outcomes

The main outcome was the occurrence of statin discontinuation, measured as the occurrence of a gap of at least 2 months without statin exposure. Periods of statin exposure were estimated by number of days in possession of statins after each dispensation as defined in the eligibility criteria. The date of discontinuation corresponded to the estimated last day of treatment covered by the last dispensation before the gap.

Statistical analyses

The three cohorts were described according to age at inclusion, sex, comorbidities, cardiovascular treatments and cardiovascular risk groups. Statin discontinuation in the total population and among the three cardiovascular risk groups was described using Kaplan–Meier survival analysis. Discontinuation rates were estimated at 9 months of follow-up. Relative risks were the ratio of statin discontinuation or mortality rates between 2013 and 2012, and between 2012 and 2011, with their 95% confidence intervals (CIs).

All analyses were performed using SAS ^® software (SAS Institute, version 9.4, North Carolina, USA).

Ethical approval

In accordance with regulations in place at the time of the study, the National Institute of Health and Medical Research ( Institut national de la santé et de la recherche médicale ) was informed of the study that was to be performed using the EGB database. There was no further requirement for ethical approval or data protection agency approval for this study, which was done in fully anonymized data.