Change in the regulatory framework of zolpidem: Assessing the impact on the prescription of hypnotic and sedative molecules using state sequence analysis

• Published in: Revue d'épidémiologie et de santé publique Vol. 68; p. S108
• Main Authors: Istvan, M., Caillet, P., Rousselet, M., Gérardin, M., Guerlais, M., Jolliet, P., Victorri-Vigneau, C.
• Format: Journal Article
• Language: English
• Published: Elsevier Masson SAS 01.09.2020
• Subjects: Clustering; Drug regulation; Hypnotics; State sequence analysis; Trajectories; State sequence analysis; Trajectories; Hypnotics; Clustering; Drug regulation
• ISSN: 0398-7620
• DOI: 10.1016/j.respe.2020.03.036

In recent years, zolpidem has been the most prescribed hypnotic drug in France and has been the subject of numerous reports of misuse, abuse and dependence. In view of these risks, the French drug agency (ANSM) decreed in April, 2017 the implementation of secure prescription pads. The objective of this study was to evaluate the impact of the measure on the prescription of hypnotic/sedative and anxiolytic molecules by examining the treatment trajectories and the factors associated with these trajectories. We conducted an observational prospective study usizng the Generalist Sample of Beneficiaries (EGB) (a permanent sample of the French healthcare system database (SNDS)). Our study consisted of all patients aged over 18 years old who started zolpidem treatment in 2016 and who were long-term users before the measure (at least one reimbursement for zolpidem in the three consecutive months: October, November and December, 2016). We studied reimbursement trajectories for the psychoactive drugs most likely to be prescribed after zolpidem was stopped up to two years after the measure (from January 2017 to December 2018). We performed a state sequence analysis using the TraMineR R library. We computed pairwise distances using optimal matching method and performed agglomerative hierarchical clustering to identify clusters of treatment trajectories. We set the number of clusters using the fall in inertia and evaluated the clustering quality using the average silhouette width (ASW) and the Hubert's C index (HC) implemented in the WeightedCluster R library. We then assessed the association between individual characteristics and the patterns of treatment trajectories using univariate and multivariate analyses. Overall, we analysed 2503 treatment trajectories of long-term users before the measure. A four-cluster typology was found with a good quality (ASW=0.45 and HC=0.08). The four patterns of treatment trajectories were: continuation of zolpidem (46.3%), discontinuation of hypnotic/sedative and anxiolytic molecules (33.3%), changed to zopiclone (14.5%) or hypnotic benzodiazepines (5.9%). We identified individual characteristics such as age and sex that were associated with the different patterns of treatment trajectories. To our knowledge, this is the first study using state sequence analysis to identify clusters of drugs reimbursement trajectories in the French national health insurance database. State sequence analysis is a promising method in pharmacoepidemiology to study different patterns of drug use in the general population.