Introduction

In recent years, zolpidem has been the best-selling hypnotic drug in France. Worldwide, a number of cases of misuse of zolpidem have been described (in Europe and in the USA1 2). The WHO believes that the frequency of cases of abuse or addiction to zolpidem is similar to that associated with hypnotic benzodiazepines.3 As a result, zolpidem has been the target of a number of regulatory framework changes in both French national and international spheres. In particular, the United Nations has placed zolpidem in table IV of the Vienna convention, which aims to control the abuse and trafficking of psychotropic substances (ruling of 15 July 2002).

In France, the French Addictovigilance Network (FAN), piloted by the French National Agency for Medicines and Health Products Safety (Agence Nationale de Sécurité du Médicament et des Produits de Santé (ANSM)), is in charge of the surveillance of cases of abuse and addiction associated with any drug or substance with a psychoactive effect. The surveillance is based on a network of 13 centres for evaluation of and information on drug dependence and addiction monitoring (Centres d’Evaluation et d’Information sur la Pharmacodépendance-Addictovigilance (CEIP-A)), which evaluate the addictive potential of a given drug via notifications provided by health professionals4 and via specifically developed pharmacoepidemiology tools.5 6

Some controlled medicines and psychotropic substances (including zolpidem) are under reinforced surveillance by the ANSM as they are associated with a risk of misuse and addiction. In France, zolpidem is enlisted on the list I of harmful substances, that is to say, it is considered as a substance associated with a health risk. An initial national survey of the addictovigilance network in 2002 found serious and worrying cases of abuse and addiction to zolpidem. The 2002 survey revealed the existence of two consumer groups: a population of chronic high dosage consumers with a therapeutic usage of zolpidem and a population of ‘misusers’ in search of an effect other than hypnotic (euphoria, well-being or stimulant effect). The same survey also found, via the analysis of the FAN pharmacoepidemiology tools, that zolpidem is a substance prone to abuse.7 Following this conclusion, the Summary of Product Characteristics (SPC) of zolpidem was modified, with notably the addition of a warning with respect to addiction. In June 2011, an update of data relative to the addictive potential of zolpidem found the same two consumer groups as in the 2002 survey with cases of increasing severity associated with the consumption of particularly high dosages.8 In light of these results, the prescription of zolpidem on special secure prescription pads was put forward by the National Commission of Narcotics and Psychotropic Substances (Commission Nationale des Stupéfiants et Psychotropes (CNSP)). In 2012, the FAN tools all, once again, proclaim zolpidem as a problematic substance.

The surveillance tools of the FAN allow for the identification of the problem of addiction in specific population groups, but they do not provide a general population risk profile. However, the analysis of quantitative data, in the French National Health Insurance database, relative to the usage of zolpidem and zopiclone in the general population,9 provided the identification of a number of different clinical profiles of zolpidem consumers: (1) ‘non-problematic’ consumers, the largest group; (2) individuals who could have developed a tolerance to the hypnotic effects of zolpidem, for whom the prescription of alternative hypnotic/anxiolytic medication is justified; (3) potential problematic consumers of zolpidem (1%) (high rate of fraudulent behaviour, excessive usage, non-respect of guidelines and medical-pharmaceutical nomadism). In 2017, another research programme gave insight into the characteristics of the two aforementioned consumer groups via the analysis of reports from health professionals.10

Following these results, on 11 January 2017, the ANSM decreed that as from 10 April 2017, the prescription of zolpidem was to be done on secure prescription pads.11 The ANSM stated that ‘this measure is taken in order to limit the risk of abuse and misuse’ and ‘to encourage correct usage’ . In a country where the consumption of psychotropic drugs is high, a ruling that impacts the most sold hypnotic drug12 13 will disrupt not only its usage but also on a larger scale the overall prescription of sedative substances (hypnotics and anxiolytics). The current project aims to develop a means to measure the impact of this new ruling, in the scope of works of the ANSM, that is to say, in terms of the reduction of the risk of abuse, the improvement of correct use of zolpidem and the change in prescriptions of sedative molecules. This project forms part of the evaluation of zolpidem done by the Nantes CEIP-A, the organisation in charge of its follow-up. In addition to the tools used by the FAN,5 this project will allow for a longitudinal evaluation of the trajectories of different patients, as well as providing insight into the general population.

Methods and analysis

Statistical methods

All variables will undergo a descriptive analysis. Quantitative variables will be described using usual position (mean or median) and dispersion (SD, IQR) parameters. The normality of their distribution will be assessed numerically (normality test) and graphically. For normally distributed quantitative variables, mean and SD will be used. For non-normally distributed variables, median and interquartile ranges will be used. Qualitative variables will be described using number and frequency tables for each parameter. All analysis will be conducted with SAS software (version 9.4). Specific statistical methods will be implemented in order to answer each question adequately.

Impact of the measure on the number of consumers: estimation of prevalence and incidence of zolpidem users within the SNDS database before and after the regulatory framework change

A number of periods will be studied (figure 1). The proportion of patients having received at least one delivery of zolpidem during the period 2 and during the period 4 will be compared using a McNemar test for paired proportion. Patients missing from one of the two periods will be recorded as non-users. The significance threshold will be fixed at 5%. An incident user will be defined as a patient receiving a first delivery of zolpidem without any prior delivery over the preceding 6 months. The number of incident users within each period will be compared using a Poisson model. The significance threshold for each coefficient will be fixed at 5%.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Periods of study of the Système National des Données de Santé (SNDS) database.

Impact on the type of consumption regarding changes in treatment duration

Within the SNDS database

The length of the first treatment episode will be evaluated by calculating the number of days covered by the initial delivery (theoretical length of treatment). The predicted variable will be the duration of treatment over a threshold (yes/no). The period will be entered as a covariate in the model, enabling the study of the effect of the period on the probability of the treatment being chronic while taking into account the correlation of treatment characteristics for a given patient.

Within the field study among problematic consumers of zolpidem

A descriptive analysis will be performed, with the characterisation of the duration of treatment with zolpidem before the regulatory framework change and the duration of treatment with zolpidem or of the replacement drug/substance after the regulatory framework change.

Impact on the type of consumption addressing changes in the type of consumption, problematic or non-problematic

Within the SNDS database

A latent class analysis will be conducted within each period, including the following variables: age, sex, presence of a chronic disease, poor economic status, prescribing practitioners specialty (only whether or not a general practitioner), number of different prescribing practitioners (doctor shopping), number of dispensing pharmacies (pharmacy shopping), excess use (mean monthly medication possession ratio19 >1 during the period), adherence to French good practice guidelines regarding hypnotics (encompassing the absence of association with others benzodiazepines) and presence of an associated psychiatric disorder (identified by concomitant drug use, ie, opioids substitution treatments, psycholeptic and psychoanaleptic drugs). The analysis will be repeated during periods 2 and 4. In order to study the transitions between clusters over time, a latent transition analysis will be performed. The choice of the best model will be made considering Bayesian Information Criterion (BIC). The choice of the best model will also be made with consideration to the stability of the model (proportion of convergences among the 5000 iterations), the BIC (lower is better) and the interpretability of the model.

Within the field observational study among problematic consumers of zolpidem

A descriptive analysis will be performed. We will compare the number and the distribution of the positive criteria of problematic consumers (patients and users) before (for zolpidem treatment) and after the coming into force of the new regulatory framework (for zolpidem treatment or replacement substances): duration of consumption, dosage, manner in which zolpidem or other substance is obtained, route of administration or pursued effects different from the expected effect of the treatment or substance. Parametric or non-parametric paired tests will be used for comparisons, according to the distributions of each variable. For each hypothesis test, an alpha risk of 5% will be used. In case of multiple testing, a correction of the significance threshold will be applied to avoid alpha risk inflation (Hochberg’s method).

Impact on the consumption of other sedative molecules: analysis of prescription deferrals and switches

Within the SNDS database, regarding characterisation of consumption trajectories

First, a time series analysis will be performed on aggregated monthly data (proportion of users per month) to compare the changes in the consumption of zolpidem and other sedatives across all the study periods. Cross-correlation between the different time series will be studied in order to identify if zolpidem users have shifted their consumption to other drugs since the change in regulatory framework. Second, a sequence analysis will be performed using dedicated tools (TraMineR, SeqHMM and arulesSequences packages in R software). This will include a cluster analysis of the sequences, in order to identify typical trajectories in consumption and their modification following the change in the regulatory framework.

Within the field study among physicians

A descriptive analysis of changes in prescription behaviour and motives for change will be performed.

Within the field study among problematic consumers of zolpidem

A descriptive analysis of the number of molecules tried as a replacement and the molecules that best replaced zolpidem, if applicable, after the change in regulatory framework, will be performed. In patients stopping zolpidem, but switching to another molecule, a univariate analysis of the same variables will be conducted in order to describe the use of zolpidem before the coming into force of the new regulatory framework and the use of other sedative drugs used in place of zolpidem after the coming into force of the new regulatory framework. For each hypothesis test, an alpha risk of 5% will be used. In case of multiple testing, a correction of the significance threshold will be applied to avoid alpha risk inflation (Hochberg’s method).

Discussion

The evaluation of the addictive power of zolpidem by the addictovigilance network required over the past, above and beyond the tools of the CEIP-A, the implementation of specific research programmes. The evaluation of the impact of the change in the regulatory framework will similarly require the implementation of specific research programmes. This project offers a design and a methodology, which are complementary to the tools of the CEIP-A,4 indispensable to the measurement of the impact, which we believe to be major, of a change in the prescription requirements of the most sold hypnotic in France. The ZORRO project aims to develop a method to measure the impact of the change in regulatory framework on practitioners’ prescriptions of sedative molecules. This evaluation is complex, as in order to be through it must precisely measure the different aspects of the consequences of the regulatory framework change, both from a quantitative and a qualitative point of view. Rather than doing a single study, we prefer to employ a strategy based on a number of different, and complementary, methodological approaches. We have anticipated some possible bias: for the analysis of information from a database, we chose the data from the SNDS database as it contains information close to that of the real consumption. In the absence of available data on the drugs actually taken by patients, this database provides information on the drugs that patients obtain from pharmacies, which is more accurate than sales data, for example. Concerning the periods of reference before and after, we have voluntarily chosen periods distant from the times of announcement and the coming into force of the change in regulatory framework in order to minimise bias linked to the transition period. Although the regulatory framework change came into force in April 2017, the ANSM had published information on the measure as from January 2017. A part of the practitioners prescribing zolpidem therefore anticipated the change in the regulatory framework and started to change their prescription strategy as from January 2017. One of the limits of the SNDS database, that justifies our multimodal approach, is the complete absence of clinical information concerning the effects pursued or felt by patients, as well as the modification in routes of administration. Concerning the in-field clinical study, the principal bias is a memory bias of the questioned subjects. The time between the change in the regulatory framework and the implementation of the study is however incompressible as it is necessary to give patients and users sufficient hindsight in order to evaluate the changes in their consumption of zolpidem. In fact, the questions have been formulated in a simplified manner, and our project targets the most problematic consumers, who should remember with little difficulty the changes, having an impact on their daily lives, following the new regulatory framework. A possible declarative bias does exist among patients and users, although this bias was taken into account, in order to minimise it, in the conception of the questionnaire. On the one hand, the questionnaire is completely anonymous. On the other hand, for patients, the questionnaire is filled-in away from any medical presence and handed back in a sealed envelope that is opened only by the Nantes CEIP-A for data analysis. The CEIP-A personnel are accustomed to interviewing subjects recruited via specialised care centres dedicated to drug dependence on their substance consumption, in the framework of their mission of surveillance of addition and abuse of psychoactive substances. Our personnel have therefore developed an expertise in the realisation of projects of this type among the users of specialised care centres dedicated to drug dependence. The strengths of this project lie within its’ multimodal approach. It allows, on the one hand, to document numerous possible consequences of the regulatory framework change and, on the other hand, to insure the overall coherence of the different studies via the management by an expert team in the field, coordinated by the French national reference centre on the addictive potential of zolpidem. This project may very well have a double impact: on the one hand, it will provide additional data essential to the ANSMs’ mission of surveillance of the risk relative to overdose, abuse, addiction and misuse of sedative substances; on the other hand, this project could be the defining point of a series of steps (eg, communication and information campaigns) designed to manage the public health issues surrounding zolpidem and to measure their overall impact.

Ethics and dissemination