Examples of Use of Psychopharmacoepidemiology

Clinical trials are usually done in selected populations for limited durations. They are invaluable to establish that a drug may indeed be effective and reasonably safe, but they are only vaguely and hopefully predictive of what may really happen when the drug is marketed in larger, unselected populations. That can only be studied using population-based methods that influence as little as possible the real-life use of the drugs. In that context, pharmaco-epidemiology will be useful to study patterns of drug utilisation, risks associated with the drugs, and the possible utility of the drug on a medical and/or economic basis, to compare with the predictions from the premarketing clinical trials.

The methods used cover the whole range of epidemiology, including cross-sectional studies, case-control and cohort studies and large-scale simple pragmatic trials. We will give examples of each of these from actual studies done or under way in our department:

For instance, we did a telephone survey of a representative sample of the French population, to ascertain benzodiazepine usage and the related psychiatric diagnoses, using the Mini interview, in over 4000 subjects, showing that benzodiazepines were used by 9% of thepopulation, for more than 6 months in 88% of those above 60, and that 25% of the patients with severe depressive episode (7% of the population) were using them. In another study, psychiatrists and GPs were asked to rank the antidepressants for safety and efficacy, then give the treatment history of the present episode in 4 depressed patients each. 247 psychiatrists and almost a thousand patients were studied, and the results show that overall antidepressants are more effective and better tolerated than prescribers believe, when efficacy and tolerability were measured by rate of treatment discontinuation for lack of effect or poor tolerability. In addition, the range of adverse symptoms experienced by treated and untreated patients were described, confirming the typical anticholinergic pattern associated with the tricyclic antidepressants. Studies can also look for specific safety-related endpoints. For instance, we did a case-control study of the association of benzodiazepine use and hip fracture in the elderly, in over 300 cases and 700 controls, and found no consistent association, though some individual drugs seem at greater risk. Other case-control studies are being done to study the role of benzodiazepines in workplace accidents and in selected traffic accidents. We also studied cause-specific mortality in a cohort of over 3000 schizophrenic patients followed 4 years, and confirmed the predominance of suicide as the main cause of death in the initial years. In a nested case-control study, higher rates of suicide deaths were found in users of thioxanthenes, and of non-suicide, non-cardiovascular deaths with atypical neuroleptics, though it is not possible to state with certitude whether this was related to the drug or to patient selection biases.

Pharmacoepidemiological studies can explore alerts from spontaneous reporting: sertindole was suspended from the market because of a suspicion of excess sudden death reported in the UK. Careful analysis of reporting patterns, and further epidemiological studies such as prescription-event monitoring in the UK, and a European-wide post-marketing cohort study showed that this could be explained by selective reporting biases. The death rates were not higher during sertindole treatment than after switching to other neuroleptics. Along with other data, this resulted in the readmission of sertindole to the market. A large-scale simple randomized controlled survival trial may give a definitive answer to this question. This large scale simple trial methodology which has been used in other contexts is made possible by recent evolution in data management technology.

Though the conventional clinical trial will remain the cornerstone of drug licensing for the foreseeable future, post marketing confirmation of drug effects, both positive and negative, will become the norm, relying on the full array of modern pharmacoepidemiological methods.