Title : Flexible analyses of population-based pharmaco-epidemiological studies of comparative effectiveness and safety of drugs
Recent large governmental initiatives e.g. Canadian Drug Safety and Effectiveness Network or the US Sentinel program, emphasize the need to assess the real-life effectiveness and safety of drugs in populationbased studies. Increasing availability of large databases, generated through electronic health records makes such studies more and more popular, and enhances their impact on the policy makers decisions regarding which drugs are withdrawn from the market, as well as on clinicians decisions regarding how and to whom the specific drugs are prescribed. Whereas such large database studies ensure adequate statistical power, they typically rely on conventional statistical models, and use simplistic metrics of drug use, e.g. current dose, or any use in the past 3 months. Yet, these metrics ignore important variation in the patterns of drug use in real-life clinical practice, where daily doses, treatment duration, frequency of treatment interruptions, all vary substantially between subjects and within-subjects over time. We present both simulation-based and real-life data illustrating how the above limitations of the models used in comparative effectiveness and safety studies seriously hamper the ability to detect the important benefits or adverse effects of a drug. We then demonstrate how our new, flexible statistical methodology, that accounts for variation in drug use patterns and for cumulative effects of past drug use, helps assessing accurately the causal effects of drugs in population-based studies and produces results consistent with the drug pharmaco-dynamics/-kinetics characteristics. The advantages of our new methodology are illustrated with empirical pharmacoepidemiological studies of the associations between: (1) benzodiazepines and fractures, (2) oral glucocorticoids and infections, (3) thiazide diuretics and cardiovascular disease.
Audience Take Away:
• This talk will help the audience understand the analytical challenges involved in assessing real-life safety and effectiveness of drugs and, as a consequence, improve the accuracy and validity of their own studies.
• By becoming more aware of the methodological issues arising in post-marketing studies of the drugs, the audience will be able to critically assessed the methods and conclusions of published studies.
• For those conference participants that may be (now or in future) actively involved in post-marketing studies, I will provide concrete information that will allow them to both (a) better design their analyses, and (b) use the user-friendly, publicly available software that implements our new state-of-the-art methods.