The role of Meta-analysis in Pharmacovigilance: how best to combine different sources of evidence about harms?

Abstract

The assessment of the benefit/risk relation is conducted throughout the entire drug life cycle. Before a market authorization is granted, randomized clinical trials are designed to evaluate the efficacy and safety of a drug in a specific therapeutic indication. These studies are able to detect the most frequent adverse events. However, rare and/or long-latency harmful events are usually detected after a drug becomes available in the market. The increased seriousness of some adverse events may lead to label update with warnings or even to a drug withdrawal after being marketed for some years. Post-marketing observational studies may better reflect the nature of adverse events occurring in clinical practice since they include populations usually underrepresented in clinical trials, such as the elderly, pregnant women or patients with comorbidities. The investigation of uncommon or long-term adverse events associated with pharmacological interventions has been discussed as a potential important application of meta-analysis. Meta-analysis is a systematic approach to synthesize and combine the results of selected studies. It is used to identify sources of variation among study findings and to provide an overall measure of effect to reach conclusions about a body of research. The meta-analytic technique has been applied with increasing frequency to clinical trials when efficacy assessments are needed. Although not frequently, meta-analysis conducted for safety purposes have also found increased risk estimates associated with some drugs, such cardiovascular adverse events due to rosiglitazone. This project carried out in order to identify the role of meta-analysis as a Pharmacovigilance approach and to evaluate how best to combine safety information from both experimental and observational studies through this statistical technique. Only a limited number of meta-analyses are currently devoted to evaluate drug safety as a primary outcome. Of these, very few combine data from both observational and experimental studies. Although statistical significant risk estimates could be reached with the inclusion of observational studies in meta-analysis, isolated or in combination with clinical trials, the increased between-studies heterogeneity usually associated may preclude any definitive conclusions. Authorities do not rely solely on risk estimates produced by meta-analysis and usually review additional sources of information to support benefit/risk ratio reevaluations due to safety issues. It was also demonstrated that cumulative meta-analysis was able to estimates increased iatrogenic risks years earlier than regulatory decisions have been taken by authorities for the corresponding safety issues. However, excessive heterogeneity resulting from different study designs included in these set of meta-analyses may be one of the reasons delaying the acceptance of this technique by regulatory authorities when evaluating medicines safety profile. Although reliable risk estimates have shown to be produced from meta-analyses conducted to evaluate drug safety issues, between-studies heterogeneity may not let drawing robust conclusions from those results, in particular when observational studies are included. The results of this work do not let recommend that a meta-analysis of the existing evidence should be conducted whenever a safety alert is issued. Moreover, this technique does not replace further assessments when the benefit/risk ratio profile of a medicine needs to be revised due to safety issues.