Statistical methods for analysis of adverse effects associated with drugs
Lead Research Organisation:
London School of Hygiene & Tropical Medicine
Department Name: Epidemiology and Population Health
Abstract
Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.
Technical Summary
Statistical methodology has concentrated on beneficial effects of interventions. Adverse effects have not had the same effort and the methodological problems are much greater.
Adverse effects of drugs are particularly difficult, because it is usually impossible to pre-specify what effect will be found and conventional adjustment methods for multiple comparisons can dismiss important effects in randomised trials. Observational studies can focus on a single adverse effect, but there is a need for better analysis strategies particularly in trials. Such a strategy would also be applicable in observational and ?Yellow card? databases.
A Bayesian approach has been suggested by Berry and Berry (2003) but their implementation has practical and medical difficulties.
This research is intended to develop methods to ?borrow? information based on the medical classification of adverse effects, typically using a dictionary such as MedDRA. Medically similar effects are each more likely than physiologically unrelated effects. Statistical approaches that ignore medical knowledge will not be sensible. This research brings together experts in Bayesian statistics (Spiegelhalter & Whittaker) with clinical expertise (Aronson, Smeeth). The principal investigator has been involved in detection of new adverse reactions from a statistical perspective.
Adverse effects of drugs are particularly difficult, because it is usually impossible to pre-specify what effect will be found and conventional adjustment methods for multiple comparisons can dismiss important effects in randomised trials. Observational studies can focus on a single adverse effect, but there is a need for better analysis strategies particularly in trials. Such a strategy would also be applicable in observational and ?Yellow card? databases.
A Bayesian approach has been suggested by Berry and Berry (2003) but their implementation has practical and medical difficulties.
This research is intended to develop methods to ?borrow? information based on the medical classification of adverse effects, typically using a dictionary such as MedDRA. Medically similar effects are each more likely than physiologically unrelated effects. Statistical approaches that ignore medical knowledge will not be sensible. This research brings together experts in Bayesian statistics (Spiegelhalter & Whittaker) with clinical expertise (Aronson, Smeeth). The principal investigator has been involved in detection of new adverse reactions from a statistical perspective.