Statistical Aspects of Monitoring for Dangers in Drug Therapy)

Abstract

A therapeutic drug has a complicated life history, and some of the important stages of this are described in the paper. Before it can even be referred to as a drug, it must be identified as a chemical compound worth investigation as possibly having useful pharmacological properties. From among large numbers in this category, a few must be selected by a screening process. Toxicity tests and clinical trials each bring their special problems for the statistician. When a new drug has had its usefulness established, standardization of its potency by biological assay may be necessary, and this also uses statistical techniques.Most of the paper is concerned with adverse reactions to drugs when used in medical practice, and methods for detecting these. In response to recent recognition of dangers, systems for monitoring for adverse reactions to drugs that occur during normal therapeutic use have been developed in many countries. The paper outlines the logical basis of monitoring and discusses its practical implementation. New problems of statistical inference, as well as large data handling activities, arise in connexion with the interpretation of evidence that is very heterogeneous in origin. The basic statistical problem appears to be one of collecting and evaluating information that may constitute a degree of suspicion calling for intensive and more objectively planned research. This type of inference seems closer to the mode of thought of a criminal investigation department than to orthodox practice of statistical estimation or decision-making; I think it has new features that deserve further study.National monitoring organizations are now cooperating, through the World Health Organization, in an international project for the combination of evidence.