Statistical Logic in the Monitoring of Reactions to Therapeutic Drugs

Abstract

National and international organizations have now been established for the acquisition and interpretation of reports of suspected adverse reactions to drugs. These differ in detail but have the common feature of a strong element of spontaneity in the initiation of reports by individual physicians. The typical report relates to one person and contains information on drugs administered, any adverse events that might be reactions to the drugs, and additional identificatory and classificatory material. Interpretation of these voluntarily submitted records is a problem of statistico-medical detection, more complex and less certain than that of a planned clinical trial despite superficial resemblances. The paper emphasizes again the distinction between patient, drug, and event ascertainment. It then presents examples of the logical argument appropriate to each. For any new problem, quantitative analysis must be allied to thorough consideration of independence, representativeness, susceptibility equivalence, background equivalence, and unbiased reporting of the data. Monitoring can never lead to complete demonstration that a drug causes an adverse reaction; with strict attention to the whole medical background, it can provide a convincing case for caution in therapeutic practice pending the completion of new research directed at an identified problem.