Set and serendipity in the detection of drug hazards.

This paper deals with influences which affect the recognition of drug hazards and examines the policy implications of the analysis. Decision-making about the control of potentially dangerous medicines presents problems for policy makers. Action in this area provokes controversy because the interests of those concerned differ widely and because judgements are made in the face of uncertainty. The problem can be described in terms of models derived from signal detection theory. The signal (an adverse drug reaction, ADR) must be differentiated from background noise (disease due to other causes). This approach directs attention to the two general factors that influence detection: the discriminability of the signal and the observer's operating criterion. How clearly an ADR can be discriminated from other illness largely depends on the nature of the reaction, but the criterion is determined by social and behavioural forces including the costs and benefits of each type of decision. These in turn depend on the interests and attitudes of the institutions involved. It is concluded that detection of drug-induced disease is hampered by the attitudes of those charged with monitoring the effects of drugs. The shared assumption that medicines are good produces an excessively high criterion for the recognition of their dangers.