After marketing, drugs should be evaluated for safety and for efficacy. Present national systems are inadequate for both tasks. Safety can be evaluated by means of nonexperimental research, whereas evaluation of efficacy in a variety of settings representing normal medical practice generally requires experiments, randomized and blinded. Research into safety requires the maintenance of routine, yet flexible, multipurpose data systems, different for established and newly marketed drugs. Ad hoc studies on safety and on efficacy can be mounted more swiftly and economically through the maintenance of "standby" capabilities, in addition to maximal use of resources outside the actual national system.