A review of the trials in which 'disease-modifying' drugs have been tested leads to the following conclusions: (a) Most of these trials do not provide acceptable evidence of the efficacy of these drugs. (b) A well-designed, placebo-controlled, study of gold (Empire Rheumatism Council, 1960, 1961) provides strong evidence that the drug has a beneficial effect lasting about 18 months. It does not establish any advantage beyond that period. Two reports which claim to show a favourable influence of gold on radiological progression are suspect because of faults in trial design. (c) There has been only one placebo-controlled study of penicillamine (Multicentre Trial Group, 1973). This also provides evidence of medium-term efficacy. No information is available about x-ray progression. (d) Studies of other drugs for which 'disease-modifying' activity has been claimed (antimalarials, antiproliferative agents, corticosteroids, etc.) have similarly provided evidence only of medium-term efficacy. With information now available it is possible to identify some of the reasons why these trials have failed to answer the fundamentally important question of whether the drugs can modify the long-term course of RA. (a) Trial designs have concentrated on following process measures (e.g., ESR) rather than outcome measures (e.g., disability and deformity). Observers have thus come to accept the former as being important in themselves, to the neglect of the latter. (b) Psychological pressures to provide relief for patients, combined with unjustified assumptions about the long-term efficacy of these drugs, have produced a climate of expectations amongst both clinicians and patients which makes it difficult to sustain long-term trials against placebo. (c) Traditional radiological assessment methods have proved insensitive. (d) Recent advances in clinical trial design and analysis (developed largely in specialties outside rheumatology) were not available at the time of most of these trials. At this stage clinical trials of 'disease-modifying' drugs should seek to answer the following questions: (a) Is it worth giving drugs such as gold and penicillamine for periods longer than 6 months? (b) How long does the effect of such treatment last? (c) Is any information obtainable before or during treatment which predicts favourable or unfavourable responses? To answer these question, trial designs, will need to include: (a) Larger numbers of patients. (b) Longer duration of treatment and follow up.(ABSTRACT TRUNCATED AT 400 WORDS)