Non-steroidal anti-inflammatory drugs (NSAIDs) effectively control the symptoms of many of the rheumatic diseases although they have little effect on the underlying causes. Their effect is mainly on the mediators of the inflammatory process. Unfortunately, these mediators have important physiological roles in the maintenance of health, particularly in the gastrointestinal tract and the kidney, so that their inhibition results in many unwanted reactions of varying severity. The mechanisms underlying these reactions are described. Their occurrence varies, both qualitatively and quantitatively, and an attempt is made to assess these differences, although it may be that they are related directly to differences in dosage and therapeutic efficacy. In addition, immunologically mediated adverse reactions occur. These mechanisms are outlined and related to the clinical picture. There are considerable differences in frequency of reactions between the compounds: in particular there is a wide variation in the rate of dermatological reactions of this type. Agranulocytosis has been particularly associated with the pyrazolone compounds, although it has been reported with most others. Aplastic anaemia, which may not be an immune-mediated reaction, is also thought of as a pyrazolone reaction, but the incidence with indomethacin approaches a similar level. Although all drugs analysed may cause hepatic reactions, these are rare except with the now withdrawn benoxaprofen. Several types of immunologically mediated renal reactions occur and these rarities are also described. Paracetamol does not have any effect on the inflammatory mediators. Anxieties about this substance relates to the parent compound phenacetin and its necrotic effect on the renal papillae. There is extensive literature on this subject concerning not only paracetamol, but also aspirin and other NSAIDs. This is also assessed and summarised. The danger of paracetamol as a direct hepatic toxin in self-poisoning is discussed. Novel NSAIDs are introduced and others withdrawn with frequent and monotonous regularity. Sometimes the reasons have some medical or scientific plausibility, but often they are over-reactions by registration authorities or pharmaceutical companies in response to uninformed media publicity. The problems of the numerically and scientifically accurate collection and assessment of adverse reaction data are legion and as a result useful agents have been lost. Some of these difficulties are described, and some non-drug 'adverse reactions' are described.(ABSTRACT TRUNCATED AT 400 WORDS)