Elemental diet is as effective in producing remission of Crohn's disease (CD) as is corticosteroid treatment, but most patients relapse soon after resumption of a normal diet. We have investigated the efficacies of dietary modification and oral corticosteroids in maintaining remission achieved with elemental diet. In a multicentre trial, 136 patients with active CD were started on elemental diet and other treatment was withdrawn. 43 (31%) declined to continue elemental diet for 14 days, but 78 (84%) of the remaining 93 achieved remission and were randomly assigned corticosteroids (38) or diet (40). Corticosteroid treatment started at 40 mg prednisolone daily, which was tapered and stopped after 12 weeks; that group received dietary advice on healthy eating. The diet group received "tapered" placebo and were instructed to introduce one new food daily, excluding any that precipitated symptoms. Assessment of progress for up to 2 years was made by physicians unaware of group assignment. Intention-to-treat analysis showed median lengths of remission of 3.8 (interquartile range 5.0) months in the corticosteroid group and 7.5 (15.3) months on diet, and relapse rates at 2 years, adjusted for withdrawals, of 79% and 62%, respectively (p = 0.048). Clinical improvement in the diet group was associated with significant changes in plasma albumin and alpha 1-antichymotrypsin concentrations and erythrocyte sedimentation rate. Food intolerances discovered were predominantly to cereals, dairy products, and yeast. Diet provides a further therapeutic strategy in active Crohn's disease.