OBJECTIVE To perform a further Cox proportional hazards logistic regression analysis of data from two large-scale placebo-controlled trials with cisapride as maintenance treatment in reflux disease. RESULTS Analysis of each of the two databases, allowing the model to operate freely, led to the identification of a number of unexpected putative predictors of outcome in the 6 to 12 months following initial healing of oesophagitis with an H2-receptor antagonist or omeprazole. This allowed us to delineate more accurately the patient population that is likely to respond to long-term continuous treatment with low or standard dose cisapride. The analysis revealed that symptom severity may be more useful than endoscopic severity in predicting relapse or in guiding therapy. Reflux oesophagitis outcome is particularly poor in the presence of treatment-recalcitrant symptoms or severe mucosal damage. Analysis showed cisapride to be effective in the maintenance treatment of patients with non-refractory symptoms, irrespective of the initial severity of oesophagitis, the healing agent used, or a history of previous endoscopic relapses.