Thirty-seven children, in whom celiac disease had been diagnosed because of flat mucosa on a gluten-containing diet and recovery on a gluten-free diet, were challenged with gluten powder, 10 g/day, in addition to an otherwise gluten-free diet. A small intestinal biopsy was performed before the challenge; clinical symptoms, a 1-h blood xylose test, and gliadin antibody measurement were used to establish the timing of the confirmatory biopsy. All but one case relapsed within 205 days (mode, 60 days). In no case was the relapse clinically evident. Raised levels of gliadin antibodies, a fall in xylose absorption, or both predicted the relapse in 37%, 7%, and 57% of cases, respectively. Evaluated individually, each test gave a considerable rate of false negative results. Discriminant coefficients produced for each test were used to compute a score that allowed the classification of patients into relapse/no relapse categories with a good degree of accuracy. The discriminant score rose sharply after only 15 days of challenge, indicating that it is possible to predict the relapse long before any clinical symptom appears.