Children with hyperthyroidism often require prolonged courses of antithyroid medication to achieve remission, and long-term compliance is problematic. To determine which clinical and laboratory features predict early remission, we reviewed the records of 191 patients less than 19 yr old with Graves' disease. We compared patients achieving remission within 2 yr (group 1, n = 27) with those who completed more than 2 yr of medical therapy but did not achieve a remission (group 2, n = 79). Patients who were in neither of the above categories (n = 85) were excluded from the statistical analysis. Variables that were measurable at the time of diagnosis, recorded in more than 50% of the study population and associated with early remission in the univariate analysis (P < or = 0.05), were entered into a stepwise multiple logistic regression analysis. Variables retaining a significant association with early remission (P < 0.05) were considered independent predictors of early remission. Patients achieving early remission were older (mean, 12.5 vs. 10.9 yr, P = 0.039) and had higher body mass indexes (BMI, 19.0 vs. 16.6, P = 0.002), higher BMI SD scores (-0.03 vs. -0.60, P = 0.004), lower heart rates (110 vs. 121, P = 0.023), smaller goiters (group 1: 60% with moderate/large goiter; group 2: 83%, P = 0.050), lower platelet counts (272 vs. 339 K/microL, P = 0.006), lower serum T4 and T3 concentrations at presentation (T4: 18.3 vs. 22.5 microg/dL, P = 0.015; T3: 439 vs. 613 ng/dL, P = 0.008), and were less likely to have a positive test for thyroid stimulating Igs (group 1: 50% vs. group 2: 93%, P = 0.008). Regression analysis identified BMI SD score and goiter size as independent predictors of early remission (P < 0.05). Eighty-six percent of patients with BMI SD score above -0.5 SD and minimal/small goiters achieved early remission, compared with 13% of those with BMI SD score below -0.5 SD and moderate/large goiters. We conclude that, of multiple clinical and laboratory variables associated with early remission, BMI SD score and goiter size are independent predictors. Algorithms employing these two variables can be used to facilitate counseling of patients and expedite therapeutic decisions.