Certain limitations of this study should be considered. First, diet, exercise, and behavioral modification were not standardized. However, the absence of a significant center by treatment interaction suggests that the treatment effect across centers was similar. Second, the study was performed in a predominantly white population and as is common in such studies, most participants were female. The average participant was at the 98th percentile for BMI, a significant degree of obesity, so it is not known if less obese adolescents would achieve similar results. The present study excluded participants with certain characteristics more likely to be associated with metabolic syndrome, although one quarter of the participants did have the metabolic syndrome at randomization. How these characteristics affect the generalizability of our results is not clear. However, recent results in obese adults attending community weight clinics or in general practice were similar to those from double-blind, randomized trials.Third, the number of participants and the study duration do not allow adequate assessment of safety beyond 1 year. Fourth, quality of life was not investigated in this study, making an objective assessment of tolerability difficult. It should be noted that, although similar between groups, the dropout rate was 35% to 36%. This rate is well within that usually seen in obesity trials, particularly those of more than 1-year duration, in which dropout rates range from 10% to 80%. Trials of obesity therapies face the added problem of patients stopping treatment when weight loss plateaus in addition to the common issue of patient perseverance seen in most long-term trials. It would be useful in the future to study other adolescent populations for longer periods.