This study investigates early onset of treatment response as predictor of symptomatic and functional outcome 3 years after initiation of methylphenidate (MPH) administration in a naturalistic, clinical cohort of children and adolescents with ADHD. Children were followed across an initial 12-week MPH treatment trial and after 3 years, with ratings of symptoms and impairment. Associations between a clinically significant MPH treatment response in week 3 (defined as ≥ 20% reduction in clinician-rated symptoms) and in week 12 (defined as ≥ 40% reduction), and 3-year outcome were tested in multivariate linear regression models, adjusting for sex, age, comorbidity, IQ, maternal education, parental psychiatric disorder, and baseline symptoms and function. We did not have information on treatment adherence or the nature of treatments beyond 12 weeks. 148 children, mean age 12.4 years (range 10–16 years), 77% males, participated in the follow-up. We found a significant decrease in symptom score from baseline [M = 41.9 (SD = 13.2)] to 3-year follow-up [M = 27.5 (SD = 12.7), p < 0.001, and in impairment score from baseline (M = 41.6 (SD = 19.4)] to 3-year follow-up [M = 35.6 (SD = 20.2), p = 0.005]. Treatment responses in week 3 and week 12 were significant predictors of the long-term outcome of symptoms, but not of impairment at 3-year follow-up, when adjusting for other well-known predictors. Early treatment response predicts long-term outcome over and above other well-known predictors. Clinicians should follow-up patients carefully, during the first months of treatment, and detect non-responders, since there might be a window of opportunity to alter the outcome, by changing the treatment strategy.