OBJECTIVE: To determine long-term growth response to growth hormone (GH) therapy in patients with isolated GH deficiency who had been small for gestational age and in those who had been appropriate in size for gestational age. DESIGN: Longitudinal, case-control study. SETTING: Pediatric clinic, endocrinology center, University of Bologna, Italy. PATIENTS: Sixteen GH-deficient children, small for gestational age with unknown cause, and 16 GH-deficient children, appropriate in size for gestational age, who were matched for chronologic age, bone age, pubertal stage, and target height at the beginning of treatment and were treated for 36 months. INTERVENTION: Recombinant human GH given subcutaneously at a dose of 20 IU/m2 per week in six doses per week for 36 months. MEASUREMENTS: Growth hormone levels (fluoroimmunoenzymatic method), levels of insulin-like growth factor I (radioimmunoassay), and complete 36-months auxologic follow-up. RESULTS: Patients who were small for gestational age had a modest improvement in height for chronologic age but no increase in predicted final height. Patients who were appropriate in size for gestational age had significantly better improvement in both measurements (multivariate analysis of variance: F = 6.3 (p < 0.001) and F = 3.8 (p < 0.05), respectively). Catch-up growth was similar during the first year of therapy for the two groups, after which the linear growth velocity decreased more rapidly in the small-for-gestational-age patients (multivariate analysis of variance: F = 4.9 (p < 0.05)). CONCLUSIONS: The constitutional component of the statural deficiency of small-for-gestational-age children seemed to prevail over hormonal deficiency during treatment with GH. Further follow-up to final height is necessary to evaluate these different responses.