During a 10-year period 39 patients with acromegaly, aged 23-73 years, underwent selective adenomectomy via a trans-sphenoidal or transfrontal (one case) approach. Six to 12 months after the operation, the serum level of growth hormone (GH) was reduced to less than 5 micrograms l-1 in 28 patients (74%) in at least two of three random samples and/or suppressed to less than 3 micrograms l-1 during an oral glucose load, thus fulfilling the commonly used criteria for a successful operation. In 10 patients these criteria for adequate GH reduction were not fulfilled, but their median S-GH level was reduced from 38 to 11 micrograms l-1 (P less than 0.01) after the operation. Surgery was successful in 11 of 13 (85%) patients with a microadenoma (less than 10 mm in diameter), in 10 of 14 (71%) patients with an adenoma of diameter greater than 10 mm but still enclosed in the sella, and in seven of 11 (64%) patients with locally invasive tumours. Impaired pituitary function was observed in 23% of the patients after surgery, independent of tumour size. In one patient the postoperative period was complicated by a lethal intracranial infection. During follow-up for 1-10 years, four patients relapsed, after 1, 1.5, 6 and 9 years, respectively. Patients for whom surgery appeared to have been ineffective at the evaluation 6-12 months postoperatively, or who later relapsed were identified by early (within 7 d) postoperative serum GH with a sensitivity of 90%. The accuracy for identification of a satisfactory outcome of surgery was 85%, and the predictive value was 90%. The corresponding values for the GH response to TRH measured 6-12 months postoperatively were 47, 40 and 54%, respectively. It is concluded that the basal level of serum GH measured 1-7 d postoperatively has higher sensitivity and specificity than the GH response to TRH 6-12 months postoperatively for evaluation of the effect of surgery on GH overproduction, and that it has a higher predictive power with regard to the long-term outcome of surgery for acromegaly.