Acromegaly is associated with increased morbidity and mortality unless serum GH levels are persistently less than 5 mU/L ( approximately 2 ng/mL) after treatment. Transsphenoidal surgical resection is the best available treatment for restoring GH to such "safe" levels; however, criteria for the assessment of the response to treatment are not uniform. To determine the clinically most useful method of assessing disease activity postoperatively and identify predictors of a favorable response to surgical treatment, we have analyzed 67 patients with acromegaly who underwent transsphenoidal surgery between 1993 and 1998. We used three different definitions of a satisfactory or safe response: 1) a postoperative mean GH less than 5 mU/L obtained from averaging five serum GH values obtained throughout one day; 2) a random single GH less than 5 mU/L; or 3) a serum insulin-like growth factor I (IGF-I) level within the normal range. Relying on a single GH measurement alone, 9 of the 23 patients with a single postoperative mean GH level less than 5 mU/L obtained at least one GH value of more than 5 mU/L (false positive rate, 28%) and 8 of the patients with a postoperative mean GH value of more than 5 mU/L obtained a single GH value of less than 5 mU/L (false negative rate, 15%). Postoperatively, a significant increase in the fluctuation of random GH values around the mean was observed in patients who were rendered safe (coefficient of variation, from 26 +/- 2% to 53 +/- 6%; P < 0.001) compared with patients with persistence of inadequately controlled disease. However, 13% (3 of 23) of patients with mean postoperative GH levels of less than 5 mU/L had elevated serum IGF-I levels postoperatively, and 17% (8 of 44) of patients with mean serum GH levels more than 5 mU/L had postoperative IGF-I levels within the normal range. There was no difference in the rate of agreement between mean GH less than 5 mU/L and normalization of IGF-I in relation to the interval since operation when IGF-I levels were measured. Preoperative tumor size and pretreatment mean GH levels were the major determinants of the outcome of surgery, as patients who were rendered safe had significantly lower preoperative mean GH levels than patients who were not cured (median, 31 mU/L vs. 78.5 mU/L, P < 0.01). IGF-I levels were weakly correlated with tumor size and could not be used to predict the patients who would be rendered safe. Preoperative PRL levels were higher in patients who failed to achieve a surgical satisfactory outcome [498 mU/L (187-857) vs. 196 mU/L (136-315), P < 0.01]. In summary, although single random GH values and IGF-I values are both significantly correlated with mean GH levels, they should not be used as an alternative to averaging several GH values to assess disease activity, because of the pulsatile nature of GH secretion and the multiple factors that may influence serum IGF-I. Because significant discrepancies occur, particularly postoperatively, mean GH levels remain the more reliable indicator of surgical outcome and disease activity. As there is considerably more evidence relating long-term prognosis to serum GH levels than to IGF-I and discrepancies occur between GH levels and IGF-I, we suggest that mean serum GH levels and single IGF-I levels, measured early in the postoperative period, are currently the best biochemical guide to the adequacy of surgery and, hence, the need for further treatment.