What is the optimal treatment for metastatic brain tumors (MBTs)? We present our experience with gamma knife (GK) treatments for patients with five or more MBTs. Our new formula for predicting patient survival time (ST), which was derived by combining tumor control probability (TCP) calculated by Colombo's formula and normal tissue complication probability (NTCP) estimated by Flickinger's integrated logistic formula, was also evaluated. ST=a*[(C-NTCP)*TCP]+b; a, b, C: const. Forty-one patients (23 male, 18 female) with more than five MBTs were treated between March 1992 and February 2000. The tumors originated in the lung in 15 cases, in the breast in 8. Four patients had previously undergone whole brain irradiation (WBI). Ten patients were given concomitant WBI. Thirteen patients had additional extracranial metastatic lesions. TCP and NTCP were calculated using Excel add-in software. Cox's proportional hazards model was used to evaluate correlations between certain variables and ST. The independent variables evaluated were patient factors (age in years and performance status), tumor factors (total volume and number of tumors in each patient), treatment factors (TCP, NTCP and marginal dose) and the values of (C-NTCP)*TCP. Total tumor number was 403 (median 7, range 5-56). The median total tumor volume was 9.8 cm3 (range 0.8-111.8 cm3). The marginal dose ranged from 8 to 22 Gy (median 16.0Gy), TCP from 0.0% to 83% (median 15%) and NTCP from 0.0% to 31% (median 6.0%). (0.39-NTCP)*TCP ranged from 0.0 to 0.21 (median 0.055). Follow-up was 0.2 to 26.2 months, with a median of 5.4 months. Multiple-sample tests revealed no differences in STs among patients with MBTs of different origins (p=0.50). The 50% STs of patients with MBTs originating from the breast, lung and other sites were 5.9, 7.8 and 3.5 months, respectively. Only TCP and (0.39-NTCP)*TCP were statistically significant covariates (p=0.014, 0.001, respectively), and the latter was a more important predictor of ST than the former (Beta= -2.2, -14.1, respectively). The relationship between (0.39-NTCP)*TCP and ST was significant. Linear regression analysis showed this value to predict ST (p=0.002, R2=0.22). The slope of the regression line for patients with MBTs originating from the breast was steeper (a=218.2, p=0.08, R2=0.41) than the slopes of regression lines for patients with tumors of other origins (lung; a=56.8, p=0.004, R2=0.49, others; a=50.4, p=0.03, R2=0.25). In treating multiple lesions, the maximum doses and dose distribution for individual lesions were often different. The formula described by Colombo is used to calculate the residual clonogenic cell number of every sub-volume of the tumor, with different doses. NTCP must also integrate every complication probability for each sub-volume of normal brain tissue in the relatively high dose area in proximity to the tumor. Herein, we present a method for determining the irradiation dose necessary for cases with multiple brain metastases. A personal computer-aided calculation is employed.