A group of 668 stage II melanoma patients was entered into a randomized prospective study aimed at evaluating the efficacy of adjuvant BCG, 5-(dimethyltriazeno)imidazole-4-carboxamide (DTIC), or a combination of the two, given immediately after radical lymph node dissection. Of these, 176 patients received BCG and 164 BCG plus DTIC. These 340 patients had histologically proven metastatic nodes and 156 had a negative skin reactivity to BCG at the beginning of treatment. The distribution of known prognostic factors (sex, age, number of positive nodes, extracapsular invasion) was balanced in the groups of patients either with initially negative or with positive skin reactivity. All patients who were initially non-reactive to BCG developed skin reactivity after 6.7 +/- 9 BCG vaccinations. Disease-free and overall survival of patients receiving BCG or BCG + DTIC with an initially negative skin reactivity to BCG was significantly (P = 7 x 10(-3) better than that observed in patients with an initial positive skin reactivity. This finding was still evident after adjustment for other known prognostic criteria (P = 0.02). It seems likely that the initial BCG skin reactivity as such marks the prognosis; however, some therapeutic effect of BCG treatment in patients having initially no skin reactivity to BCG, can not be ruled out.