This study was performed to assess the value of prolonged, as opposed to short-pulse, in vitro exposure of leukemic cells to chemotherapeutic drugs in leukemic clonogenic assay for prediction of clinical response. In 21 patients with acute nonlymphocytic leukemia treated with intensive combination chemotherapy based on an anthracycline and 1-beta-D-arabinofuranosylcytosine infusion, chemotherapy sensitivity of leukemic clonogenic cells was assessed in comparison with that of normal myeloid clonogenic cells by the in vitro continuous exposure to drugs throughout the entire culture period. Analysis of these in vitro data in terms of prediction of achieving clinical complete remission was carried out in comparison with data on 22 cases in which in vitro sensitivity was assessed by the pulse 1-hr exposure. The in vitro sensitivity index, expressed as a log odds ratio, was positive (greater than 0) in 8 of 11 patients achieving complete remission and negative (less than 0) in 7 of 10 patients failing to achieve complete remission, with an overall correlation of 71%. This is at least as good as the pulse exposure method, which has a correlation of 68%. If sensitivity indexes of marginal magnitudes (--1.0 approximately +1.0) are excluded, the correlation increases to 92% (12 of 13 patients). The correlation appears to improve especially for 1-beta-D-arabinofuranosylcytosine by the continuous exposure method (71%) as compared with the pulse method (57%). This study establishes the feasibility of an in vitro chemotherapy sensitivity testing of leukemic clonogenic cells by continuous in vitro drug exposure and suggests that the continuous exposure method may be better than the pulse method for antimetabolites such as 1-beta-D-arabinofuranosylcytosine. The data also suggest that simulation of the in vivo drug schedule may be important in this in vitro test.