Summary A system for the prediction of clinical response in acute myelocytic leukaemia (AML), based on inhibition of growth of colony forming cells (CFC) was studied. If the product of initial drug concentration and time of exposure (C x T) was constant, the response to adriamycin (Adr) was constant, at T values <48 h. No constancy of response to the phase-specific agents cytosine arabinoside (Ara-C) and 6-thioguanine (6TG) was demonstrated with constant C x T (T value range 0.25-48 h). Hence in the predictive test, a 1 h incubation with Adr was employed whilst a continuous exposure to Ara-C and 6TG, with these drugs incorporated in the agar medium, was used. The in vitro sensitivity to Adr, Ara-C and 6TG of 19 AML patients and the predictive value of several parameters of sensitivity were evaluated. 6TG sensitivity was not useful for prediction of remission. Adr sensitivity in vitro made a greater contribution to prediction of remission than did Ara-C sensitivity. Seventy-nine percent of patients were correctly classified if Adr data alone were considered. A multivariate function including Adr and Ara-C results was obtained which resulted in 84% of patients correctly classified as sensitive or resistant to the agents received in remission-induction therapy.