Peripheral blood mononuclear cells (PBMC) from 42 patients with acute myelogenous leukaemia (AML) in complete remission (CR) and from normal donors were activated into LAK cells in the presence of 1000 U/ml of recombinant interleukin-2 (rIL-2). Cytotoxicity of LAK cells was assayed against K562, Daudi, and Raji cell lines, and autologous and/or allogeneic thawed leukaemic blasts. Fresh unactivated PBMC from normal donors and AML patients served as controls. Mean +/- standard deviation (SD) percentage lysis of the different targets by patient LAK cells were: K562 61 +/- 20%, Daudi 62 +/- 23%, Raji 48 +/- 24%, autologous blast cells 12 +/- 16% and allogeneic blast cells 13 +/- 10%. Lysis of the different targets by LAK cells from normal donors was similar to that achieved with LAK cells from AML patients. Overall there was a good correlation between the lysis of the different targets. There was no significant difference between the percentage lysis of autologous and allogeneic thawed blast cells, although LAK cells from seven out of the 18 patients tested were unable to lyse autologous leukaemic cells. Activity of patient LAK cells did not correlate with the initial characteristics of the patient nor with the time spent in CR before harvesting PBMC for activation. At the time of analysis, 32 patients were in continuing CR and 10 had relapsed. Multivariant analysis for prognostic factors showed that patients whose LAK cells had more lytic activity on K562 (P = 0.005) and fresh blast cell (P = 0.02) targets had significantly less risk of relapse than patients with little inducible LAK cell activity.(ABSTRACT TRUNCATED AT 250 WORDS)