92 patients with acute myeloid leukemia were classified according to the FAB classification (M1 n = 20, M2 n = 43, M3 n = 1, M4 n = 19, M5a n = 2, M5b n = 2, and M6 n = 5 patients). Serum measurements of lactoferrin (LF), myeloperoxidase (MPO) and lysozyme (LYS) were performed before the start of treatment. LF was significantly lower in M1 when compared with M2 but not as compared to M4, MPO was significantly higher in M2 and M4 than in M1, but comparable MPO levels were found in M2 and M4. LYS was significantly elevated in M2 in comparison with M1, and in M4 when compared to both M1 and M2. Polymorphonuclear granulocytes (PMNs) in M1 were significantly reduced when compared with M2 and M4, whereas mononuclear cells were significantly increased in M4 in comparison with both M1 and M2. FAB classification did not generate any prognostic information. When the patients were, instead, subdivided according to LF levels were found prognostically significant differences. Of patients below 100 micrograms/l, 44% went into remission as compared to 77% with LF from 101 to 400 micrograms/l. In patients with LF levels above 400 micrograms/l the remission frequency was only 14%. Multivariate statistical analysis on the data further suggested that lactoferrin may be used as an independent prognostic indicator. We conclude that although determination of the serum-levels of lactoferrin, lysozyme and myeloperoxidase in certain cases may be valuable as a supplement to the morphological examination of acute myeloid leukemia, it is evident that none of the three determinations can be used alone to distinguish between the FAB groups.