A high relapse rate in AML suggests that present therapies are ineffective in eliminating leukemic stem cells. These may be present in the side population (SP) cells, which are able initiate leukemia in NOD/SCID mice. LSC detection may offer opportunities for detection of stem cells under conditions of minimal residual disease (MRD). Separately (Moshaver, this meeting) we show that leukemia associated immunophenotypes (LAP; used for MRD detection, Feller, Leukemia 18:1380, 2004) and CD34+CD38- stem cell marker C-type Lectin-like molecule-1 (CLL-1, van Rhenen, Blood 106: 6a, 2005), and Il-3 receptor α-chain CD123, are able to discriminate between normal and leukemic SP stem cells at diagnosis. In addition, LAP marker and CLL-1 expression, but not CD123 expression, on SP cells in control bone marrows (n=12) was lacking. SP cells could be detected in 13/17 diagnosis AML patients. The contribution of the AML SP to total SP was calculated based on expression of above-mentioned AML markers and checked with FISH analysis. We further use “AML SP frequency” (AML SP as % of WBC) and “AML SP fraction” (AML SP as fraction of total SP). We obtained 12 follow-up samples of 6 of these patients. At diagnosis median AML SP frequency was 0.015 and median AML SP fraction was 0.7. At follow-up, median total SP frequency was 0.01% (0.0007%–2.0 %). AML SP frequency herein was median 0.004% (<0.00001 – 1.05). AML SP frequency detection was compared with MRD detection method, known to predict prognosis (Feller, 2004). In 5/6 patients with low AML SP frequencies (< 0.004%) MRD was undetectable (≤0.01%). In 4/6 patients with high frequencies, MRD>0.01% was found. 3/6 patients showed a remarkable phenomenon: AML SP frequency was relatively high (0.006%, 0.007% and 0.014%, respectively) with apparently non-matching, relatively low MRD frequencies (0.02%, 0.02% and 0.01%, resp.), but with a drastic increase of MRD frequency at the next sampling time point (to 0.17%, 0.11% and 0.21%, resp). Such increases as well as the absolute percentages (>0.1%) inevitably lead to relapse (Feller 2004), which at present short follow up time already occurred for one patient. In addition, AML SP fraction was compared with MRD: 5/5 with low AML SP fraction (<0.21) had no detectable MRD (MRD≤0.01%). For high AML SP fraction (≥0.21) 4/6 had MRD>0.01%. AML SP fraction identified the same 3 special patients as did AML SP frequency: the AML SP fraction was high (0.33, 0.64 and 1.0, resp.) with the low MRD frequencies that subsequently strongly increased. Both stem cell parameters thus add additional prognostic value to MRD assessment. In conclusion, AML and normal stem cells can be detected with great specificity in AML remission bone marrow using LAP markers and CLL-1, that specifically stain AML SP stem cells and not normal SP stem cells. Additional predictive power for forthcoming relapses comes from the selective sparing of AML SP stem cells after therapy compared to both the more mature AML cells and the normal stem cells. The introduction of such stem cell parameters may thus contribute to MRD based risk stratification during treatment, offer guidelines for future clinical intervention time points, while studying cellular characteristics of both AML and normal stem cells may allow to design stem cell directed therapies that are effective for eradication of AML stem cells with minimal toxicity towards normal stem cells.