Quantification of minimal residual disease (MRD) is becoming increasingly important to guide therapy in patients with acute myeloid leukemia (AML). As only 25% of all AML have fusion genes that have been proven useful as MRD-markers, new targets are urgently needed to make high sensitive PCR studies possible in more AML cases. As length mutations in exon 14 of the FLT3 gene (FLT3-LM) are the most frequent genetic aberrations detected in AML and are predominantly found in AML with normal karyotype this is a highly promissing marker to assess prognosis predicted by MRD in this prognostically intermediate group. Recently, FLT3-LM as a follow up marker was discussed controversly because of possible instabillity. To further evaluate this issue we analyzed 62 paired samples at diagnosis and at relapse that were positive at both or only one of these two time points. 53 were positive at both time points. In these cases the mutation (mt) to wildtype (wt) ratio was increasing in 25 and decreasing in 6 cases. Five cases were mt- at diagnosis and gained the mutation at relapse. Four cases (6.5%) lost the mt at relapse. However, these cases also had a change in cytogenetics (n=2), or immunophenotype (n=2), and FAB subtype (n=4). Thus, these cases may be regarded as different leukemias. In addition, we analyzed 517 follow up samples (day 16 – day 1548) of 148 different patients with FLT3-LM at diagnosis by conventional RT-PCR (sensitivity: up to 10−3). First, all follow up samples were rated positive (n=281) or negative (n=236) for the FLT3-LM as determined at any time point. The negative group had a longer OS (median not reached vs 902 days; p=0.0008) and EFS (413 vs. 279 days, p=0.0004). In a next step, five different follow up checkpoints were defined: 1 (up to day 21 after start of therapy), 2 (day 22–60), 3 (day 61–120), 4 (day 121–365), and 5 (after day 365). Survival was best predicted by interval 4 for OS (median not reached vs. 580 days, p=0.0031) and EFS: (406 vs. 221 days; p=0.0001) in 79 negative and 57 positive samples analyzed at this checkpoint. Also prognostically relevant for OS were checkpoints 2 (31 negatives, 16 positivs, 614 vs. 325 days, p=0.0213), and checkpoint 3 (56 negatives and 23 positives, 614 vs. 325 days, p=0.0215). Thus, sole FLT3-LM status during follow up is a highly predictive parameter for prognosis. In 13 cases that were positive for a second marker 72 samples were analyzed by quantitative real time PCR in parallel (PML-RARA: n=5, MLL-PTD: n=5, DEK-CAN: n=2, MLL-AF6: n=1). A high correlation of the results was found (p=0.01). In addition, 10 cases were 8–15 available follow up samples including at least one relapse were selected for real time PCR. In these cases patient specific primers spanning parts of the individual mutation were designed. With these PCRs a high sensitivity of 10e5–10e7 was riched. The data obtained clearly indicate the relapses 3–4 months before clinical relapse. In conclusion, these data show that 1) FLT3-LM is a very powerful marker for the analysis of therapy response and early detection of relapse. 2) Instability is a minor problem especially because the instabilily in most cases is in the direction of accumulation of the mutation.