Relapse prediction in acute myeloid leukaemia patients in complete remission using WT1 as a molecular marker: development of a mathematical model to predict time from molecular to clinical relapse and define optimal sampling intervals.

We hypothesized that Wilms tumour 1 gene (WT1) expression levels in acute myeloid leukaemia (AML) patients might have predictive value and reveal molecular relapse kinetics. WT1 level was determined at diagnosis, during therapy and post-therapy follow-up in 89 patients who reached first complete remission (CR1) (952 samples, median 8 samples/patient, range 2-38). CR1 bone marrow (BM) WT1 level above normal (based on 39 healthy donors) was an independent adverse prognostic factor regarding both disease-free survival [hazard ratio (HR) 4.46, P = 0.001] and overall survival (HR 2.62, P = 0.019). By grouping 34 BM and 99 peripheral blood (PB) complete remission samples in monthly intervals prior to clinical relapse, disease development was delineated and a simple mathematical model constructed, that allowed for the prediction of relapse detection rates (RDRs) as well as median times [t(m)s] from WT1 positivity to clinical relapse. BM sampling was required to obtain RDRs above 93% and t(m)s above 67 d. Acceptable RDRs and t(m)s (81% and 44 d, respectively) could be acquired by bimonthly PB sampling. In conclusion, CR1 WT1 expression is an independent prognostic factor in AML. According to our model, BM is superior for relapse prediction, but PB samples are useful when shorter sampling intervals are possible.