Minimal residual disease (MRD) has been found to offer an independent post-remission prognostic marker in acute myeloid leukaemia (AML). We have confirmed this in a large blinded study in which MRD was assessed in parallel to a HOVON therapy protocol. Subsequently, MRD is now included as part of a new risk classification in a HOVON/SAKK treatment protocol where intermediate risk patients, previously all scheduled for allogeneic stem cell transplantation, are scheduled for autologous transplantation in case of MRD-negativity. In parallel, leukaemia stem cell load at clinical follow up was identified as a factor that adds significantly to the prognostic value of MRD load. Especially relapse in MRD-negative patients can now be better predicted based on leukaemia stem cell (LSC) frequency. A single tube assay was therefore developed which facilitates stem cell load assessment and allows this in the majority of patients. Lastly, false-negative MRD and false-negative LSC cases may be identified by monitoring upcoming aberrant populations that, due to low frequencies, were not properly identified at diagnosis. Introduction Treatment of AML patients usually is performed according to classification in risk groups defined by molecular, cytogenetic and clinical parameters, which thereby identify patients with favourable, intermediate and poor/very poor risk.1 Although very useful, this classification is still far from the ideal situation in which patients should have individualized risk assessment, potentially allowing individualized therapy: all risk groups still harbour patients with good and poor outcome, even the most favourable patient group, while the poor risk patient group still harbours good performing patients. Parameters that would integrate all known and unknown factors contributing to individual patient's clinical outcome were therefore thought to be urgently needed. A post-remission parameter like MRD has been proposed to represent such a factor and is meanwhile extensively studied. 2 Our recent study in the HOVON/SAKK consortium (http://www. HOVON.nl) has shown for young patients (HO42A, for details see website) that MRD assessed by flowcytometry in an unbiased way (prospective sampling and MRD quantitation in a multiinstitutional way, with complete lack of knowledge on clinical outcome until final analysis) is an independent prognostic factor in all individual risk groups. However, despite these improvements, still a considerable portion of MRD-negative group (defined as having MRD below a threshold level of 0.1% of WBC) relapse. 3 Also, the terms MRDnegative and MRD-positive (above the threshold level of 0.1%) have a different meaning in the different risk groups: MRDpositive patients in the favourable risk group still have relatively good prognosis, while MRD-negative patients in the poor risk group still have relatively poor prognosis. Intrinsic properties of MRD cells may underlie this phenomenon and one of the factors may be the LSC load. LSC are hypothesized to initiate leukaemia, to be therapy resistant, and at the basis of re-growth of the leukaemia after treatment. 4 As a single independent parameter it has been shown that stem cell load at diagnosis is a prognostic marker5,6,7,8 and, more recently in the HO42A study, we also showed prognostic impact after therapy. 7 The stem cell load, when taken into account in a risk assessment that also includes MRD, may further improve MRD based risk assessment. Methods Flowcytometric MRD was performed using the approach in which leukaemia associated immunophenotypes (LAIP) were defined at diagnosis.3 In the HOVON42A study risk was based on the combination of cytogenetics, molecular aberrations and clinical parameters. LSC can be defined as CD34+CD38-. Strategies to define such population at diagnosis and at follow up have been described elsewhere. 7 LSC specific markers (not expressed on CD34+ CD38- normal hematopoietic stem cells, HSC), included lineage markers like CD7, CD19, CD569and, in addition, CLL- 1 (CLEC12A)10 , were originally included in the LSC panel. 7 In a new protocol, tested in more than 500 AML samples, additional markers have been included (e.g CD123, CD33, CD96, TIM-3, CD44). The combination of MRD and LSC frequencies after induction therapy was tested for relapse free survival in a small (n=91) patient group. Threshold levels used for LSC were much lower than for MRD: 1 in 100,000 cells to one in a million cells.7 Results MRD: Based on the prognostic impact of MRD in the HO42A clinical study, the present HO132 protocol contains a risk group previously designated as intermediate risk, but now, based on MRD positivity (MRD % >0.1) after induction courses, is part of the poor risk group. Since MRD had the strongest prognostic impact after the second induction course, for logistical and financial reasons it was decided to restrict the protocol toMRD assessment after the second induction course. MRD status after consolidation, although informative 3,11 is not considered in decision making, since it does not comply with the timely finding of an allogeneic donor for part of the patients. Based on previous validation study 12, it was decided to perform MRD assessment in a largely centralized way. Accrual of patients has started (target number 800; 83 patients already included). LSC +/-MRD: When combiningMRD load and stem cell load (Figure 1) it was found that: 1. The MRD negative group consists of two different groups with different prognosis based on LSC load; 2. The LSC negative group can be divided in a MRD-positive and aMRD-negative group with different prognosis; 3.Double negativity goes along with good prognosis, while double positivity goes along with very poor prognosis. Fig1: Combination of MRD and LSC in prognosis of patients. ForMRD a threshold value of <0.1% (ofWBC) defines MRD-negative group and > 0.1% defines MRD positive group. For LSC similarly <0.0002% (of WBC) defines LSC negative group and > 0.0002% defines LSC positive group. Data for another threshold value have been published.7 LSC could be accurately determined in part of the patients only.7 Therefore, additional markers were studied with the aim to increase that part. Using the markers indicated inMethods, it is now possible to assess LSC in >90% of the patients. Based on extensive study of the redundancy of markers, we are now able to assess total LSC load in a givenAML both at diagnosis and after therapy in a one tube assay that includes 13 different markers. This approach is included as a side study in the HO132 protocol, where the possible contribution of LSC load assessment to prognosis will be studied. Immunophenotype shifts in MRD and LSC assessment: Although LSC markedly improves risk stratification based on well-known risk factors and now also includingMRD load, still the double negative patient group (MRD- LSC-) contains relapsing patients while the two single positive groups still contain both relapsing and non-relapsing patients (Figure 1). We and others found immunophenotype shifts in around 60% (median in different studies,13) of patients, and these contribute to MRD- and LSC false negatives. With the knowledge of immunophenotype and molecular shifts in mind, the present HO132 protocol, apart from the inclusion of LSC load assessment, also assesses the contribution of immunophenotype shifts to MRD (and LSC) negativity. Conclusions We conclude that, based on many studies, and especially on our recent unbiased MRD study, MRD contributes significantly to the existing risk group stratification that is based on cytogenetic and molecular factors as well as on clinical factors. The decision-making process whether a patient is allocated to an allogeneic stem cell transplantation can now be guided by MRD assessment. Moreover, the inclusion of other factors like stem cell load and immunophenotype changes, as seen forMRD and LSC, will contribute to the final goal of individualized therapy.