An International Multi-Center Microarray Study for the Molecular Classification of Leukemia Identifies Novel Sub-Groupings in MDS Overlapping with AML.

Microarrays can identify robust gene expression signatures associated with distinct sub-classes of paediatric and adult leukemias. Recently, the MILE (Microarray Innovations in LEukemia) study has analysed 1901 expression profiles from retrospective samples in 11 centres (ELN: 7, USA: 3, Singapore: 1). MILE has compared the microarray classification accuracy of 16 acute and chronic leukaemia subclasses, MDS, and non-leukaemia as control group, to routine diagnostic workup. The achieved cross-validation accuracy was very high for the leukaemia subclasses: ~96%.

Included in the study were 175 samples diagnosed as MDS, however, only 49.1% of these samples were correctly called as MDS from their underlying gene expression profiles. The remainder were approximately equally split between a call of “non-leukaemia” (24%) and “AML” (24.6%). A further sub-division of MDS samples called “AML”: 81% called as “AML with normal or other cytogenetics”; and the remainder as “AML with complex cytogenetics”.

MDS is a heterogeneous group of disorders with a wide range in blast cell count, cytogenetics and number of cytopenias. Our analysis showed that neither study centre nor age were a factor in differentiating between “MDS”, “AML” or “non-leukaemia”. However, WHO classification was highly correlated with the microarray classification result; specifically RAEB(I or II) was associated with “AML” call (p < 0.0001). RA/RARS was highly correlated with “MDS” or “non-leukaemia” calls. Furthermore, IPSS was significantly correlated with call (p>0.0001): 65% of patients with an IPSS score of Int-2 or above were classified as “AML”. Examination of the individual components of the IPSS showed that two patients classified as “AML” had a blast count of >20%, under the WHO definition these would be defined as AML and were excluded. Individually, the blast, karyotype and cytopenia contributions were highly significant (p<0.0001, <0.013 and <0.0001 respectively) when comparing “AML”, “MDS” and “non-leukemia” calls. All the “non-leukemia” patients had <10% blasts, with 85% (34/40) having a cytogenetic score of 0 (Normal or Good (1 of: -Y, del(5q), del(20q))) and 82.5% having only 0 or 1 cytopenias. In contrast, 45% of the “AML” samples had between 11 & 20% blasts, 32% with intermediate (0.5) or poor/complex (1) cytogenetic score and 79% had 2 or 3 cytopenias. Furthermore, survival data was available for 122 of the diagnosed MDS patients and showed that MDS patients called “AML” had a trend towards shorter survival (2P=0.2) than those called “MDS” or “non-leukaemia”. These analyses, in combination with gene expression signatures, may contribute to a redefinition of MDS classifications.