Interobserver Variability with the Diagnosis of Acute Myeloid Leukemia (AML) and Myelodysplastic Syndrome (MDS) ¿Is the Threshold of 20% Bone Marrow Blasts Reproducible?

Introduction. The boundaries between MDS and AML are still a matter of debate. In the 2001 WHO Classification, the myeloblast count distinguishing AML and MDS was lowered from 30% to 20% of the bone marrow (BM) cells or peripheral blood (PB) leukocytes. It was justified on the basis that treating patients with 20-29% BM blasts with intensive chemotherapy showed a similar outcome to those with > 30% BM blasts.

However, the better knowledge of the biology of both diseases is showing that in several cases AML and high risk MDS share identical genetic profiles, as it is well known in AML with myelodysplasia- related changes (AML-MRC). Currently there are new therapeutic options, less toxic, and suitable for elderly people.The threshold of 20% BM blast is artificial, but it is still the main criterion used in clinical trials and also in real life to discriminate patients that probably belong to the spectrum of the same biological entity. Treatment of patients with MDS or AML is widely based in this relatively arbitrary condition.

Objective: To study if the threshold of 20% bone marrow blasts, distinguishing MDS with excess of blasts type 2 (MDS EB 2) and AML, is reproducible among different observers.

Methods. 120 bone marrow samples from patients previously diagnosed with MDS-EB-2, AML or therapy-related myeloid neoplasms (t-MN) according to 2016 WHO classification were included. The diagnosis of MDS required cytogenetics and/or FISH, and the cases with AML should have been classified following the 2017 ELN recommendations, regarding immunophenotyping, cytogenetics and molecular biology. The design of the study was established to include cases with <40% BM blasts, WBC <25x109/L and less than 20% myeloblasts in peripheral blood. The proportion of samples from each category was not predefined. Specimens were collected from 12 hospitals and were evaluated by 12 morphologists. Each observer evaluated 20 samples, and each sample was analyzed independently by two morphologists. The second observer was blinded to the clinical and laboratory data, except for the peripheral blood (PB) counts. The interobserver concordance was evaluated using the Cohen kappa test.

Results. Finally 116/120 samples were considered suitable for the study. Regarding 2016 WHO categories, 55 cases showed MDS EB-2, 44 AML-MRC, 8 t-MN, 4 AML- NOS, 2 NPM1-mutated AML, 2 RUNX1-RUNX1T1 AML, 1 BCR-ABL1+ AML. Next generation sequencing was performed in 79 cases. Discordance was observed in 34/116 cases (29.3%). 14 cases with MDS-EB2 (1 NPM1+) were classified as AML-MRC by the second observer, 16 AML cases as MDS EB-2, 3 MDS EB-2 as MDS- EB1 and 1 AML as MDS- EB1. The genetic and /or molecular profile of the discordant cases was heterogeneous. Regarding the threshold of 20% BM blasts, discrepancies were 31/116 (26.7%, I Kappa test = 0.46, moderate agreement). The agreement between MDS-EB-2 and AML-MRC, with discordance in 28/98 cases (28.6%), was moderate-fair (Kappa test= 0.42).

Conclusion. The threshold of 20% BM blasts did not accurately separate AML from MDS EB-2. Particularly less concordance was seen for AML-MRC. Incorporation of genetic and molecular characteristics to the morphologic diagnosis is needed to optimize the definition of both entities.

Acknowledgment: Angel Cedillo, Secretaría Técnica AMHH.