Das PDF wurde in Ihrem Download Ordner gespeichert oder in einem neuen Tab Ihres Browsers geöffnet. Den Download-Ordner finden Sie auf Ihrem Laufwerk oder in der Browserleiste von Firefox und Safari rechts oben bei dem Pfeil-Icon.
Estimated reading time: 6 minutes
For nearly half a century acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS) have been separated according to the percentage of bone marrow blasts. However, the last ten years have shown that the incorporation of genetic information in the classification improves its accuracy in diagnosis and prognosis. Some cases with specific genetic aberrations are classified as AML even irrespective of the blast counts. The estimation of the blast counts itself might vary depending on the experience of the hematopathologist and other methods that also quantify pathologic cell populations, such as flow cytometry, might report divergent results. The research work of the MLL on this topic was summarized in three abstracts for ASH 2019:
Comprehensive diagnostics of hematologic malignancies is based on an integrated application of various methods such as cytomorphology, cytogenetics, immunophenotyping and molecular genetics. Cytomorphology and flow cytometry are both applied to identify and quantify pathologic cell populations with, in the majority of the cases, consistent results. However, especially in MDS, AML, and related diseases it is sometimes challenging to align cytomorphologically defined blasts (%blasts) and flow cytometrically defined myeloid progenitor cells (MPC, %MPC). Drawing from the collection of analyzed myeloid malignancies at MLL Kern et al. analyzed the genetic background of 49 cases with %blasts > %MPC (group1) compared to 83 AML cases with consistently high percentages by both methods (group2) and 53 MDS cases with matching low %blasts and %MPC (group3). With 76% normal karyotypes were most frequently observed in group 1, followed by group 3 (64%) and group 2 (51%). Hence, the karyotypes of group 1 were more similar to MDS. 34 genes associated with myeloid malignancies were analyzed regarding their mutational frequencies in the different groups. It was found that the mutation spectrum of cases from group1 were similar to MDS developing into secondary AML with higher frequencies of ASXL1, SRSF2, TET2, and RUNX1. SF3B1 mutations that are associated with favorable prognosis in MDS were only detected with low frequencies in groups 1 and 2. Genes typically associated with de novo AML were found most frequently in group 2. Hence, the study revealed a higher similarity of cases with %blasts > %MPC to MDS than to AML, encouraging a more extensive integration of mutation data into the classification of these myeloid neoplasm.
Although the percentage of bone marrow blasts is an arbitrary threshold, it led to different therapeutic strategies and the discovery of more or less directly related cytogenetic changes. Recent years expanded the spectrum by adding molecular genetic information for diagnosis, prognosis and increasingly targeted treatment options. However, many of the identified cytogenetic and molecular genetic findings are shared between both diseases, independent of blast cell counts. The analysis of molecular features was so far somewhat limited by the sequencing cost of whole genome sequencing (WGS) which led to the more cost efficient selection of targeted gene panel sequencing. Now, the decreasing costs of WGS provide researchers with unprecedented possibilities to investigate disease-specific genetic patterns more comprehensively and allow also for the identification of patient-specific molecular fingerprints that might guide targeted treatment approaches. Meggendorfer et al. applied machine learning techniques to WGS data to identify the most discriminative features in a cohort of 591 AML and 701 MDS cases (randomly divided into a training (90%) and validation (10%) set). The authors used LASSO regression to select the features that optimized the classification accuracy of AML versus MDS and subsequently trained a Naïve Bayes classifier to stratify patients of the validation cohort. The initial dataset consisted of 2,918 genes of which 66 genes were deemed important. In addition 8 cytogenetic aberrations were included in the final model. Relational self-organizing maps were used to group the molecular features of the final model and to identify entity-specific co-occurrence networks. For AML the dominant features were: normal karyotype with NPM1 mutation, t(8;21), t(15;17) inv(16), WT1 mutation, KRAS mutation, and TP53 mutation with co-occurring del(7) and del(5q). In MDS the dominant features were: del(5q), SF3B1 mutation, normal karyotype, and del(20q). From the validation cohort 35/132 AML cases and 10/140 MDS cases were assigned to their respective counterpart by the classifier, resulting in an accuracy of 83.7%. The divergent results between cytomorphology and genetics could easily be explained by the molecular profile of the misclassified cases that matched the respective entity-defining molecular features. In summary, it could be shown that the molecular profiles indicate a considerable genetic similarity of the two diseases independent of the blast count.
WHO-based AML classification is hierarchically structured and integrates different features, such as genetics and multilineage dysplasia (MLD). An AML WHO category with poor prognosis is “AML with myelodysplasia-related changes” (AML-MRC). The category is defined by blasts > 20% and the presence of MLD or MDS related cytogenetic abnormalities. Although it has been shown by various groups that survival rates for patients diagnosed with AML-MRC are lower, MLD as the sole criteria of AML-MRC does not significantly affect overall survival (OS). Shorter OS of the AML-MRC group is usually attributed to the presence of unfavorable cytogenetics such as complex and/or monosomal karyotypes. Different FDA and EMA approved drugs exist for the treatment of AML-MRC. However, the diagnosis of AML-MRC in clinical routine is hampered by limited reproducibility and the turnaround times can be rather long (5-10 days for cytogenetics). Hence, Baer et al. aimed to identify a molecular AML-MRC signature in a cohort of 739 AML patients (165 AML-MRC, 574 AML-NOS, not otherwise specified) that allows fast and accurate diagnosis and prognosis. Molecular genetics analysis was based on WGS data but limited to 73 frequently mutated genes, to identify a signature that’s applicable with common routine panels. The AML-MRC group was characterized by an average of 2.7 gene mutations per patient and the most frequently mutated gene was TP53 (38%). TP53 mutations were associated with shorter OS and correlated with a complex karyotype in the AML-MRC group. Subsequently, LASSO regression was used to identify the most discriminative features between AML-MRC and AML-NOS. Mutations in TP53, RUNX1, SETBP1, splicing factors and epigenetic regulators as well as the absence of NPM1 and CEBPA mutations had the highest impact on AML-MRC prediction. To simplify AML-MRC prediction in a routine diagnostic workflow the identified features were transformed into an easy to use point system. The final molecular model and the simplified point system allowed a re-identification of 83% and 70% cases currently defined as MRC, respectively. However, the molecular models also identified additional 112 (model) and 80 (point system) cases as being AML-MRC for which OS analysis also revealed a significantly poorer outcome compared to AML-NOS (6 vs. 35 months). If cytogenetic data and patient’s history were added to the molecular models 99% (model) and 96% (point system) of true positive AML-MRC were identified. In conclusion, the study showed that informed genetic models allow the identification of 96-99% of AML-MRC as defined by WHO today. The molecular MRC-like pattern reliably identifies cases with comparably poor OS that might also benefit from AML-MRC treatment.