Large genomically annotated acute myeloid leukemia (AML) datasets exist, but there are relatively few population-based contemporary cohorts with integrated clinical, genomic, and outcome data. The REFORM-AML effort assembled a Danish, real-world cohort to characterize clinicopathological features, the molecular landscape, treatment patterns, and the prognostic performance of established ELN risk models.
A total of 2,512 AML patients were identified for the REFORM-AML cohort covering the years 2015–2022. Of these patients, 33.8% had available next-generation sequencing (NGS) data (denoted NGS+ in the study). The dataset therefore represents a mix of genomically profiled and non-profiled cases within a national population-based sample.
When patients were stratified by age, differences in treatment pattern emerged related to NGS availability. Among patients aged 70 years or younger, baseline characteristics and clinical outcomes were comparable between patients with NGS data (NGS+) and those without (NGS-). In patients older than 70 years, a larger proportion of NGS+ patients received intensive treatment compared with NGS- patients. Despite this difference in treatment allocation for the older subgroup, survival among intensively treated older patients did not differ by NGS status.
The study found that the distribution of somatic mutations varied significantly according to both age and sex. Specifically, older age and male sex were associated with higher frequencies of adverse-risk gene mutations. The report emphasizes these demographic associations with molecular risk features but does not provide full mutation-by-mutation frequencies in the abstract; detailed mutation-level data are available in the full manuscript and underlying dataset.
In the subset of intensively treated patients with NGS data, the authors applied the ELN2017 risk classification and reported 5-year overall survival (OS) stratified by ELN2017 groups: 58.4% for favorable risk, 43.4% for intermediate risk, and 28.2% for adverse risk. Hazard ratios (HRs) for OS using intermediate risk as the reference were 0.63 for the favorable group and 1.45 for the adverse group, indicating separation of survival outcomes across ELN2017 categories in this treated, genomically characterized subgroup.
Applying the updated ELN2022 classification to the same intensively treated, NGS+ subgroup produced 5-year OS estimates of 56.9% for favorable risk, 51.8% for intermediate risk, and 29.7% for adverse risk. Reported hazard ratios versus intermediate risk were 0.78 for favorable and 1.86 for adverse, demonstrating prognostic discrimination under the ELN2022 scheme as well.
The investigators compared the predictive performance of ELN2017 and ELN2022 in a time-dependent survival model. They report that the two models had comparable predictive performance for overall survival in intensively treated, NGS+ patients. This finding suggests both ELN frameworks reliably stratify risk for OS in this real-world, genomically profiled subgroup, although point estimates of survival and hazard ratios differed modestly between the two systems.
The study received approval from the North Denmark Region (approval/registration IDs: 2021-011009 and 2021-197). Under Danish health-care legislation, individual patient consent was waived according to the approval. Data cannot be shared directly per Danish law, but access is possible through application; additional information can be requested from the corresponding author. Funding for the work was declared from the Danish Cancer Society. Several authors reported competing interests, including consultancy or advisory roles, travel grants, speaker fees, and research funding; these are listed in the manuscript's competing interest statement.
Key conclusions reported by the authors are:
Outcomes of intensively treated AML patients in the REFORM-AML database were comparable regardless of whether NGS profiling was available, supporting the cohort's representativeness for the Danish AML population.
Age and male sex correlated with higher frequencies of adverse-risk mutations, underlining demographic variability in the molecular landscape of AML.
Both ELN2017 and ELN2022 robustly predicted overall survival in intensively treated, NGS+ patients, with similar time-dependent predictive performance.
These findings highlight the value of population-based genomic annotation for understanding mutation distribution and validating established risk classification systems in real-world settings. The manuscript is a preprint and has not been peer reviewed; the authors note that the data and detailed results are available in the full text and via application under Danish data-access rules.