Paper Details

ABSTRACT

BACKGROUND

WHO 2022 and ICC 2022 classifications use distinct blast percentages to differentiate acute myeloid leukemia (AML) from myelodysplastic syndrome (MDS). Blast thresholds vary according to the defining genetic abnormality and can range from no required minimum to 20%. The ICC introduced MDS/AML for adult patients with 10-19% blasts to reflect the continuum between MDS and AML, but the prognostic validity of 10% and 20% thresholds remains uncertain. We evaluated the prognostic effect of blast counts on survival in different genomic contexts in patients with both MDS and AML.

METHODS

We pooled four independent public cohorts with blast count and overall survival (OS) data – including both patients with MDS (IPSS-M, International Working Group) and AML (AMLSG and UK-NCRI from the publication by Tazi Y et al., TCGA and Beat AML). For the primary analysis, we included adults without TP53 mutations or AML-defining recurrent genetic abnormalities and classified cases according to ICC groups as MDS-related gene mutation, MDS-related cytogenetic abnormality or not otherwise specified. Cases with 10–19% blasts were categorized as MDS/AML and those with ≥20% as AML; cases with <10% were included only in a separate blast-continuum analysis OS was modeled with study-stratified Cox regression and adjusted for age and ICC group. Cohort-specific estimates from datasets containing both blast groups were combined using random-effects meta-analysis, with heterogeneity quantified by I². Blast percentage was also modeled continuously using natural cubic splines, with tests at the clinically defined 10% and 20% thresholds. The primary analysis was pre-specified; analyses of NPM1-mutated and recurrent rearrangement AML were exploratory.

RESULTS

Among 2,199 evaluable patients, including 389 with MDS/AML and 1,810 with AML, classification as MDS/AML was not independently associated with OS after adjustment for age and ICC group (study-stratified Cox HR 0.91, 95% CI 0.80–1.03; p=0.15). There was no evidence that this association differed across ICC groups (interaction p=0.61). Four cohorts included patients in both blast groups (n=2,116); their estimates yielded a pooled HR of 0.92 (95% CI 0.77–1.10), with no heterogeneity (I²=0%). In leave-one-study-out analyses of the individual-level model, HRs ranged from 0.88 to 1.00.

In a separate continuum analysis of 2,917 patients with MDS-related mutations or cytogenetic abnormalities, including 1,769 deaths, the association between blast percentage and OS was strongly nonlinear (p=7.3×10?³²). A significant step term was observed at 10% blasts (HR 2.45; p<0.001), whereas no step was detected at 20% (HR 0.91; p=0.37). Exploratory analyses included 1,340 patients with NPM1-mutated AML (744 deaths; 84 with <20% blasts) and 680 with recurrent rearrangement-defined AML excluding APL (323 deaths; 45 with <20% blasts). Neither group showed evidence of a threshold effect at 10% or 20%. Blast percentage was not clearly associated with OS in NPM1-mutated AML (HR 1.03 per 10% increase; p=0.056) but showed a modest continuous association in rearrangement-defined AML (HR 1.07 per 10% increase; p=0.004). These exploratory analyses were limited by the small number of genetically defined AML cases with <20% blasts.

CONCLUSION

In this analysis, ICC-defined MDS/AML with MDS-related genetic abnormalities or an NOS phenotype did not have a distinct outcome compared with AML with the same genetic profile. Within this population, blast percentage showed a nonlinear association with survival, with a significant increase in risk at the 10% threshold but no additional prognostic separation at 20%. In contrast, exploratory analyses suggested a more modest, predominantly linear association between blast percentage and outcome in NPM1-mutated AML and AML defined by recurrent gene rearrangements, without evidence of a discrete threshold effect. These findings support the incorporation of genetic context, rather than a universal blast cutoff, into future classification systems for MDS and AML.