Introduction: Ara-C (cytarabine), daunorubicin, and etoposide (ADE) remain key chemotherapies used in remission induction and consolidation therapy of pediatric acute myeloid leukemia (pAML). However, multiple drug resistance is a major cause of therapeutic failure in pAML. Several ADE-significant pharmacokinetic/pharmacodynamic (PK/PD) genes have been identified. However, it is unknown which microRNAs (miRNA) have a significant impact on the regulation of gene (mRNA) expression levels involved in ADE's pharmacology and the treatment outcomes of pAML patients.

Methods: pAML patients with available bone marrow mRNA and miRNA expression levels from the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) database were included in the analysis. For step 1, we assembled a list of significant PK/PD genes (n=67) to ADE. In step 2, correlation analyses between mRNA expression levels of target PK/PD genes and miRNA expression levels were performed using mRNA-seq and miRNA-seq data from the TARGET database. miRNA-mRNA pairs with significant anti-correlation for expression levels using Spearman's rho (FDR < 0.05) were included for further analysis. In step 3, the least absolute shrinkage and selection operator (LASSO) was used to fit the Cox regression model with overall survival (OS) as the outcome variable on miRNAs with significant negative correlation with PK/PD genes from step 2. A thousand bootstraps of LASSO Cox regression were performed to identify miRNAs represented in at least 75% of the models. Coefficients from select miRNAs were used to generate a miRNA signature equation for the ADE response (ADEmiR) score. In step 4, patients were classified into low or high score groups according to the median value from the ADEmiR score, and analysis was performed for association with event-free survival (EFS), OS, minimal residual disease after the end of induction 1 (MRD1), and complete remission after the end of induction1 (CR1). Lastly, multivariate analysis was conducted using confounding factors such as age, clinically identified risk groups, FLT3 mutation status, and white blood cell count (WBC). The ADEmiR score equation was then validated using miRNA expression levels data from 164 adult AML patients from The Cancer Genome Atlas (TCGA) database.

Results: 219 patients were identified with a mean age of 10±6 years, and 54% were male. Within TARGET cohort, the high ADEmiR score group was associated with significantly inferior EFS (HR=3.26, p<0.001) and OS (HR=6.51, p<0.001) (Figure 1), higher MRD1 (44% vs. 22%, p=0.0054) and a greater proportion of patients not achieving CR1 (29% vs. 12%, p=0.0037) (Figure 2). In a multivariate Cox regression model that included ADEmiR score, age, clinically identified risk groups, FLT3 mutation status, and WBC, the high ADEmiR score group remained significantly worse for EFS (HR=2.54, p<0.001) and OS (HR=5.85, p<0.001) (Table 1). Among clinically identified risk groups, the high ADEmiR score remained a significant predictor of poor OS in the low-risk group (HR=4.80, p<0.001), the standard-risk group (HR=7.00, p<0.001), and the high-risk (HR=9.03, p=0.035) (Figure 3). The ADEmiR score was further validated in the TCGA cohort with consistent results showing the high ADEmiR score group was associated with inferior OS (HR=1.97, p<0.001) (Figure 1). In a multivariate Cox regression of the TCGA cohort, the high ADEmiR score group stayed significant with worse OS (HR=1.61, p=0.024). In the clinically identified risk groups of the TCGA cohort, the high ADEmiR score was associated with poor OS in both intermediate-risk (HR=1.71, p=0.033) and poor-risk groups (HR=2.66, p=0.037) with a trend in the low-risk group (HR=2.10, p=0.2) (Figure 4).

Conclusions: The ADEmiR signature scoring system utilizing miRNA could serve as a prognostic tool to predict treatment outcomes in AML patients receiving ADE. Ongoing studies conducted by our group are focused on validating the ADEmiR score in other cohorts as well as its integration with our other gene-expression based prognostic tools such as ADE-Response Score (ADE-RS), and pediatric leukemic stem cell (pLSC6) score to predict optimal ADE treatment outcomes in pediatric AML.

Disclosures

No relevant conflicts of interest to declare.

Author notes

*

Asterisk with author names denotes non-ASH members.

Sign in via your Institution