AUTHORS
Satyavardhan Rao Nittu, PharmD, Francisco Marchi, PhD2, Nam Nguyen, MS, PharmD3, Vivek Shastri, Research Assistant Professor4, Xueyuan Cao, PhD5, Raul Ribeiro, MD6, Jeffrey Rubnitz, MD, PhD7, Alan S. Gamis, MD8, E. Anders Kolb, MD9, Richard Aplenc I, MD, PhD10, Jeffery M. Klco, Member; Director, Division of Hematopathology and Molecular Pathology11, Stanley Pounds, PhD12, Todd A. Alonzo, PhD13, Rhonda E. Ries, MA14, Soheil Meshinchi, MD, PhD15, Jatinder K Lamba, PhD16
AFFILIATIONS
1 University of Florida, Gainesville, United States
2 University of Florida, Gainesville, United States
3 University of Florida, Gainesville, United States
4 University of Florida, Gainesville, United States
5 University of Tennessee Health Science Center College of Nursing, Memphis, Tennessee, United States
6 Ascentage Pharma, Rockville, Maryland, United States
7 Ascentage Pharma, Rockville, Maryland, United States
8 Children's Mercy, Kansas, United States
9 Leukemia & Lymphoma Society, Washington, United States
10 Division of Oncology, Children's Hospital of Philadelphia; Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, United States
11 St. Jude Children’s Research Hospital, Tennessee, United States
12 St. Jude Children’s Research Hospital, Memphis, United States
13 Children's Oncology Group, Arcadia, United States
14 Fred Hutchinson Cancer Center, Seattle, United States
15 Fred Hutchinson Cancer Center, Seattle, United States
16 University of Florida, Gainesville, United States
ABSTRACT
Purpose: To identify relapse-associated differentially expressed genes using paired diagnosis-relapse specimens and evaluate their prognostic utility for predicting relapse and survival outcomes in pediatric acute myeloid leukemia (AML).
Experimental Design: Publicly available RNAseq data from 2,116 pediatric AML patients enrolled in Children's Oncology Group clinical trials were analyzed. Differential gene expressions were evaluated for 275 matched diagnosis-relapse specimen pairs. Genes differentially expressed at relapse were assessed for association with cumulative incidence of relapse (CIR) in an independent dataset, and relapse-associated genes were subsequently incorporated into a Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression model to develop a relapse-risk signature.
Results: Among 227 genes differentially expressed between diagnosis and relapse (FDR <0.01), 213 demonstrated increased expression at relapse, and of these, 31 had an existing targeting drug. Of these, 67 genes were significantly associated with relapse risk among 900 patients who achieved complete remission after induction I. A 15-gene relapse-risk signature (15-GeneRR) was derived using LASSO regression. Patients with high 15-GeneRR scores had significantly higher cumulative incidence of relapse compared with those with low scores (HR=2.30, 95% CI: 1.79–2.94, P<0.001). The association was validated in an independent AML08 cohort (HR=3.60, 95% CI: 1.27–10.19, P=0.016). Multivariable analyses confirmed the independent prognostic value of the 15-GeneRR signature after adjustment for age, risk group, race, and other established clinical factors. Notably, among standard-risk patients, 39% of those with high 15-GeneRR scores relapsed within one year of achieving remission compared with only 14% of those with low scores.
Conclusions: Relapse-associated transcriptional alterations identified from paired diagnosis-relapse specimens provide important therapeutic and prognostic information in pediatric AML. The 15-GeneRR signature independently predicts relapse risk and adverse clinical outcomes and may improve risk stratification, particularly among standard-risk patients where therapeutic decision-making remains challenging.
DISCLOSURES
Satyavardhan Rao Nittu, PharmD: No relevant conflicts of interest to declare.