Integrated Molecular and Pharmacogenetic Profiles in Prediction of Early Response to Therapy in Children with Acute Lymphoblastic Leukemia (ALL): A Report From Malaysia-Singapore ALL Study Group.

Abstract 2624

Poster Board II-600

Although childhood ALL is the prototype for a drug-responsive malignancy, some patients relapse or develop severe toxicity due to significant inter-individual variability. Leukemic factors like the molecular subtype account for the majority of the variation in prognosis; some studies have suggested that polymorphisms in key genes involved in drug transport, uptake, metabolism or targeting are responsible for smaller magnitude of variance in terms of therapeutic efficacy. Thus far, genetic characteristics are not widely applied in modern risk-stratifying system for ALL patients. And due to the significant differences of ethnic genotype/allele frequencies, the variance is probably much larger than leukemic factors alone. We integrated molecular and pharmacogenetic profiles to assess the contribution of germline variations in early treatment response using minimal residual disease MRD status in the Malaysia-Singapore ALL 2003 study. Three hundred and seventy children with ALL (200 Chinese and 170 Malays) were included and tested for t(12;21), t(9;22), t(1;19) and t(11q23)/MLL rearrangements. Genotypes of 20 polymorphisms in 11 different candidate genes were collected (Phase II enzymes: GSTM1, GSTT1, GSTP1, NQO1; folate metabolism: MTHFR, MTHFD1, TYMS, SLC19A1; transporter: ABCB1; candidates reported in other studies: CCR5, IL15). MRD levels were measured by real-time quantitative PCR at Week-5 (TP1, end-of-induction) and Week-12 (TP2) after the start of therapy. Chinese and Malays were analyzed separately to minimize confounding effect as most studied polymorphisms had significantly different genotype/allele frequencies between the two ethnic groups. All genotypes, together with patient characteristics of sex, leukemia lineage, molecular subtype and NCI risk group, were included as variables to build a prediction model using Classification and Regression Tree (CART) analysis. Patients with ≥0.1% TP1 MRD and ≥0.01% TP2 MRD had significantly shorter event-free survival (EFS) compared to those with undetectable MRD level (<0.01%) at two time-points respectively. Chi-squared test did not show significant association with MRD status for any studied polymorphism, indicating that no single pharmacogenetic factor was an independent predictor. CART analyses revealed particular conditions associated with good or poor MRD status at two time points respectively. In the Chinese, 1) patients carrying t(12;21) or t(1;19) have 1.3- and 1.6-fold of higher chances to have undetectable MRD at TP1 and TP2 respectively (P<0.001); 2) at TP1, carriers of t(9;22) or 11q23/MLL rearrangements or those without common translocations but carrying both ABCB1 −2352G/A and SLC19A1 G/G have a 2.5-fold of higher risk for ≥0.1% MRD (P=0.005 and 0.001 respectively); and 3) at TP2, patients carrying t(9;22) or t(11q23)/MLL rearrangements or those without common translocations but simultaneously carrying SLC19A1 G-allele, GSTM1 present and IL15 rs17015014 C/C also have a 2.5-fold of higher risk for ≥0.01% MRD (P=0.017 and 0.006 respectively). In the Malays, 4) NCI high risk patient carrying IL15 rs17015014 C-allele have a 1.9-fold of higher risk for ≥0.1% TP1 MRD (P=0.004) whilst those carrying both GSTM1 null genotype and MTHFR 677C/T have a 3.6-fold of higher risk for ≥0.01% TP2 MRD (P<0.001). Correct classification rates for all built models vary from 73.6% to 79.4%. Models built by using only genotypes or patient characteristics are inferior to the classifier consisting of both features. We conclude that inherited germline variations affect therapeutic efficacy but their importance vary in different races probably due to the differences in population frequencies. These pharmacogenetic factors rank below the more powerful leukemic factors such as molecular subtypes but provide a distinct albeit smaller impact.