Quantitative Analysis of Minimal Residual Disease at the Completion of Induction Therapy Predicts Relapse in Children with B-Lineage Acute Lymphoblastic Leukemia in DFCI ALL Consortium Protocol 95-01.

Quantitative analysis of minimal residual disease (MRD) is emerging as among the most important predictors of subsequent relapse in childhood ALL. In 381 of 438 (87%) sequential cases of childhood B-lineage ALL enrolled on DFCI Consortium protocol 95-01, we could sequence IgH and/or TCR γ/d gene rearrangements at diagnosis to design allele specific oligonucleotide (ASO) primers at complementarity determining region 3 (CDR3). Real-time quantitative PCR (RQ-PCR) was performed using ASO and consensus primers and consensus TaqMan probes. The MRD value was calculated as the mean of triplicates of target gene normalized by the mean of triplicates of GAPDH control gene. We present here initial analysis of assessment of RQ-PCR MRD using 202 peripheral blood (PB) and bone marrow (BM) samples obtained at the end of the induction period on day 30. BM and PB samples at day 30 were available in 71% of patients, PB only available in 8% and BM only available in 21% of the patients. In cases with both BM and PB samples or in whom multiple markers for MRD assessment were available, the highest MRD value at day 30 was used to analyze relapse-free survival (RFS). 60% of cases had no detectable MRD at day 30 in either BM or PB, whereas 40% had detectable MRD ranging from 2.07 x 10⁻⁶ to 0.11. At day 30, BM samples were more likely to have detectable MRD than their matched PB samples (p=0.0043 by McNemar’s test). The 5-year RFS (+/− standard error) for patients with undetectable MRD is 0.93 (+/−0.023), versus patients with detectable MRD of > 0 to < 0.001 is 0.57 (+/− 0.08), versus detectable MRD of >= 0.001 is 0.25 (+/− 0.09) (log-rank p-value <0.0001). Stratified log-rank tests comparing the distribution of time to relapse among these three levels of MRD, controlling separately for risk status, treatment group, WBC, sex and age, remain highly significant (p-value < 0.0001). Patients with detectable MRD at day 30 have 9 times the risk of relapse compared to those with negative MRD (hazard ratio 95% CI of 4.1 to 20.2). Multivariable analysis of the prognostic value of MRD for the hazard of relapse was performed using Cox proportional hazards regression models, with covariates including MRD, risk status, treatment group, and MRD detection was demonstrated to be a highly significant independent prognostic factor in all models (p<0.0001). A Cox proportional hazards model demonstrated that the actual value of detectable MRD after controlling for the presence or absence of MRD is associated with relapse (p=0.004). We conclude that patients with undetectable MRD at day 30 have clearly superior outcome and patients with MRD more than 0.001 have significantly worse outcome. Quantification of MRD at completion of induction therapy using RQ-PCR provides a significant, reliable and independent clinical prognostic parameter to predict clinical outcome. Day 30 BM samples are more likely to have detectable disease than PB. Data on the impact of RQ-PCR at all time points will be presented. Based upon these results, risk-stratification therapy based on MRD level at completion of induction therapy will be incorporated into the next trial to attempt to improve the long-term survival rate of children at high risk of relapse and reduce overall the chemotherapy toxicity in children with B-lineage ALL.