In recent series of adults with acute lymphoblastic leukemia (ALL), the GRAALL (ASH 2009, abstract 577) and other cooperative groups have confirmed the strong prognostic value of Ig/TCR minimal residual disease (MRD) on patient outcome. Despite this, age, WBC, CNS involvement, recurrent chromosomal translocations, and early response to steroids and chemotherapy remain frequently used to tailor post-remission therapy and envision allogeneic stem cell transplantation (SCT) in most adult ALL trials. We updated our MRD study, now with 262 patients who all achieved complete remission (CR) after the first induction and were assessed for MRD after induction (MRD1, at 6 weeks) and consolidation (MRD2, at 12 weeks). One hundred and fifty-eight patients had Philadelphia chromosome (Ph)-negative B-cell precursor ALL (BCP-ALL), while 104 had T-cell ALL (T-ALL). Since 107 of the BCP-ALL (68%) were studied for IKZF1 deletion and 90 of the T-ALL patients (87%) for NOTCH1/FBXW7 mutations, we were able to reassess the MRD significance according to these newly described oncogenic markers. These two covariates (i.e. MRD and IKZF1/NOTCH1/FBXW7 genetics) allowed us to redefine a much simpler yet more powerful stratification of disease risk in both BCP- ALL and T-ALL subsets.
All 262 patients studied (median age, 31.5 years) were treated in the GRAALL-2003 and GRAALL-2005 trials. Although they were younger and had more frequently circulating blasts, other characteristics and outcome did not differ from patients treated in the same trials but not assessed for MRD. Ig/TCR MRD levels were determined according to Euro-MRD guidelines (Leukemia 2007;21:604). IKZF1 deletions were assessed by multiplex multi-fluorescent PCR. NOTCH1/FBXW7 mutations were assessed as previously described (Blood 2009;113:3918). Multivariate backward stepwise selection Cox models were used for the cumulative incidence of relapse (CIR), disease-free (DFS) and overall survival (OS) endpoints, after censoring transplanted patients at SCT. Models were always adjusted on age (35-year cutoff), WBC (30 and 100 G/L cutoff for BCP- and T-ALL, respectively), CNS involvement, and trial. Additional BCP-specific covariates included CD20 expression, t(4;11) and t(1;19) translocations, and IKZF1 deletion. Additional T-specific covariates included cortical immunophenotype according to the EGIL classification, TLX1 overexpression, and NOTCH1/FBXW7 mutation. Finally, allogeneic SCT was re-evaluated in the newly defined risk subsets, as a time-dependent covariate.
An initial multivariate analysis revealed that among blood response after 1 week of steroid, bone marrow response after 2 weeks of therapy, and molecular response at both MRD1 and MRD2 time-points, the MRD2 level was the main and sole independent predictor of relapse (P=0.003). In BCP-ALL patients, persistent MRD2 and IKZF1 deletion were the only two independent factors identified, the presence of at least one factor defining 51% high-risk patients with 52% versus 15% CIR (HR, 3.8; P= 0.008), 41% versus 81% DFS (HR, 3.6; P= 0.005), and 54% versus 80% OS (HR, 3.9; P= 0.015) at 4 years. Allogeneic SCT in first CR significantly decreased relapse incidence and prolonged DFS in these new high-risk BCP-ALL patients (HR, 0.23 and 0.40; P= 0.016 and 0.05, respectively). In T-ALL patients, persistent MRD2 and lack of NOTCH1/FBXW7 mutation were the only two independent factors identified, the presence of at least one factor defining 49% high-risk patients with 64% versus 12% CIR (HR, 6.4; P= 0.002), 36% versus 88% DFS (HR, 6.4; P= 0.002), and 41% versus 95% OS (HR, 7.3; P= 0.015) at 4 years. SCT had no significant effect on relapse incidence and DFS in these new high-risk T-ALL patients.
In adult patients with Ph-negative ALL treated with the pediatric-inspired GRAALL regimen, IKZF1 deletion in BCP-ALL, NOTCH1/FBXW7 mutation in T-ALL, and MRD at 3 months in both subsets replace all classical risk factors, leading to a new simplified prognostic scoring system based only on IKZF1 and NOTCH1/FBXW7 genetics and MRD clearance. This new risk score identifies approximately half of the patients as good-risk, with a relapse incidence as low as 10–15%. It will be validated and used prospectively in the next generation of GRAALL trials, to stratify both new drug evaluation and SCT in first CR.
No relevant conflicts of interest to declare.
Asterisk with author names denotes non-ASH members.