Comparison of the Predictive Value of MRD and PVA Score in Pediatric ALL.

Early identification of patients (pts) at risk for relapse allows for development of risk-adapted treatment strategies, thus steadily improving the outcome in pediatric acute lymphoblastic leukemia (ALL). Besides classic prognostic factors such as age, initial leukocyte count (WBC), genetic alterations and the immune phenotype, the so called PVA Score, summarizing the in vitro resistance of blasts against prednisone, vincristine and asparaginase, has been applied for treatment stratification in the CoALL protocol, a German multicenter study for children with ALL. Over the past years it has become increasingly clear that the in vivo response to chemotherapy assessed by detection of residual malignant cells (MRD) by PCR technique can be predictive of prognosis. Here we compare for the first time the relevance of in vitro (PVA Score) and in vivo (MRD) treatment response in a large cohort of 275 children with ALL, age 1–17 years, uniformly treated according to the CoALL protocols 05–92 to 07–03. Children with B cell precursor ALL (BCP-ALL) and T-ALL were analyzed separately. Bone marrow samples of 160 children with BCP-ALL and of 115 T-ALL pts diagnosed between 1992–2005 were prospectively assessed for PVA Score at diagnosis and MRD levels at day (d) 15, 29 and 43 after informed consent was obtained from the parents or legal guardians at the time of enrolment. Of note, 7 of the BCP-ALL and 14 of the T-ALL pts with late morphological response were excluded from analysis. Overall median MRD levels in BCP-ALL pts (MRDd15: 6×10e-4; MRDd29: 2×10e-5) were one log lower than in T-ALL (MRDd15: 9×10e-3; MRDd29: 3×10e-4). We detected no association between PVA Score and MRD level in BCP-ALL (correlation coefficient: r=0.15; p=0.15) and only a weak correlation in T-ALL pts (correlation coefficient: r=0.43; p=0.0003). When assessing the impact of the PVA Score on relapse free survival (RFS), in BCP-ALL only score 3+4 (good response) vs. 8+9 (poor response) was prognostically relevant (RFS 0.86±0.05 vs. 0.59±0.12; p=0.03), whereas in T-ALL no significant difference between these subgroups was found (RFS 0.71±0.1 vs. 0.68±0.1; p=0.62). In multivariate analysis PVA Score 3+4 vs. 8+9 remained the most relevant parameter for RFS in BCP-ALL (p=0.05) when compared to age and initial WBC. However, MRD levels were of even higher predictive power, especially at later time points: MRD negativity at d29 in BCP-ALL identified pts with significantly superior RFS (RFS MRD neg.: 0.9±0.05 vs. pos.: 0.7±0.05; p=0.003) and low MRD levels indicated a favorable outcome in T-ALL (RFS MRD <10e-3: 0.89±0.05 vs. MRD >10e-3: 0.68±0.07; p=0.001). Moreover, both BCP-ALL and T-ALL pts characterized by MRD levels >10e-3 on d43 exhibited a poor outcome (RFS BCP-ALL: 0.42±0.17; RFS T-ALL: 0.47±0.14). MRD remained an independent marker in multivariate analysis including initial WBC and age, both in BCP- (MRDd29: p=0.006; MRDd43: p=0.001) and T-ALL (MRDd29: p=0.003; MRDd43: p=0.015). By multivariate analysis, in T-ALL low MRD levels on d29 predicted superior RFS independently from the PVA Score (MRD: p=0.002 vs. PVA: p=0.09), whereas in BPC-ALL these parameters were not completely independent from each other at that early time point (MRD: p= 0.059 vs. PVA: p= 0.063) but became independent at d43 (MRD: p= 0.018 vs. PVA: p= 0.253). While the predictive value of the PVA Score was limited to BCP-ALL, MRD was an independent prognostic marker for both BCP- and T-ALL and reliably identified pts at low and high risk for relapse.