Overall survival (OS) for children with ALL treated on contemporary protocols is now > 90%. To further improve outcomes, stratification must direct intensive treatment to those patients with the highest risk of relapse. Whilst minimal residual disease (MRD) has emerged as the most powerful predictor of outcome, its use in isolation is insufficient to identify those patients with the greatest risk. To address this, we performed a detailed analysis of the entire patient cohort from the recently reported large multi-center randomized controlled trial, UKALL 2003. Integration of end of induction (EOI) MRD with other predictive factors - cytogenetics, age and white cell count (WCC) - allowed the identification of very high risk groups that are likely to benefit from early treatment intensification or novel therapies.
The UKALL 2003 trial recruited 3126 patients aged 1-25 years with Philadelphia-negative ALL between Oct 1, 2003, and June 30, 2011. Treatment was initially stratified based on NCI risk and cytogenetic results. Subsequent treatment was directed by EOI MRD measured using EuroMRD approved real-time quantitative PCR for immunoglobulin and T-cell receptor gene rearrangements. Analysis was performed on 2666 patients with available MRD results and focused on the systematic integration of MRD results with other patient characteristics to identify subgroups with a very poor OS (<75% at 5 years). MRD was considered positive at a level ≥0.005% and low risk if <0.005% unless otherwise stated.
Both MRD and cytogenetic risk group were independently predictive of survival. In particular, patients with high risk (HR) genetic abnormalities - low hypodiploidy, near haploidy, KMT2A (MLL) rearranged, iAMP21 and t(17;19)(q23;p13) (n=100) had a poor outcome with 5 yr OS of 76.0%. Combining these factors showed that MRD could stratify the HR cytogenetic subgroup with a significant difference in survival between those with low risk MRD and positive MRD (5yr OS 93.0 vs. 63.1%, p=0.002, n=43 vs. 57). Crucially, this demonstrates that HR cytogenetics are only truly poor risk in the context of poor treatment response. Importantly, there were almost no survivors following relapse in the HR cytogenetic/MRD positive group, whilst 50% of those with low risk MRD could be salvaged after relapse. In addition, whilst t(1;19)(q23;p13)was not considered a HR cytogenetic factor in this trial, combination with positive EOI MRD identified an additional subgroup of 20 patients with a 5yr OS of only 74.7% (vs. 96.7% in MRD low risk patients, p=0.003).
Whilst WCC at presentation was also predictive of outcome, this difference was significant only in those with B-ALL and not T-ALL. In B-ALL, WCC>100 x109/L was associated with significantly poorer survival (5yr OS 95.1 vs. 84.0%, p<0.0001) whilst high WCC in T-ALL had little effect on survival, even at WCC >200 x109/L (5yr OS 90.2 vs. 86.7%, p=0.283). Integration of MRD results allowed the identification of a very high risk group with B-ALL, WCC>100 and positive EOI MRD (n=80) which had a 5yr OS of only 71.3%. Further breakdown by cytogenetic risk group showed those patients with good risk cytogenetics (ETV6-RUNX1, high hyperdiploidy) (n=19) had a remarkably good survival despite WCC>100 (5yr OS 94%). In contrast, those with intermediate (n=34) or HR cytogenetics (n=15) had a very poor 5yr OS (65.7% and 60.0% respectively).
Older age (≥10yr) was also associated with poorer outcomes compared to those <10yr (5yr OS 95.8 vs. 87.4%, p<0.0001). Integration with MRD results further separated these groups but did not identify a very high risk subgroup. However, combining older age with a higher EOI MRD cut-off of 1% successfully identified a subgroup (n=96) with a very poor 5yr OS of 51.3% (vs. 91.4% in those with MRD<1%, p<0.0001). Interestingly, this was independent of cytogenetic risk group.
Through the integration of EOI MRD with other risk factors in a large pediatric ALL cohort we have identified a population of patients with a very poor OS. When combined, these criteria select only 6% of all patients, but capture 35% of all deaths suggesting current treatment is suboptimal in this group. This information is essential for the planning of future trials to ensure that treatment intensification and novel therapies are targeted at those patients at greatest risk, whilst avoiding unnecessary treatment in those patients that already experience good outcomes.
No relevant conflicts of interest to declare.
Asterisk with author names denotes non-ASH members.