Abstract
High hyperdiploidy (HeH) in acute lymphoblastic leukaemia (ALL) is defined by the non-random acquisition of multiple chromosomes, resulting in a modal number of 51-65 chromosomes. HeH is the most common primary abnormality in paediatric ALL and is associated with an excellent outcome. However, relapses do occur and prevalence of the subgroup means that they account for a sizeable proportion of relapsed cases. Hence identifying risk factors within this subtype remains important. Numerous studies have identified specific trisomies, modal numbers and structural rearrangements as prognostic factors. Despite these efforts, a consensus of the key risk factors has failed to emerge. We present a comprehensive analysis of 1181 children with HeH ALL, comprising discovery and validation cohorts: ALL97/99 (n=456) and UKALL2003 (n=725). The event-free survival (EFS), relapse risk (RR) and overall survival (OS) of HeH patients were: 84%/13%/93% and 91%/6%/95%, respectively.
Initially, we examined previously reported risk factors in both cohorts separately. We did not observe any impact on outcome when patients were sub-divided using our copy number alteration (CNA) risk profile, which is based on the deletion status of the 8 most often affected genes. Similarly, there was no relationship between modal number and outcome whether we examined modal number as a continuous variable or in categories (51-53 v 54-57 v 58-65). Although the EFS hazard ratios (HZR) for patients with a double (+4, +10) or triple trisomy (+4, +10, +17) were <1, the difference was not statistically significant in either cohort.
HeH karyotypes harbour 5-19 additional chromosomes and although some chromosomes are more likely to be gained than others, there are thousands of combinations. In order to assess the optimal number of chromosomes required for predicting outcome, we used the discovery cohort to calculate the area under the curve (AUC) associated with each individual trisomy as well as all possible combinations of 2-6 chromosomes. This analysis revealed that the average AUC (an estimate for predictive power) increased with the number of chromosomes, but only up to 4, suggesting that there was no advantage in considering 5 or more chromosomes. Next, we derived EFS HZR from univariate Cox regression models for each chromosome and ranked them in order of p value. The HZR for the top 4 chromosomes were +18 (0.43, p<0.001), +20 (2.33, p=0.01), +17 (0.68, p=0.09) and +5 (1.52, p=0.09). Therefore, patients with +17 and +18 but not +5 or +20 (HeH good risk, HeHgr) are predicted to have a better outcome compared to HeH poor risk (HeHpr) patients. This was true for patients in the discovery cohort but also, more importantly, in the validation cohort (table). In UKALL2003, the HeHpr group was associated with a higher incidence of end of induction (EOI) MRD (>0.01%) (56% v 47%, p=0.04) and IKZF1 deletions (12% v 2%, p=0.002). Factoring in EOI MRD, revealed that all UKALL2003 HeH patients had an excellent outcome except those with EOI MRD plus a poor risk HeH profile (table). This subset represented 33% of HeH patients but captured 71% (15/21) of the relapses recorded among UKALL2003 patients.
In conclusion, we demonstrate that with the exception of EOI MRD the most important risk factor in HeH is the pattern of chromosomal gain. This result is in keeping with recent genomic studies, which have failed to identify a common underlying mutation, suggesting that the key driving event is disordered gene expression caused by the pattern of chromosomal gain. It is reassuring that the trisomies identified in this study have all previously been proposed as risk factors, providing a framework for further investigations aimed at elucidating precisely which genes are determining treatment response in HeH. Table.GroupHeH good riskHeH poor riskALL97/99Number of patients188268EFS (95% CI)91% (86-95)79% (74-84%)Hazard ratio (95% CI), p value12.81 (1.64-4.80%), p<0.001UKALL2003Number of patients306419EFS (95% CI)95% (91-97%)88% (83-91%)Hazard ratio (95% CI), p value12.07 (1.17-3.67%), p=0.01EOI MRD (>0.01%) Number of patients*95162EFS (95% CI)95%(88-98)83%(75-88)Hazard ratio (95% CI), p value13.55 (1.23-10.29), p=0.02EOI MRD (<0.01%) Number of patients*109127EFS (95% CI)97%(90-99)94%(87-97)Hazard ratio (95% CI), p value13.05 (0.63-14.73), p=0.2*An MRD result was only available for 493 UKALL2003 patients.
Disclosures
No relevant conflicts of interest to declare.
A study on the predictability of acute lymphoblastic leukaemia response to treatment using a hybrid oncosimulator