Page 141 - 2019_10 resto del Mondo_web

Meta-analysis of the impact of MRD in adult B-ALL
prolong RFS compared with chemotherapy among those who did not achieve an early MRD response, but was no better than chemotherapy in patients who did achieve an early MRD response.59 Likewise, the PETHEMA ALL- AR03 trial used MRD to guide treatment decisions at the end of consolidation, and found that HSCT could be avoided in patients who reached MRD negativity without adversely affecting their prognosis.60 Similarly, a study using the Northern Italy Leukemia Group trial protocol 10/07 implemented risk-stratification to define a cohort of patients with an MRD response who could achieve good outcomes with conventional maintenance instead of HSCT after induction therapy.61 These studies provide evi- dence that MRD can play a role in sparing patients from the risks associated with transplantation, without nega- tively affecting survival outcomes.6 New systemic thera- pies may offer alternatives to HSCT; for example, a phase II study (published in full since the literature review was conducted) found that blinatumomab could induce a com- plete MRD response in over three-quarters of patients with MRD-positive ALL in first-line treatment;62 blinatu- momab is now approved for use in patients with MRD- positive ALL in the USA.63 There is a need to establish the role of MRD testing in guiding treatment in each subgroup of patients with ALL and to determine an evidence-based treatment protocol to optimize patients’ outcomes.
Ph-positive and Ph-negative disease differ not just in terms of prognosis, but also in treatment regimens and method of MRD assessment. We analyzed RFS separately for Ph-positive and Ph-negative studies, and the prognos- tic ability of MRD negativity remained the same regard- less of Ph status, with no discernible difference among subgroups. Subdividing the total population according to Ph status led to shrinkage or elimination of certain sub- groups, but these results again point to the consistency of effect of MRD negativity across clinically relevant popula- tions. It will be important in future research to further dis- sect the Ph-like subset of patients (who have a similar gene expression profile to Ph-positive ALL but without the targetable BCR-ABL translocation), given their poor prognosis and persistence of MRD.64
The question of how sensitive MRD assays need to be remains to be resolved. In a study of patients with untreat- ed Ph-negative B-ALL or T-ALL, MRD levels of 10−3 or greater were associated with a poor prognosis.45 In our study, no real differences in the association between MRD level and outcome were seen, regardless of whether the sensitivity threshold was 10−3, 10−4 or 10−5. Therefore, the threshold of 10−4 recommended by ESMO seems appropri- ate.6
The studies included in this systematic review and meta-analysis represent a broad range of patient sub- groups and treatment regimens. Consequently, the find- ings should be generalizable across the whole B-ALL pop- ulation. Nevertheless, these analyses have some limita- tions. Given the wide range of treatment regimens, fol- low-up times and methodologies for assessing MRD, the data should be interpreted with caution. Heterogeneity was high (as measured using I2), but this was expected because it reflects the nature of studies included: they were not randomized and included different populations of patients, methods of analysis, follow-up times and study designs. In addition, treatments administered after MRD assessment, particularly HSCT, would be a major confounding factor when assessing the relationship
between MRD and survival outcomes. We investigated sources of heterogeneity by using predetermined sub- group analyses of factors that might affect MRD and sur- vival, and by using meta-regression. The subgroup analy- ses support the overall conclusion that MRD negativity is associated with better survival outcomes than MRD posi- tivity. Heterogeneity remained high after accounting for key variables, indicating that there could be additional fac- tors that affect the relative effect of MRD status on out- comes, which we were unable to account for in this analy- sis. Furthermore, for some subgroups very few studies were available, and data were sparse, meaning that the reliability of some of the subgroup results was subopti- mal.
In some cases, time-to-event data were missing, and hazard ratios were calculated using published methodolo- gies to account for this.30 However, the meta-regression for RFS indicated that the Tierney method used to calcu- late hazard ratios was unlikely to influence the magnitude of the effect of MRD negativity.
It would be of interest to study the impact of MRD on bone marrow relapse. Although RFS is a good proxy for relapse, deaths in remission (e.g., following HSCT) may affect the results. This meta-analysis focused on MRD as a risk factor for systemic relapse and did not examine MRD as a predictor of extramedullary relapse or death from other causes, such as transplant-related mortality; therefore, different rates of death across subgroups from causes other than bone marrow relapse could confound the results. The extent of confounding due to HSCT-relat- ed mortality is unclear, however, and only two of the studies included in the meta-analysis comprised solely patients who did not undergo HSCT. Overall, the strong relationship observed in our study, and that by Berry et al.,28 highlights the high prognostic value of MRD, despite this confounding factor.
In four studies, it was not possible to separate the B-cell and T-cell populations; however, the majority of patients in these studies had B-ALL (66% to 79%), and there was no difference in the prognostic value of MRD between the B-ALL and mixed studies for either RFS or OS. It should also be noted that our analysis only included two studies in CR2 or later; however, results in this subgroup were consistent with the overall findings.
In conclusion, this systematic review and subsequent meta-analysis have generated support for the use of MRD as a prognostic marker in the management of patients with ALL. Overall, and in all subgroups analyzed, MRD negativity was associated with better survival outcomes than MRD positivity. This finding appeared relatively unaffected by variation in the timing or sensitivity thresh- old of the MRD assay applied, though other factors – such as the window of opportunity for treatment before relapse – may need to be considered when refining MRD assess- ment methodology to optimize its value in practice. Detection of MRD offers a promising clinical tool that may help clinicians to harness the potential of emerging targeted therapies for ALL.
Acknowledgments
The authors would like to thank James O’Kelly PhD of Amgen Ltd who provided editorial support, and Kim Allcott PhD of Oxford PharmaGenesis who assisted with the systematic litera- ture review and provided editorial support. This research was funded by Amgen.
haematologica | 2019; 104(10)
2037