ARF way to Ph⁺ ALL stratification?

In this issue of Blood, Pfeifer et al have determined that CDKN2A/2B deletions are “a strong and independent prognostic marker for predicting risk of relapse and overall survival” when adults with Philadelphia chromosome positive acute lymphoblastic leukemia (Ph⁺ ALL) are treated with both imatinib and allogenic stem cell transplantation (aSCT).¹ 

In 2018, the presence of t(9;22) in newly diagnosed ALL has positive connotations. We now know that when tyrosine kinase inhibitors (TKIs) are added immediately to the cytotoxic therapy of Ph⁺ ALL, the rate of complete remission (CR) reaches 95% to 100% with therapy failure being more likely because of treatment-related mortality (TRM) than therapy resistance. Indeed, convincing randomized controlled trial data²  demonstrated that a reduction in the intensity of the initial cytotoxic therapy can significantly reduce early TRM. Now, almost 100% of patients with Ph⁺ ALL will reach CR and become assessable for aSCT. Can we refine our understanding of who are the best candidates? Ongoing clinical studies are addressing the question of whether patients with Ph⁺ ALL achieving early complete molecular remissions with later-generation TKIs can be spared the toxicity of aSCT. However, the current standard of care remains aSCT, where possible.

It has long been known that frequent, nonrandom “additional chromosome abnormalities” impact long-term outcome in Ph⁺ ALL; gain of a second Ph (+der 22), high hyperdiploidy, and loss of chromosomes 7, 7p, and/or 9p all impact relapse risk and survival. Loss of 9p, the locus for cyclin-dependent kinase inhibitor 2 A and B (CDKN2A/2B), has already been associated with a poorer relapse-free survival in 3 clinical studies of Ph⁺ ALL.^3-5  Pfeifer et al used single nucleotide polymorphism arrays and multiplex ligation-dependent probe amplification to uncover the impact of common ALL-related gene deletions of 97 patients treated on 2 German Multicenter ALL Study Group protocols for Ph⁺ ALL. The biological plausibility of their finding is supported not only by the prior findings of poor outcome upon loss of 9p, but also by insight into the function of the genes. CDKN2A encodes several proteins, the most well-studied of which are p16(INK4a) and p14(ARF). INK4a controls cell growth and division via blocking the activity of CDK4 and CDK6 on cell cycle progression, and ARF is a physiological inhibitor of MDM2, an E3 ubiquitin ligase that controls the activity and stability of P53. CDKN2B encodes a cyclin-dependent kinase inhibitor that also forms a complex with CDK4 or CDK6 and controls cell cycle G1 progression. There is already work that associates loss of ARF function with enhanced oncogenicity and loss of imatinib response in murine models of BCR-ABL–induced ALL.⁶ 

It is important to note the caveats that limit the applicability of, but by no means negate, the findings. Although patients in this study had been treated “uniformly” with chemotherapy, a significant subset started imatinib later during therapy than others. Late imatinib start was clearly associated with a very highly significantly inferior overall survival (OS), disease-free survival (DFS), and remission duration in univariate analysis and remained a highly significant outcome factor in the multivariable analysis. Additionally, although all patients in this study received aSCT, conditioning regimens were not uniform. In particular, reduced intensity conditioning was associated with a significantly inferior OS and DFS than myeloablative conditioning. It is impossible to quantify the bias that is introduced by individual aSCT conditioning regimen choices within a trial protocol, but it is highly likely that these decisions are nonrandom. Heterogeneity of conditioning regimens is a common problem in studies involving aSCT and always complicates and confounds analyses by therapy received.

What the authors did not find is also of interest. There was no apparent relationship between minimal residual disease (MRD) level by BCR-ABL1 quantification prior to aSCT and outcome, although not all cases were analyzable for MRD. Hence, it appears that the poor prognostic relevance of CDKN2A/B deletions cannot necessarily be accounted for during the standard monitoring process, which is of obvious concern whenever molecular remission is used to define a group of patients who might have a treatment reduction. However, it may also suggest that aSCT can overcome the poor predictive outcome of pre-aSCT MRD. Alternatively, it may simply reflect that the effect size is modest and complex and that the number of patients and events analyzed here is simply too small to deconvolute it. Additionally, the authors did not find a relationship between IKZF1 deletions and outcome. On this point, it is also difficult to be sure whether this is a matter of study population size, combined with the unexpectedly high proportion of patients (75%) reported with IKZF1 deletions.

What is the implication for the treating clinician? Although the association of CDKN2A/B deletion with inferior outcome in the therapeutic circumstances reported is highly significant (hazard ratio [HR] for remission duration, 4.066; P = .0061), in the multivariable analysis it is nonetheless dwarfed by the impact of giving appropriate therapy at the appropriate time; for example, a late imatinib start is associated with HR for remission duration of 7.156 (P < .0001). The modest clinical data set in this study also shows that aSCT conditioning regimen is important, reflecting data from large registry studies.⁷  So, although it would be of interest to treating clinicians to know the CDKN2A/B status, it is important that CDKN2A/B deletions are not interpreted as a reason not to carry out aSCT in patients with Ph⁺ ALL. Although there is an inferior outcome for patients with CDKN2A/B deletions as compared with those without the deletion, the Kaplan-Meier curves for OS, DFS, and remission duration all reach an apparent plateau at a level that exceeds expectations from the pre-TKI era; certainly, the data do not suggest aSCT is futile. Furthermore, the impact of CDKN2A/B deletions in a population that did not receive aSCT has not yet been evaluated. Finally, we do not know if the finding will be generalizable to patients who received initial therapy with second- or subsequent generation TKIs.

Moving forward from this very interesting article, trialists and scientists need to rush to their existing data sets and models to further interrogate this fascinating clinical finding. I echo the authors’ careful conclusions that their data warrant the further prospective evaluation of CDKN2A/2B for risk stratification of adult Ph⁺ ALL.