How complete must an AML remission be?

In this issue of Blood, Le et al discuss the clinical benefit associated with achievement of complete remission with partial hematological recovery (CRh) for patients with acute myeloid leukemia (AML) treated with molecularly targeted drugs.¹ They show that, despite blood neutrophil and platelet levels as low as half of that conventionally required for a full complete remission (CR), those achieving CRh experienced potentially important benefits of transfusion independence and increased time to severe infection or severe bleeding.

Broadly speaking, new medicines are approved by the US Food and Drug Administration (FDA) with the expectation that they will help patients live longer or better. Such approvals do not require direct demonstration of this clinical benefit (ie, how a patient feels, functions, or survives) but may instead rely on either validated established or even, in accelerated approvals, “reasonably likely to predict” surrogate measures of clinical benefit. Response assessment is an important intermediate end point, frequently used in oncology clinical trials, that has been the basis of multiple cancer drug approvals,² including for patients with AML.³ Response assessments, of course, also have other important clinical uses, including early assessment of the antitumor efficacy of initial therapy and quantification of hematopoietic recovery.^4,5 

Le et al and colleagues from the FDA analyzed data submitted to their agency in support of marketing applications for 4 molecularly targeted therapies (gilteritinib, ivosidenib, enasidenib, and olutasidenib), each used as a single agent, for adults with AML that was predominantly relapsed/refractory. Both the conventional CR and the CRh remission response criteria evaluated here required bone marrow blasts to be <5%, with no detection of circulating blasts or extramedullary disease, and no transfusions in the past 7 days. Rather than the historical, albeit arbitrary, requirement of absolute neutrophil count of >1 × 10⁹ per liter and 100 × 10⁹ platelets per liter for a “full” CR,⁵ the response criterion threshold chosen for CRh allowed blood counts down to half of these levels. Overall, 42.4% of patients in these trials achieved a response on treatment (no overt evidence of persistent leukemia), with 19.4% having a best response of CR, only 6.2% achieving a best response of CRh, and the remaining 16.9% having neither (ie, a “lesser hematological recovery” [CRL]). Unsurprisingly, those with response (CR, CRh, or CRL) had superior overall survival to those who did not respond to treatment, but those with a best response of CR had superior survival to those with CRh and CRL. Again, unsurprisingly, those with better hematological recovery (CR and CRh) were more likely to be transfusion independent for at least 56 days than those with lower blood counts (CRL) and nonresponders. Compared with nonresponders, those achieving CRh also had reduced risk over time of severe bleeding or severe infection that was more consistent with that seen for those achieving a best response of CR rather than those with only CRL.

Our regulatory colleagues deserve our congratulations and thanks for producing this analysis in support of a CRh best response in patients with AML treated with such agents being associated with important palliative clinical benefit and hence representing, together with CR, an appropriate end point goal in AML clinical trials. It is important that the FDA performs, and makes public, research like this providing an evidentiary basis for past and future decisions. There are limitations, however.

Most obviously, the leukemias of patients in these studies represent only a subset of the genetic causes of AML, and the single-agent, nonintensive, therapy approach represents only 1 way in which such patients may be treated. Although eradication of any trace of leukemia at the expense of an infectious or bleeding death in aplasia would obviously represent an undesirable pyrrhic victory, achievement of hematological count recovery may be less important for those who subsequently undergo potentially curative allogeneic hematopoietic cell transplant (alloHCT) without delay. Trial investigators were not blinded to the results of response assessments, and remission achievement is often 1 criterion used in determining eligibility for subsequent alloHCT. Further compounding the uncertainty regarding generalizability is the relapsed and refractory disease setting, in which patients had received heterogeneous prior therapy, including in some cases alloHCT, resulting in potential differences in persistent disease biology, degree of residual normal hematopoietic reserve, and time (and opportunity for mucosal repair) since last therapy. Most fundamentally, though, both CR and CRh are extremely “low-fidelity” measurements of antileukemic efficacy and can encompass wide ranges of residual disease burden.^6,7 Persistence of low-level residual disease is 1 explanation of suboptimal blood count recovery. Attempts to correlate best response categories with molecular assessments of measurable residual disease (MRD) were, as with patient-reported outcomes, limited by the tools and data sets available. The testing methods used for AML MRD were heterogeneous, and only a few patients achieved MRD negativity. Additionally, although persistence of FLT3-ITD in CR now has well-established clinical implications,⁸ the association of mutated IDH1 and IDH2 detection in remission and subsequent clinical outcome is less clear.⁹ As more effective agents are developed, or combination regimens are used, rates of response are likely to be higher than seen here with single agents, making CR a less useful tool to quantify and discriminate responses to different therapies. This has already been seen in other blood cancer types¹⁰; for example, on 12 April 2024, the FDA Oncology Drugs Advisory Committee voted 12-0 in support of the use of MRD as an accelerated approval end point in multiple myeloma clinical trials.

Ultimately, all response criterion thresholds are somewhat arbitrary and context dependent; we treat patients, not numbers. For now, though, CRh, like CR, represents not only a level of response for patients with AML that many clinicians would feel comfortable limiting prophylactic antibiotic use and platelet transfusions, and deintensifying the follow-up monitoring schedule, but also a clinical trial end point response that may give helpful information in predicting the likely clinical benefit required for new drug approvals.

Conflict-of-interest disclosure: The author declares no competing financial interests.