What Is the Measure of Minimal Residual Disease in AML?

Get Permission

Relapse is the primary obstacle to cure in leukemia. The term minimal residual disease (MRD) was coined in the early 1990s to describe finding a disease-specific marker in the context of a morphologic-appearing remission. The technique first used for MRD detection was the Southern blot (!), but the advent of polymerase chain reaction technology, the windfall of mutation discovery to provide new targets for detection, and the explosion of flow cytometry advances made sensitive detection of MRD possible in many, if not most, hematologic malignancies. MRD is an established endpoint for treatment goals and milestones in chronic myeloid leukemia (CML) and, more recently, acute lymphocytic leukemia (ALL).

Recently, the MRD acronym has had its definition changed to ­measurable residual disease, to better reflect certain realities: first, it is a quantitative measure of disease burden (hence, measurable); second, there is nothing minimal about it in terms of the numbers of leukemia cells detected (millions) or in the effect of MRD on outcome (substantial). Indeed, the detection of MRD is a direct measure of disease burden (and thus, treatment efficacy) and is the strongest predictor of relapse in CML, ALL, and as the recent study by Short et al1 in JAMA Oncology demonstrates, acute myeloid leukemia (AML) as well; the study is well summarized in the current issue of The ASCO Post.

Jerald P. Radich, MD

Jerald P. Radich, MD

The carefully performed, detailed study parallels the work previously published in ALL.2 It shows the powerful association of MRD status and outcome across different disease subtypes and therapies, despite the measurement timepoint and regardless of whether assessment was done by polymerase chain reaction methods, next-generation sequencing, or flow cytometry.

Issues Surrounding MRD: Rich Avenues of Study

Now what? Is MRD now a done deal? Hardly.

Ultimately, one goal would be to use MRD as an early readout in clinical trials, which could potentially greatly shorten clinical trials (an example would be the 1-year BCR-ABL level endpoint in the registration trials of dasatinib and nilotinib in CML). However, beyond this, there are technical, clinical, and biologic issues that will be rich avenues of study.

One way to contextualize some issues is by examining Figure 1 in the Short et al article (Figure 1).1 The results show the effect of MRD status, with the hazard ratio for 5-year overall survival of 0.36. Note, however, that a substantial number of patients with MRD positivity do not relapse, whereas a substantial number of those cases of MRD negativity do relapse. Why?

Consider the top portion of the curves: patients without MRD who relapse. The vast majority of these deaths will be from relapse. Is this a technical issue, suggesting we need to build a better mousetrap that can look for MRD at deeper levels? The promise of a more sensitive assay (Note: MRD heads use “sensitivity” as a measure of detecting the needle in the background of hay, as opposed to the standard “sensitivity-specificity” of an assay) is alluring: Who wouldn’t want to be able to detect deeper and deeper responses? However, at some point, with an unlimited sensitive assay, it will be likely that all patients will have some level of MRD.

Given that at diagnosis the leukemia burden is estimated to be 1012 cells (an order of magnitude more than stars in the Milky Way), it is unlikely that “cure” ever means the 0 cells that mark as leukemia. Thus, the odds ratio of relapse for a lousy MRD assay is roughly 1; it increases as the test becomes more sensitive but then may return toward 1 as an even more sensitive assay picks up extremely low levels of cells that pose little threat to leukemia re-emergence (perhaps because they do not harbor an adequate array of mutations, or acquired additional mutations make them less pathogenic, and/or because the immune system rides herd on them).

Consider the bottom portion of the curves in Figure 1: patients with MRD who do not relapse. Various scenarios could account for this finding, including residual clones of lesser pathogenicity and/or immune restraint of clonal expansion (see above). Alternatively, could this be due to persistence of mutations associated with clonal hematopoiesis of inconclusive potential?3 Complicating this issue further is that it is likely the types of mutations likely matter in regard to the relapse rate. The amount of MRD and the risk of relapse may vary among patients with different mutation spectra.

Practical Clinical Questions

There are also practical questions in regard to clinical practice. Will changing therapy due to an MRD result actual improve outcomes (for example, moving from one chemotherapy to another, perhaps based on the changing mutation pattern after clonal selection)? Given that MRD is equally bad in the context of chemotherapy and allogeneic transplantation,4 will taking an MRD-positive patient to transplantation improve the odds of survival? Will the particular mutation burden matter in determining which cases of MRD should move to transplantation and which should not (and what type of transplant, since data suggest the effect of MRD is not as prominent in umbilical cord transplants as with other donors5)?

Reliable and reproducible MRD studies may make a change in trial design. Can using the kinetics and levels of MRD (ie, a quantitative variable) improve study efficiency compared with using MRD as a (dumb) categorical variable? For example, imagine measuring the decline in disease burden with drug A and drug B. Could rapid clinical trials call a “winner” (or at least, pick a drug that is likely not a winner) by comparing the slopes of the declining MRD? To make the patient’s life easier, will better assays allow MRD testing to be done on the peripheral blood, rather than bone marrow (as in CML)?

MRD is an important measure that has clear clinical utility. However, if MRD is to become a true surrogate endpoint, it will need to be standardized, scalable, and less expensive. The paper by Short et al shows the summation of a lot of important work by a lot of investigators. There is a lot more work to go. 

Dr. Radich is Clinical Research Division and Kurt Enslein Endowed Chair at the Fred Hutchinson Cancer Research Center, Seattle.

DISCLOSURE: Dr. Radich has served as a consultant or advisor to Amgen, Bristol Myers Squibb, Genentech, Novartis, and Takeda.


1. Short NJ, Zhou S, Fu C, et al: Association of measurable residual disease with survival outcomes in patients with acute myeloid leukemia: A systematic review and meta-analysis. JAMA Oncol. October 8, 2020 (early release online).

2. Berry DA, Zhou S, Higley H, et al: Association of minimal residual disease with clinical outcome in pediatric and adult acute lymphoblastic leukemia: A meta-analysis. JAMA Oncol 3:e170580, 2017.


Related Articles

Meta-analysis Shows Superior Disease-Free and Overall Survival With Measurable Residual Disease Negativity in AML

In a systematic review and meta-analysis reported in JAMA Oncology, Nicholas J. Short, MD, of The University of Texas MD Anderson Cancer Center, and colleagues, found that measurable residual disease (MRD) negativity is associated with superior disease-free and overall survival in patients with...