Association of Hematologic Response and Assay Sensitivity on the Prognostic Impact of Measurable Residual Disease in Acute Myeloid Leukemia: A Systematic Review and Meta-Analysis

Abstract

Measurable residual disease (MRD) is associated with relapse and survival in acute myeloid leukemia (AML). We aimed to quantify the impact of MRD on outcomes across clinical contexts, including its association with hematologic response and MRD assay sensitivity. We performed systematic literature review and meta-analysis of 48 studies that reported the association between MRD and overall survival (OS) or disease-free survival (DFS) in AML and provided information on the MRD threshold used and the hematologic response of the study population. Among studies limited to patients in complete remission (CR), the estimated 5-year OS for the MRD-negative and MRD-positive groups was 67% (95% Bayesian credible interval [CrI], 53%-77%) and 31% (95% CrI, 18%-44%), respectively. Achievement of an MRD-negative response was associated with superior DFS and OS, regardless of MRD threshold or analytic sensitivity. Among patients in CR, the benefit of MRD negativity was highest in studies using an MRD cutoff less than 0.1%. The beneficial impact of MRD negativity was observed across MRD assays and timing of MRD assessment. In patients with AML in morphological remission, achievement of MRD negativity is associated with superior DFS and OS, irrespective of hematologic response or the MRD threshold used.

Introduction

Acute myeloid leukemia (AML) is an aggressive hematologic malignancy with a variable prognosis that is influenced by both patient-related and disease-related factors.¹ While pre-treatment factors such as age, karyotype, and somatic mutations provide important information about an individual patient’s risk of disease relapse and survival, post-treatment response assessment is also strongly prognostic for long-term outcomes.² Several studies have suggested that patients who achieve a morphologic remission and also have full peripheral blood recovery (i.e. achieve a complete remission [CR]) have superior outcomes to those with incomplete hematopoietic recovery.^3-5 Given that determination of morphologic remission only requires bone marrow blasts to be less than 5%, the sensitivity of this microscopy-based response assessment is limited, which in turn limits its discrimination for relapse risk. However, more sensitive techniques to assess for small amounts of residual leukemia—referred to as “measurable residual disease” (MRD)—have been developed that provide additional prognostic information and can identify patients in morphologic and/or hematologic remission who still have high rates of relapse and AML-related death.⁶ In a meta-analysis of 81 studies, the estimated disease-free survival (DFS) for patients with and without MRD after AML-directed therapy was 25% and 64%, respectively, and the 5-year overall survival (OS) was 34% and 68%, respectively.⁷ Importantly, the significant prognostic impact of MRD was observed across age groups, AML subtypes, time of MRD assessment, specimen source, and MRD detection methods.

Given the strong association between MRD and long-term outcomes, there is hope that an MRD response may also serve as a surrogate marker for survival in clinical trials, thereby allowing for more rapid assessment of novel drugs and regimens.⁸ However, while the impact of MRD in patients with AML is well-established through results from dozens of studies, there are still unanswered questions regarding its potential use as a surrogate endpoint for regulatory decision-making.⁹ First, MRD response and hematologic recovery are both independent predictors of survival in AML,⁴ and the survival estimates for patients who achieve MRD negativity with initial therapy or who remain MRD-positive may be different, depending on their hematologic response. It is therefore important to establish the relative impact of MRD in the context of a specific hematologic response, particularly for those in CR. Second, there is substantial variability in the analytical sensitivity of the MRD assays used in many AML studies, both across different types of MRD assays (e.g. multiparameter flow cytometry or quantitative polymerase chain reaction) and within specific classes of MRD tests. Some analyses have also classified patients as “MRD-negative” despite detection of small amounts of disease below a pre-specified threshold.¹⁰ Estimates of the relative benefit of MRD negativity must therefore consider the sensitivity of the specific MRD assay and/or the threshold used to dichotomize MRD-negative and MRD-positive responses. In the previous meta-analysis of the association between MRD and survival in AML, the impact of hematologic response (i.e. CR versus lesser responses) and MRD threshold/sensitivity was not specifically evaluated. This limitation was considered a potential barrier to regulatory acceptance of MRD response as a surrogate endpoint in AML.⁹

In order to better quantify both the relative and absolute benefits of achieving MRD negativity in AML, we performed a meta-analysis of AML studies that reported the association of MRD and clinical outcomes and which also reported the hematologic response of the study population and the MRD threshold used to determine an “MRD-negative” response. By restricting our analysis to these studies, our goal was to provide more accurate survival estimates associated with MRD-negative and MRD-positive responses in these specific clinical scenarios, thereby providing more robust quantitative support for regulatory and clinical uses of MRD as a potential surrogate marker in AML.

Methods

Search strategy and selection criteria

Two investigators (N.J.S. and F.R.) conducted independent searches of PubMed, MEDLINE, and Embase for papers published between January 1, 2000 and October 1, 2018 that included the keywords “AML,” “acute myeloid leukemia,” or “acute myelogenous leukemia,” in combination with the keywords “MRD,” “minimal residual disease,” or “measurable residual disease.” Disease experts reviewed the list and provided recommendations for additional inclusions. Our search and study screening process adhered to the PRISMA Guidelines.¹¹ We excluded review articles, non-English articles, studies that included non-AML diseases in their analysis, studies that used overlapping datasets, and studies that had insufficient description of MRD information or lacked correlation with DFS or OS. We also excluded studies that only reported outcomes after hematopoietic stem cell transplant (HSCT). In total, 81 articles meeting these inclusion criteria were found, as previously described.⁷ As the purpose of this meta-analysis was to specifically evaluate the impact of MRD according to MRD threshold and hematologic response (i.e. in particular, CR versus lesser responses), we excluded an additional 6 studies that did not report the hematologic response of the study population, 22 studies that did not define the MRD threshold used to determine MRD negativity and 5 studies that did not report either of these parameters, resulting in 48 evaluable studies (Figure 1).

Figure 1 –

Flow diagram of the study selection process. Forty-eight articles were ultimately included in this meta-analysis.

Statistical analyses

We extracted data from evaluable publications and collected survival information in MRD-negative and MRD-positive subgroups as previously described.⁷ Data from Kaplan-Meier (KM) curves were used, if available. In cases in which KM curves were not available, we used reported hazard ratios (HRs) and their confidence intervals (CIs). For articles that only provided survival proportions at fixed time points, we used the reported survival rates and standard deviations.

We applied a Bayesian hierarchical model with time-varying hazard effects and allowed for random effects from different studies. The hazard functions were modeled with stepwise linear functions on every 6-month time segments and truncated at 11 years. The survival outcomes reconstructed from the reported Kaplan-Meier curves (type I), observed HRs and the corresponding standard errors (type II), and estimated survival rates (type III) jointly contributed to the estimation and inference of the overall survival probabilities and averaged HRs. To evaluate the impact of MRD by hematologic response, we repeated analyses for studies that only included patients in CR at the time of MRD assessment and examined its interactions by MRD detection methods or assessment time. We also stratified studies with three categories of MRD threshold values (<0.1%, 0.1% and >0.1%) by which patients were classified as “MRD-negative”, and conducted subgroup analyses for MRD threshold alone, or its interactions with MRD detection method or MRD assessment time, respectively.

OS was defined from the time of treatment start until death or last follow-up. DFS was defined as the time of remission until relapse, death, or last follow-up. We included disease-free, event-free, leukemia-free, recurrence-free, and relapse-free survival in the definition of DFS. The estimated DFS and OS distribution for patients in the MRD-negative and MRD-positive groups were plotted, either marginally or stratified by three cut-off categories. Shaded areas demonstrated the 95% Bayesian credible intervals (CrIs) of the estimated survival curves. The difference in estimated survival probabilities and restricted mean survival time (RMST) at 5 years were also estimated to summarize the discrepancy between the MRD-negative and MRD-positive groups.

Markov chain Monte Carlo (MCMC) methods were used to generate posterior estimates from the joint posterior distributions of model parameters. We employed rjags package in statistical software R (version 4.1.1) to interface with the JAGS (Just Another Gibbs Sampler, version 4.3.0) for data analysis.

Results

Included studies

Forty-eight distinct publications of 7,146 patients were included in this analysis (9 studies for OS, 13 studies for DFS, and 26 for both outcomes).^12-59 The characteristics of studies included in each subgroup analysis are summarized in Table 1, and the individual studies are shown in Supplemental Table 1. Studies were stratified by MRD thresholds of <0.1%, =0.1%, and >0.1%. For the MRD threshold of <0.1%, 14 studies were included in the OS and DFS analyses; for the MRD threshold of 0.1%, 11 studies were included in the OS analysis and 13 studies were included in the DFS analysis; and for the MRD threshold of >0.1%, 10 studies were included in the OS analysis and 14 studies were included in the DFS analysis. For the analysis of MRD in patients who achieved CR, 21 studies were included in the OS analysis and 23 studies were included in the DFS analysis.

Table 1 –

Included studies by subgroup

Studies included, N (%)
Subgroup	In OS analysis	In DFS analysis
MRD threshold	n=35	n=39
< 0.1%	14 (40)	14 (36)
= 0.1%	11 (31)	13 (33)
> 0.1%	10 (28)	14 (36)
MRD time point	n=34	n=38
Induction	28 (82)	33 (87)
Consolidation	5 (15)	5 (13)
After consolidation	13 (38)	12 (32)
MRD detection method	n=35	n=39
MFC	20 (57)	25 (64)
PCR	13 (37)	12 (31)
Others	5 (14)	4 (10)
CR status	n=35	n=39
CR	21 (60)	23 (59)
CR/CRi	4 (11)	4 (10)
CR/CRi/MLFS	10 (29)	12 (31)

Abbreviations: OS, overall survival; DFS, disease-free survival; MRD, measurable residual disease; MFC, multiparameter flow cytometry; PCR, polymerase chain reaction; CR, complete remission; CRi, CR with incomplete hematologic recovery; MLFS, morphological leukemia-free state.

Association of MRD status on survival outcomes

The hazard ratios for OS and DFS for each individual study are shown in Supplemental Figure 1A-B. Forty-one studies (85%) showed a statistically significant difference in OS between MRD-negative and MRD-positive patients, and 44 studies (92%) showed a statistically significant difference in DFS. The survival curves for MRD-negative and MRD-positive groups in the full dataset of 48 studies are shown in Figure 2A-B. Both OS and DFS were superior for those who achieved MRD negativity. The estimated 5-year OS for patients who achieved MRD negativity was 65% (95% CrI, 59%-71%), compared with 30% (95% CrI, 23%-36%) for those who were MRD-positive. Similarly, the estimated 5-year DFS for patients who achieved MRD negativity was 63% (95% CrI, 55%-70%), compared with 16% (95% CrI, 11%-22%) for those who were MRD-positive. The average HR for achieving MRD negativity was 0.38 (95% CrI, 0.32-0.45) for OS and 0.33 (95% CrI, 0.26-0.44) for DFS. The difference in 5-year RMST of the MRD-negative and MRD-positive groups was 1.36 years (95% Crl, 1.13-1.62) for OS and 1.98 years (95% Crl, 1.70-2.28) for DFS.

Figure 2 -. Estimated survival curves according to MRD response for the entire study population and for only studies reporting patients in CR.

(A) Overall survival for the entire study population (N=35 studies, n=6004 patients), (B) disease-free survival for the entire study population (N=39 studies, n=4663 patients), (C) overall survival for studies only in patients in CR (N=21 studies, n=3427 patients), and (D) disease-free survival for studies only in patients in CR (N=23 studies, n=2939 patients). The curves show the posterior medians of survival probabilities. The shadings of each curve show the 95% Bayesian credible intervals for the survival rates at the corresponding point in time of follow-up.

Association of MRD status on survival outcomes in the CR population

Among studies reporting outcomes of only patients who had achieved CR, the impact of MRD showed a pattern similar to that of the full dataset (Figure 2C-D). The estimated 5-year OS for the MRD-negative group was 67% (95% CrI, 53%-77%), compared with 31% (95% CrI, 18%-44%) for the MRD-positive group. For estimated 5-year DFS for the MRD-negative group was 63% (95% CrI, 53%-72%), compared with 17% (95% CrI, 10%-24%) for the MRD-positive group. In studies of only patients in CR, the average HR for achieving MRD negativity was 0.38 (95% CrI, 0.27-0.51) for OS and 0.34 (95% CrI, 0.25-0.48) for DFS. The difference in 5-year RMST for OS was 1.35 years (95% CrI, 0.97-1.78) and 1.93 years (95% CrI, 1.57-2.30) for DFS.

Association of MRD status on survival outcomes by MRD threshold

The survival curves for MRD-negative and MRD-positive groups by MRD thresholds in the full dataset are shown in Figure 3. MRD negativity was consistently associated with superior OS and DFS, regardless of the MRD threshold used. The estimated 5-year OS for the MRD-negative and MRD-positive groups from studies using an MRD threshold <0.1% was 70% (95% CrI, 59%-80%) and 30% (95% CrI, 20%-40%), respectively; using an MRD threshold of 0.1% 5-year OS was 57% (95% CrI, 38%-72%) and 20% (95% CrI, 9%-34%), respectively; and using an MRD threshold >0.1% 5-year OS was 73% (95% CrI, 34%-91%) and 48% (95% CrI, 9%-80%), respectively. The estimated 5-year DFS for the MRD-negative and MRD-positive groups from studies using an MRD threshold <0.1% was 69% (95% CrI, 57%-80%) and 14% (95% CrI, 8%-23%), respectively; using an MRD threshold of 0.1% 5-year DFS was 62% (95% CrI, 43%-76%) and 14% (95% CrI, 4%-30%), respectively; and using an MRD threshold >0.1% 5-year DFS was 56% (95% CrI, 36%-73%) and 19% (95% CrI, 7%-37%), respectively. Notably, studies using a lower MRD threshold (i.e. <0.1%) appeared to provide better discrimination for survival outcomes than did studies using higher thresholds.

Figure 3 – Estimated survival curves according to MRD response for the entire study population, stratified by MRD threshold.

(A) Overall survival using threshold <0.1% (N=14 studies, n=2723 patients), (B) disease-free survival using threshold <0.1% (N=14 studies, n=1265 patients), (C) overall survival using threshold = 0.1% (N=11 studies, n=1139 patients), (D) disease-free survival using threshold = 0.1% (N=13 studies, n=1238 patients), (E) OS using threshold >0.1% (N=10 studies, n=2205 patients), and (F) DFS using threshold >0.1% (N=14 patients, n=2288 studies). The curves show the posterior medians of survival probabilities. The shadings of each curve show the 95% Bayesian credible intervals for the survival rates at the corresponding point in time of follow-up.

Association of MRD status on survival outcomes by CR status and MRD threshold

To further understand the prognostic impact of MRD negativity in patients who achieved CR, we repeated these analyses using only those studies that exclusively reported MRD outcomes in patients in CR. The HRs for the CR-only population, stratified by the MRD threshold used, are shown in Figure 4. Among patients in CR, MRD negativity was associated with significant clinical benefit across reported MRD thresholds, although the favorable impact of MRD negativity was numerically higher in studies using an MRD threshold <0.1%, where the HR was 0.25 (95% CrI, 0.11-0.47) for OS and 0.15 (95% CrI, 0.06-0.30) for DFS.

Figure 4 –

Hazard ratios (HRs) by MRD threshold for studies that only included patients in CR. Each square shows the posterior median of HR and the horizontal bands demonstrate the corresponding 95% credible intervals (CrI).

Association of MRD status on survival outcomes by other subgroups and their interactions with CR status and MRD threshold

The HRs for studies in the full dataset, stratified by MRD threshold, MRD detection method (i.e. multiparameter flow cytometry, polymerase chain reaction, or other methods), and MRD detection times (i.e. after induction, during consolidation, or after consolidation) are shown in Figure 5. All the subgroupings suggested a beneficial effect of achieving MRD negativity, regardless of MRD detection method or time of MRD assessment. Furthermore, within these subgroups, achievement of MRD negativity was associated with superior OS and DFS, regardless of the MRD threshold used. The differences of 5-year survival rates and RMSTs for each subgroup are summarized in Supplemental Table 2.

Figure 5 – Hazard ratios (HRs) by subgroups for the entire study population, stratified by MRD threshold.

(A) Overall survival and (B) disease-free survival. Each square shows the posterior median of HR and the horizontal bands demonstrate the corresponding 95% credible intervals (CrI). Abbreviations: MFC, multiparameter flow cytometry; PCR, polymerase chain reaction.

MRD negativity was associated with superior OS and DFS across subgroups, both in studies that included only patients in CR and in studies that included patients with mixed hematologic responses (i.e. studies including patients in CRi and/or MLFS) (Figure 6). The differences of 5-year survival rates and RMSTs for each subgroup are summarized in Supplemental Table 3.

Figure 6 – Hazard ratios (HRs) by subgroups for the entire study population, stratified by inclusion of only patients in CR versus inclusion of patients with lesser responses.

(A) Overall survival and (B) disease-free survival. Each square shows the posterior median of HR and the horizontal bands demonstrate the corresponding 95% credible intervals (CrI). Abbreviations: MFC, multiparameter flow cytometry; PCR, polymerase chain reaction. “Mixed response” refers to studies including patients in CRi and/or MLFS in the MRD analyses.

Discussion

MRD has emerged as a powerful prognostic marker in a wide spectrum of leukemias, and routine assessment of MRD has been incorporated into several consensus recommendations.^10,60-62 However, despite the well-established impact of MRD on survival across dozens of studies⁷, it is not yet considered a valid endpoint for AML clinical trials. This, in part, has been based on uncertainties of the optimal MRD cutoff for risk stratification and of the magnitude by which MRD influences relapse and survival outcomes specifically in those who have achieved CR.⁹ In this meta-analysis, we focused on these associations and observed a consistently favorable and statistically significant association of MRD negativity on OS and DFS. Our findings strongly suggest that achievement of MRD negativity is an important therapeutic goal across clinical contexts and confirm that detection of MRD is associated with very poor outcomes.

There is significant heterogeneity in how MRD has been measured in published AML studies, including differences in type of MRD assay, timing of MRD assessment, and MRD threshold to determine “MRD negativity”, as well as differences in the hematologic response of patients at the time of MRD assessment (e.g. at the time of CR, CRi or MLFS). All these variables could impact both the relative and absolute impact of persistent MRD on clinical outcomes and therefore should be controlled for in order to accurately quantify this association. The studies included in this meta-analysis used a variety of MRD assessment methodologies with variable sensitivities; this variation in analytic sensitivity was driven both by inherent differences in sensitivity across assay types and improvements in sensitivity of some assays over time, particularly in the case of multiparameter flow cytometry.¹⁰ While MRD response was associated with DFS and OS regardless of MRD threshold, the favorable impact of achieving MRD negativity appeared greatest when lower MRD thresholds were used. Importantly, we also observed that these associations were largely maintained regardless of the specific MRD assay used or the time of MRD assessment.

As CR is considered a more optimal clinical endpoint of AML therapy than either CRi or MLFS based on several previous analyses suggesting a survival benefit in those achieving CR,^3-5 it was particularly important to establish the impact of MRD in the context of CR with full hematologic recovery. We observed a remarkably similar impact of MRD response with respect to both OS and DFS in the global study population (in which 40% of studies included patients with mixed hematologic responses) and in those studies that only reported outcomes in patients in CR. The similar results from these two analyses is likely because, even among studies that evaluated MRD in patients with mixed hematologic responses, patients with CRi or MLFS represented only a minority of the total study population. Among the CR-only population, a trend towards greater benefit with an MRD-negative response was again observed in the group with MRD threshold <0.1% (HR for OS: 0.25 [95% CrI, 0.11-0.47]; HR for DFS: 0.15 [95% CrI, 0.06-0.30]), as compared with MRD thresholds ≥0.1%, which is suggestive of better discrimination of outcomes with more sensitive assays for MRD detection.

The robust association of MRD on relapse and survival outcomes supports consideration of MRD response as a surrogate endpoint for DFS and OS. A recent analysis by the Food and Drug Administration (FDA) of patient-level data from 8 randomized clinical trials, showed a modest association between CR rate and OS and a strong association between event-free survival (EFS) and OS,⁶³ reinforcing a previous precedent from the agency that accepted improvement in EFS as a valid therapeutic endpoint in AML. Acknowledgment of MRD negativity—or more specifically “CR with MRD negativity”—as a surrogate endpoint in AML would have a distinct advantage over EFS because an MRD-based response endpoint can be measured relatively early in the disease course and therefore may allow for more rapid assessment of the efficacy of investigational drugs and regimens.⁸ Furthermore, the consistently poor outcomes of patients with detectable MRD highlights both the unmet need of this clinical scenario and the importance of MRD-directed clinical trials in AML. Regardless of hematologic response or MRD threshold used, we observed that persistent detection of MRD after AML-directed therapy was associated with an estimated 5-year OS that was generally <35% and a 5-year DFS <20%. Given the universally poor outcomes of patients with MRD-positive AML, a novel therapeutic that results in a clinically meaningful rate of MRD clearance and translates into superior OS among responders may have a regulatory path to approval as MRD-targeted therapy in this disease.

The FDA has offered specific guidance for the consideration of MRD as a surrogate endpoint in AML.⁶⁴ Some of these recommendations include assessment of MRD specifically in patients who achieve CR with full hematologic recovery, analysis of the impact of MRD according to specific MRD cutoffs, and the use of patient-level data. Ideally, surrogacy would be assessed using patient-level data from multiple randomized trials showing that survival outcomes are strongly correlated with MRD response, independent of the treatment received. Unfortunately, to our knowledge, no analysis of this association has yet been published from a large, randomized trial in AML. While our meta-analysis does not include patient-level data per se, we used a novel method to reversely reconstruct survival data from published Kaplan-Meier curves, allowing for more accurate statistical modeling and survival estimates without access to patient-level data and contributing to the confidence of our survival estimates. Acknowledging that some studies were inconsistent in how they defined DFS and OS—with some survival analyses starting at the time of diagnosis and others starting at the time of MRD assessment—this heterogeneity was unlikely to substantially alter our survival estimates.

Using this novel statistical method, we were able to quantify the magnitude of benefit of MRD negativity in patients with AML who achieved CR; however, it should be noted that most studies did not report survival outcomes specifically in patients with lesser responses (e.g. CRi or MLFS) and therefore we cannot estimate the impact of MRD for patients with these individual responses. Most studies also did not provide survival data according to specific levels of MRD but rather classified patients as “MRD-negative” or “MRD-positive” based on the analytical sensitivity of the assay or a pre-specified MRD threshold. We therefore cannot reconstruct survival estimates for specific MRD quantities, as was reported for the phase 2 trial that led to the approval of blinatumomab for MRD-positive B-cell ALL with MRD ≥0.1%.^65,66 While our data do not provide a specific quantity of MRD that warrants therapeutic intervention, the consistently negative impact of persistent MRD across clinical contexts, coupled with our finding that MRD is strongly associated with survival regardless of the MRD threshold used, support a conclusion that any level of MRD following AML-directed therapy is likely to be associated with worse outcomes. In fact, the greatest magnitude of benefit observed with achievement of MRD negativity was observed in studies which used an MRD threshold <0.1%, suggesting that more sensitive assays that can detect smaller quantities of MRD may have greater clinical utility than less sensitive assays. While it is challenging to compare sensitivity thresholds and MRD quantification across different assays—particularly PCR versus other MRD assays—the consistently favorable impact of MRD negativity across all analyses regardless of the MRD threshold used provides strong evidence that an MRD-negative remission should be the goal AML-directed therapy, regardless of the MRD assay that is used or its sensitivity.

While our meta-analysis confirms that MRD negativity is consistently associated with superior outcomes, it is important to note that achievement of MRD negativity with currently available assays is not tantamount to cure, nor are patients with MRD-positive disease destined to relapse. Despite recent improvements in AML MRD assays, most tests can achieve a sensitivity of 1x10⁻⁴ to 1x10⁻⁵ at best, and therefore absence of detectable MRD does not rule out the presence of residual disease that can eventually lead to overt relapse. Notably, in our analysis, even using a threshold of <0.1%, the 5-year DFS was 63% in the entire cohort (and 70% for patients in CR), highlighting that a substantial proportion of these MRD-negative patients still relapse and die from AML. Conversely, a minority of MRD-positive patients are still apparently cured, as evidenced by a 5-year DFS rate of 16%. It is reasonable to speculate that these patients with long-term DFS eventually converted to MRD negativity—either with additional cycles of chemotherapy, change of chemotherapy regimen, or by undergoing allogeneic HSCT.

In conclusion, we have shown that MRD is strongly associated with both OS and DFS, with similar benefits observed in patients achieving MRD negativity regardless of the MRD threshold used or when the analysis was restricted only to patients in CR with full hematologic recovery. In patients who achieved CR, MRD negativity was associated with a 3-fold reduction in the risk of death, and the magnitude of benefit associated with an MRD-negative response appeared greatest when a lower MRD threshold (i.e. <0.1%) was used. These associations between MRD and survival outcomes were confirmed across MRD detection methods and times of MRD assessment, highlighting the applicability of MRD monitoring across clinical contexts. Taken together, the robustness of the association between MRD and survival in AML provides further support for the potential use of MRD as a surrogate endpoint in AML that may be considered for regulatory decision-making.

Data-sharing statement

Data may be shared upon reasonable request to the corresponding author.