Minimal Residual Disease Data in Hematologic Malignancy Drug Applications and Labeling: An FDA Perspective

Abstract

Minimal residual disease (MRD) is increasingly used as a prognostic biomarker, a measure of clinical efficacy, and a guide for treatment decisions in various hematologic malignancies. We sought to characterize MRD data in registrational trials in hematologic malignancies submitted to the U.S. Food and Drug Administration (FDA) with the ultimate goal of expanding the utility of MRD data in future drug applications. We descriptively analyzed MRD data collected in registrational trials including the type of MRD endpoint, assay, disease compartment(s) assessed, and the acceptance of MRD data in the U.S. prescribing information (USPI). Of 196 drug applications submitted between January 2014 and February 2021, 55 (28%) included MRD data. Of the 55 applications, MRD data was proposed by the Applicant for inclusion in the USPI in 41 (75%) applications but was only included in 24 (59%). Despite an increasing number of applications that proposed to include MRD data in the USPI, the acceptance rate decreased over time. Although MRD data have the potential to expedite drug development, our analysis identified challenges and specific areas for improvement, including assay validation, standardization of collection methods to optimize performance, and considerations in trial design and statistical methodology.

Introduction

Multiple studies have reported the prognostic value of minimal residual disease (MRD) status in a range of hematologic malignancies, including acute myeloid leukemia (AML), acute lymphoblastic leukemia (ALL), chronic myelogenous leukemia (CML), chronic lymphocytic leukemia (CLL), and multiple myeloma (MM) (1–4). The achievement of MRD negativity has been associated with depth of clinical response and prolongation of progression-free and overall survival (1–4). There are multiple potential clinical uses for MRD data. MRD data can be used to assess response to therapy or assess likelihood of relapse. For regulatory purposes in a clinical trial, MRD data can be used to inform patient selection, as a stratification factor, for efficacy response assessment, and potentially to guide treatment escalation or de-escalation (5,6). The criteria used to support MRD assessment in clinical research and clinical care differ from the criteria needed to support regulatory decisions (6). Clinical trials in hematologic malignancies are increasingly incorporating MRD status as a prognostic biomarker to identify patients at risk of poor outcomes and as an efficacy endpoint to assess response, including clinical trials intended to support regulatory approval. Despite this, MRD data collected in registrational trials are often not fit for inclusion in the U.S. prescribing information (USPI).

In registrational clinical trials that evaluate MRD data, important considerations include the clinical validity of its use, the validity of the MRD assay, performance of the assay in disease compartments (bone marrow vs. peripheral blood), the timing and consistency of MRD monitoring, and the completeness of the data collection, as well as the adequacy of the statistical analysis plan (6). Herein, we present an analysis of the MRD data in hematologic malignancies clinical trials submitted for regulatory review as it relates to regulatory decisions made by the U.S. Food and Drug Administration (FDA), including the observed challenges. We evaluate trends in the use of MRD data in registrational trials, proposals for inclusion of MRD data in the USPI, and inclusion of those data in final labeling, with the aim of improving future submissions by increasing awareness of the deficiencies observed to date and ways to address these challenges. While there is considerable interest in the development of MRD as a surrogate endpoint, this topic is addressed in the FDA Guidance on the use of MRD data in development of drug and biological products for treatment of hematologic malignancies and is beyond the scope of this article (6).

Methods

We queried the FDA internal database and identified all new drug applications (NDAs) and biological licensing applications (BLAs) submitted between January 2014 to February 2021 to support approval of therapies for patients with hematologic malignancies. We reviewed clinical study reports, relevant datasets, FDA reviews, and the proposed and approved versions of the USPI for inclusion of MRD data. Reviewers extracted the following information regarding the applications and the associated trials with MRD data: application submission date, trial design, disease, and disease setting. MRD-specific information extracted included information on the type of MRD endpoint (primary, secondary, or exploratory), whether treatment decisions were based on MRD results (yes/no), type of MRD assay (NGS, flow, etc.), assay sensitivity, threshold for MRD-negativity, compartment(s) assessed (bone marrow and/or peripheral blood), and study population evaluated. If MRD data was included in the Applicant’s proposed label, additional details were captured to determine if the information included in the final USPI differed from what was initially proposed. If the application included MRD assessments and/or proposed inclusion of MRD data in the USPI, but MRD data was ultimately not included in the USPI, the reasons for not including MRD data in the USPI were captured from FDA review documents. Additional exploratory analysis by time, 1/2014–6/2017 (period 1) and 7/2017–2/2021 (period 2), was conducted to evaluate potential trends in MRD data. All analyses were descriptive.

Results

Of 196 NDAs or BLAs involving hematologic malignancies submitted between January 2014 to February 2021, 55 (28%) included MRD data. Twenty (20) applications with MRD data were submitted in period 1, and 35 applications with MRD data were submitted in period 2. The applications included indications in MM (31%), CLL (22%), ALL (18%), CML (16%), AML (13%), and hairy cell leukemia (2%), in the frontline setting (42%), relapsed/refractory setting (44%), both (9%), or other setting (5%), such as maintenance. The 55 applications included data from a total of 57 clinical trials (8 applications included MRD data from more than one trial and 6 trials served as the basis for more than one application), of which, 53% were randomized controlled trials and 47% were single arm trials. Eleven percent (11%) of trials included MRD status evaluation as a primary endpoint, 65% as a secondary endpoint, and 32% as an exploratory endpoint. MRD data results were used to guide treatment decisions in 9% of trials.

Of the 55 applications, MRD data was proposed by the Applicant for inclusion in the USPI in 41 (75%) applications. Of those 41 applications, the MRD data was ultimately deemed adequate for inclusion in the USPI by FDA in 24 (59%) applications (Table 1) (7). The acceptance rate for inclusion of MRD data was 80% (12/15 applications) in period 1 and 46% (12/26 applications) in period 2.

Table 1:

Products for Hematologic Malignancies with MRD Data in U.S. Prescribing Information

Product	Disease	Disease Setting	Assay(s) in USPI	MRD Endpoint Status	Clinical Threshold in USPI	Compartment in USPI
Ponatinib	ALL and CML	Relapsed/Refractory	PCR	Secondary	10−3	PB
	CML	Relapsed/Refractory	PCR	Secondary	10−3	PB
Nilotinib	CML	Frontline	PCR	Primary	10−3	PB
			PCR	Secondary	10−3	PB
			PCR	Exploratory	10−3	PB
			PCR	Primary	10−3	PB
		Relapsed/Refractory	PCR	Primary	10−3	PB
			PCR	Secondary	10−3	PB
		Both	PCR	Primary	10−3	PB
Bosutinib	CML	Frontline	PCR	Primary	10−3	PB
			PCR	Exploratory	10−3	PB
Imatinib	CML	Frontline	PCR	Secondary	10−3	PB
			PCR	Primary	10−3	PB
Dasatinib	CML	Frontline	PCR	Secondary	10−3	PB
		Relapsed/Refractory	PCR	Secondary	10−3	PB
			PCR	Secondary	10−3	PB
		Both	PCR	Secondary	10−3	PB
Venetoclax	CLL	Relapsed/Refractory	FC	Exploratory	10−4	Both
			PCR	Secondary	10−4	PB
		Frontline	NGS	Secondary	10−4	Both
Obinutuzumab	CLL	Frontline	PCR	Secondary	10−4	Both
			PCR	Secondary	10−4	Both
Blinatumomab	ALL	Relapsed/Refractory	PCR	Exploratory	10−4	BM
			FC and PCR	Exploratory	10−4	BM
			FC and PCR	Secondary	10−4	BM
			PCR	Secondary	10−4	BM
		Other (MRD+ CR1 or CR2)	PCR	Primary	10−4 and 10−5	BM
Inotuzumab ozogamicin	ALL	Relapsed/Refractory	FC	Secondary	10−4	BM
Tisagenlecleucel	ALL	Relapsed/Refractory	FC	Secondary	10−3	BM
Daratumumab	MM	Frontline	NGS	Secondary	10-5	BM
			NGS	Secondary	10−5	BM
		Relapsed/Refractory	NGS	Secondary	10−5	BM
Daratumumab and hyaluronidase	MM	Relapsed/Refractory	NGS	Secondary	10−5	BM
Idecabtagene vicleucel	MM	Relapsed/Refractory	NGS	Secondary	10−5	BM

Abbreviations: ALL=acute lymphoblastic leukemia, CML=chronic myeloid leukemia, CLL=chronic lymphocytic leukemia, MM=multiple myeloma, PCR=polymerase chain reaction, FC=flow cytometry, NGS=next generation sequencing, PB=peripheral blood, BM=bone marrow, MRD=minimal residual disease, CR1=complete remission 1, CR2=complete remission 2.

Source: U.S. prescribing information for respective products (7).

In total, 26/42 (62%) trials that had MRD data proposed for inclusion in the USPI had the data included in the USPI. The acceptance pattern for inclusion of MRD data based on disease type was evaluated. MRD data was deemed adequate for inclusion in the USPI for 42% of the trials in MM, 100% of the trials in CML, 70% of the trials in ALL, 67% of the trials in CLL, and none of the trials in AML or HCL that had MRD data proposed for inclusion in the USPI (Figure 1).

Figure 1:

MRD Data Acceptance in U.S. Prescribing Information by Disease Type. The figure represents disease indications in the trials for which MRD data was proposed versus ultimately accepted for inclusion in the USPI. One application included indications in both ALL and CML.

Abbreviations: MM=multiple myeloma, CML=chronic myelogenous leukemia, ALL=acute lymphoblastic leukemia, CLL=chronic lymphocytic leukemia, AML=acute myeloid leukemia, HCL=hairy cell leukemia.

Source: FDA analysis of pooled data from respective clinical trials.

The clinical threshold for MRD-negativity ranged from 10⁻³ to 10⁻⁶ in the trials and 10⁻³ to 10⁻⁵ for MRD data included in the USPI. Among the 26 trials with MRD data in the USPI, MRD status was evaluated regardless of clinical response in 38% of the trials, and in patients achieving a specific clinical response (e.g., partial response or better, or complete response or better) in 62% of the trials.

MRD assays used in the trials included polymerase chain reaction (PCR) in 40%, next-generation sequencing (NGS) in 26%, flow cytometry (FC) in 51%, and immunohistochemistry (IHC) in 2% of trials. Sixteen percent (16%) of trials utilized more than one type of assay to assess MRD status. Among the 26 trials that had MRD data included in the USPI, the data was based on PCR in 58%, NGS in 19%, and FC in 12% of trials; 2 trials had MRD data based on both PCR and FC, and 1 trial had MRD data based on both NGS and PCR. The acceptance pattern for inclusion of MRD data based on assay type was also evaluated. Eighty-five percent (85%) of the trials with MRD data based on PCR had the data include in the USPI, 50% of trials with MRD data based on NGS had the data included in the USPI, and 33% of the trials with MRD data based on FC had the data included in the USPI; one trial with MRD data based on IHC was not accepted for inclusion in the USPI (Figure 2).

Figure 2:

MRD Data Acceptance in U.S. Prescribing Information by Assay Type. The horizontal axis represents methods of MRD assessment in the trials for which MRD data was proposed versus ultimately accepted for inclusion in the USPI.

Abbreviations: PCR=polymerase chain reaction, NGS=next generation sequencing, FC=flow cytometry, IHC=immunohistochemistry.

Source: FDA analysis of pooled data from respective clinical trials.

Among the 17 applications with MRD data proposed but ultimately not included in the USPI, the leading reason(s) for exclusion were analytical and/or test validation issues in 65%, assay performance issues in 24%, issues with the trial design and/or MRD data collection in 12%, and one application was withdrawn due to reasons not related to MRD data.

Discussion

There has been significant interest in the use of MRD in drug development. Reflecting this interest, over a quarter of new drug and biologic licensing applications involving hematologic malignancies, including 57 registrational trials, submitted to the FDA between 2014–2021 contained MRD data. Of these, the majority were in MM, CLL, and ALL, followed by CML and AML.

MRD status has been used successfully to define a population with B-ALL with poor outcomes who might benefit from additional therapy and to determine when continuous treatment may be discontinued in patients with CML. Blinatumomab is approved for the treatment of patients with B-ALL in complete remission with MRD ≥ 0.1%. To assess the acceptability of this cutoff to define a population at high risk for relapse, FDA performed a log-group analysis of hematologic relapse-free survival using patient-level historical data which showed that patients in first remission with MRD ≥ 0.1% after 3 blocks of intensive chemotherapy had a uniformly poor prognosis and could benefit from preemptive treatment (8). In the case of nilotinib, which is approved for the treatment of CML, MRD status has been used to evaluate the potential for treatment discontinuation. Data from two prospective multicenter single-arm trials in patients with CML were submitted to support molecular response (MR) 4.5 (corresponding to BCR-ABL ≤ 0.0032% IS) after a minimum of 3 years of treatment with nilotinib as a cutoff which defined a population in whom treatment discontinuation could be considered. In these studies, about half of patients were able to maintain MR4.5 for 96 weeks post-discontinuation, and of those patients who restarted nilotinib due to loss of major molecular response (MMR), more than 90% were able to regain MMR (9).

Use of MRD data in other disease settings has proved more challenging. One of the challenges is that most hematologic malignancies are multicompartmental diseases and it may be unclear which compartment is optimal to assess MRD status. CLL, for example, may involve the marrow, blood, lymph nodes, liver, and spleen. During and following treatment, one or more sites may serve as a reservoir for residual disease. In addition, therapeutic modalities, especially targeted therapies, may differentially affect disease in the peripheral blood (PB) versus bone marrow (BM). For example, with a threshold of <10⁻⁴ for MRD-negativity, concordance between PB and BM MRD status is approximately 85% in patients with CLL treated with chemoimmunotherapy (10–12). In the CLL14 trial evaluating the combination of venetoclax plus obinutuzumab in untreated CLL, 134 patients in the venetoclax arm who were MRD negative in PB had matched BM specimens; of these, 122 (91%) were negative in both PB and BM (13). Despite a relatively high level of concordance between PB and BM, the multicompartmental nature of the disease warrants a multi-compartmental approach when evaluating MRD status. This also has further implications because there is a need for appropriate analytical data to support use of the MRD assay for each of the selected compartments. Ultimately, a multi-compartmental approach in patients with hematologic malignancies can help inform the utility of MRD evaluation, MRD kinetics, and the impact on longer-term outcomes.

While the above examples highlight a few of the successes and some of the limitations observed with the MRD data in registrational clinical trials in hematologic malignancies, our analysis highlights that overall, substantial challenges remain. Characterization of regulatory actions showed that despite the increasing number of submissions proposing MRD data for inclusion in the USPI over time, the rates of inclusion of MRD data in the USPI did not reflect this increase.

The leading reasons for excluding MRD data from the USPI were analytical and test validation deficiencies (e.g., incomplete test characteristics data, lack of test validation overall or in that disease, lack of a validated threshold for MRD-negativity), followed by performance issues (e.g., high amount of test failure, inability to identify a clone), and issues with trial conduct or design (e.g., inadequate data collection, statistical issues).

Our analysis suggested several notable patterns regarding acceptance of MRD data based on disease setting and assay type. The acceptance rate for inclusion of MRD data from registrational trials in CML was 100%. In part, this is likely a reflection that the analysis of MRD status is measured based on PCR detection of the BCR-ABL transcript, a defining molecular feature present in patients with CML and has been established as a standard of care in this disease setting. MRD data from PCR-based assays had the highest acceptance rate (85%) and the acceptance rate for data from PCR-based assays remained high (73% acceptance rate among trials in ALL and CLL) when trials in CML were excluded from the analysis. In contrast, the acceptance patterns in other disease settings and for other assay types were much lower. This may reflect both the more limited data available and additional complexities of MRD analysis in these disease settings. Molecular heterogeneity and lack of uniform defining molecular features in many hematologic malignancies poses a significant challenge.

Among the disease settings evaluated, MM had one of the lowest acceptance rates for MRD data despite having the highest number of submissions that included MRD data proposed for inclusion in the USPI. Because most of the MM applications that included MRD data were submitted in period 2, this partially accounts for the decline in acceptance rate from period 1 to period 2. The most common reason for exclusion of MRD data in this disease setting was due to assay performance issues, including failure to detect a baseline clone with the use of NGS-based assays. A challenge with the use of NGS-based MRD assays that require identification of a diagnostic clone that can be tracked over time to monitor changes in disease burden, is that calibration failure (i.e., no diagnostic clone suitable for tracking is detected) results in missing data and limits interpretability of the overall results. The issue may be further compounded when there is an imbalance in calibration rates between study arms. Reasons for calibration failure may include low disease burden, hemodilution, and inadequate DNA input. This suggests that sample collection could be improved and standardized to ensure more consistent assay performance and robust data collection.

Among the assay types evaluated, FC had the lowest acceptance rate for MRD data. The primary reason for excluding MRD data assessed by FC was insufficient analytical validation of the assay. Additional reasons included lack of a validated threshold to define MRD-negativity in a particular disease setting, inconsistent sampling of and/or discordant results between different disease compartments and insufficient/missing data. A major limitation of FC-based platforms is the lack of analytical validation, variability, and lack of standardization (e.g., differences in antibody panels, instruments, operators, and data analysis). This suggests that further efforts directed at analytical and clinical validation and standardization, would address some of the major challenges associated with the use of FC-based MRD data.

In summary, while there have been some successes, our analysis showed that a substantial proportion of the MRD data that was collected was not sufficiently robust to support inclusion in product labeling. The decrease in acceptance over time is unexpected considering the initial publication of the draft FDA guidance on the use of MRD data in development of drug and biologic products for treatment of hematologic malignancies in 2018 and the interest among stakeholders in advancing the use of MRD in clinical trials. The MRD guidance addresses many of the key issues, including technical considerations (e.g., type of platform, assay validation, sample collection), disease-specific considerations, and regulatory considerations (e.g., use of MRD as a biomarker, as an endpoint, for patient selection/enrichment). FDA is agnostic to which technology platform (i.e., molecular vs. cellular-based assays) is used to assess MRD in clinical trials, provided that the assay has adequate analytical validation. To increase the quality and utility of MRD data collected in registrational trials, FDA encourages sponsors to meet with us early in their development programs and ensure that (1) the MRD-assay(s) they propose to use have adequate analytic validation, (2) the details of the planned assay(s) and their specific uses are clearly specified, (3) the proposed thresholds are appropriate and validated in the intended disease setting/population, and (4) the sample collection, sample stability, banking and processing conditions are optimized. FDA recommends that sponsors consider submitting a request to the Center for Devices and Radiological Health (CDRH) for pre-submission feedback on these aspects of the assay prior to proposing its use in a clinical trial (14). All aspects of the proposed MRD assay and its intended use should be addressed prior to initiating a clinical trial with registrational intent. Furthermore, FDA encourages sponsors to remain engaged after trial initiation to troubleshoot problems that may arise.

Future Directions

MRD has limitless potential as a clinical tool and has the potential to expedite drug development. When used as a surrogate or intermediate clinical endpoint, it may allow for an earlier clinical trial readout. This may be especially useful in disease areas where randomized trials with overall survival or progression-free survival endpoints are less feasible. In addition to endpoint evaluation, MRD data may expedite drug development by permitting a more tailored approach to treatment. If MRD status is used as a selection tool to identify patients at higher risk of poor outcome, this may allow for selection of an appropriately high-risk population that may benefit from more intensive or prolonged therapy.

However, each of these uses of MRD requires substantial confidence and reliability in the assay results and its performance characteristics. Our analysis highlights that there remain considerable areas for improvement in the clinical trial use of MRD status. This reflects valuable information that was not able to be incorporated in the labels of these products. Moving forward, improvements in assay validation, rigorous collection of MRD data, and appropriate statistical planning are needed to enable greater reliance on MRD data in registrational clinical trials.