Relapse Despite MRD Negativity in Multiple Myeloma: Toward Integrated Risk Prediction

Minimal residual disease (MRD) negativity has emerged as one of the most powerful prognostic biomarkers in multiple myeloma (MM), consistently associated with prolonged progression-free survival (PFS) and overall survival (OS). The clinical significance of MRD negativity was first systematically demonstrated in a landmark meta-analysis of 21 studies (15 included in the quantitative analysis), comprising patients receiving conventional chemotherapy, novel agents, and autologous stem cell transplantation (ASCT). MRD was primarily assessed using multiparameter flow cytometry or PCR-based methods. MRD-negative patients experienced a substantial survival advantage, with a 69% reduction in the risk of progression and a 59% reduction in the risk of death compared with MRD-positive patients 1. Importantly, the prognostic value of MRD negativity was consistent across treatment modalities, establishing it as a unifying marker of deep therapeutic response. At the same time, it has become clear that MRD negativity, although strongly prognostic, is not synonymous with cure, as a clinically relevant proportion of patients who achieve MRD negativity ultimately relapse. A subsequent meta-analysis incorporating 44 datasets and more than 8000 patients treated with contemporary regimens—including proteasome inhibitors, immunomodulatory drugs, monoclonal antibodies, and combination therapies—confirmed and expanded these findings 2. Using next-generation sequencing (NGS) at 10−5 sensitivity across multiple time points, patients maintaining MRD negativity for 6–12 months demonstrated significantly prolonged PFS compared with those who reverted to MRD positivity, with more than a 60% reduction in progression risk. Sustained MRD negativity correlated with deeper molecular remission and more profound M-protein suppression. Across modern regimens, including combinations with immunomodulatory drugs, proteasome inhibitors, monoclonal antibodies, and high-dose chemotherapy with ASCT, progressively higher rates and longer durability of MRD negativity have been achieved, underscoring the central role of both intensive cytoreduction and immune-mediated mechanisms in attaining deep responses 2, 3. Despite these advances, relapse after achieving MRD negativity is increasingly recognized in clinical practice. Rather than representing a true “paradox,” this observation of relapse despite MRD negativity reflects the interplay of biological complexity, spatial heterogeneity, and technical limitations of MRD assessment. Understanding why some patients relapse despite apparently optimal depth of response is therefore the central rationale for moving beyond a purely bone-marrow–centric, single-modality definition of remission toward integrated relapse prediction. Relapse after MRD negativity reflects the heterogeneous and evolving biology of MM. The disease arises from a post–germinal center plasma cell undergoing progressive genetic and epigenetic diversification, resulting in a complex subclonal architecture. sequencing studies frequently reveal branching patterns of clonal evolution in MM, but linear and more complex or mixed trajectories are also observed, and the impact of modern therapies on clonal dynamics remains to be fully elucidated 4, 5. Therapeutic pressure can selectively eliminate sensitive clones while resistant subpopulations—often enriched for alterations in TP53, DNA repair pathways, or drug targets—survive and expand. These resistant clones may remain below the detection threshold of even highly sensitive MRD assays, particularly when they are confined to spatially distinct niches not captured by standard posterior iliac crest sampling. In this context, MRD negativity often reflects the absence of detectable dominant clones in the sampled compartment, rather than true eradication of all biologically competent subclones. Single-time-point MRD assessments are limited by bone marrow sampling variability and spatial heterogeneity. Residual disease may remain undetected if localized in unsampled marrow regions, extramedullary sites, or below assay detection thresholds. Consequently, serial MRD monitoring has emerged as a strategy to refine prognostic precision. Landmark analyses from the MAIA and ALCYONE trials evaluated sustained MRD negativity in newly diagnosed patients treated with daratumumab-containing regimens 3. These findings underscore that the durability of MRD negativity provides prognostic information beyond single-time-point assessments and can guide individualized maintenance therapy and surveillance strategies. Similarly, real-world evidence supports the prognostic relevance of sustained MRD negativity beyond controlled clinical trials. In an observational cohort of 275 newly diagnosed multiple myeloma patients, longitudinal MRD monitoring revealed that patients who maintained MRD negativity for 12 to 24 months experienced markedly prolonged PFS and OS, independent of induction regimen, autologous stem cell transplant status, or baseline cytogenetic risk. Notably, the benefit of durable MRD negativity extended across both standard- and high-risk cytogenetic subgroups, suggesting that persistent eradication of measurable disease confers a protective effect even in biologically aggressive myeloma 6. In a single-arm, single-center phase 2 trial evaluating continuous lenalidomide maintenance, Diamond and colleagues performed serial MRD measurements and observed that reappearance of detectable disease by NGS or flow cytometry anticipated biochemical and clinical relapse by several months. Patients whose MRD status converted from negative to positive during maintenance therapy were at significantly higher risk of progression, emphasizing the value of repeated MRD measurements to capture clonal resurgence and evolving disease kinetics 7. Together, these data highlight two principles: the durability of MRD negativity is a stronger predictor of long-term outcomes than single-time-point assessment, and MRD is a dynamic biomarker, with conversions from negative to positive status serving as early warning signs for relapse. Serial MRD monitoring thus provides both prognostic information and a window for pre-emptive therapeutic intervention. Thus, assessing MRD at regular intervals (e.g., every 3–6 months during maintenance) illustrates how both the timing and frequency of MRD evaluation influence the ability to detect molecular relapse at an early stage. A concise overview of the main clinical and methodological characteristics of selected studies illustrating the prognostic role and limitations of MRD, PET/CT, and liquid biopsy is provided in Table 1. Even serial bone marrow assessments may fail to detect disease outside the sampled compartment. Functional imaging with 18F-FDG PET/CT provides systemic evaluation of metabolically active lesions, including focal and extramedullary disease. Zamagni et al. 8 demonstrated that residual PET positivity after autologous transplantation independently predicted shorter PFS and OS, even in patients who were MRD-negative in the marrow. Subsequent studies confirmed the independent prognostic value of post-transplant PET positivity 10. Standardized interpretation systems such as the Italian Myeloma Criteria for PET Use (IMPeTUs) allow structured evaluation of lesion number and metabolic intensity 11. These findings illustrate that functional imaging complements marrow MRD assessment by identifying biologically active disease that remains clinically silent and may underlie relapse despite MRD negativity. Liquid biopsy offers a minimally invasive, systemic approach to residual disease detection in MM. By analyzing circulating tumor DNA (ctDNA) and circulating myeloma cells (CMMCs), it samples multiple marrow regions and extramedullary sites, better capturing spatial heterogeneity than single-site aspirates 12. Several studies suggest that CMMCs can display phenotypic and biological features distinct from marrow-resident plasma cells, including altered differentiation state, adhesion, and homing properties, consistent with potentially more mobile and therapy-resistant subpopulations 12. ctDNA identifies tumor-specific genomic alterations, including subclones, and rising ctDNA levels, variant allele fractions, or detectable CMMCs frequently precede biochemical or clinical relapse. Thus, liquid biopsy may signal molecular relapse even in marrow MRD-negative patients. Conversely, persistent circulating disease despite marrow MRD negativity suggests residual clones outside the sampled niche, refining prognostic assessment 12. Emerging approaches—including cell-free RNA and epigenetic profiling (e.g., 5-hydroxymethylcytosine signatures)—may further clarify clonal evolution and resistance. Clinically, circulating biomarkers correlate with tumor burden indicators such as ISS/R-ISS, β2-microglobulin, LDH, and PET-CT findings, especially in focal or extramedullary disease 12. Despite promising data, studies of liquid biopsy in MM remain small and methodologically heterogeneous. The meta-analysis by Ye et al. 9 sought to clarify the clinical value of circulating cell-free DNA (cfDNA) and ctDNA by systematically reviewing the literature through July 2021. Seven studies involving 235 patients were included, encompassing different disease stages and treatment settings. Techniques varied (allele-specific PCR vs. next-generation sequencing), as did genomic targets (mainly immunoglobulin gene rearrangements), contributing to inter-study heterogeneity. For MRD detection, pooled analysis showed high specificity (91%) but only moderate sensitivity (58%), with a summary AUC of 0.95. While ctDNA positivity strongly predicted marrow MRD positivity, its limited sensitivity indicated that many MRD-positive cases would be missed if relying on plasma alone. Prognostically, elevated cfDNA levels were consistently associated with inferior outcomes. Higher cfDNA conferred nearly a five-fold increased risk of progression and approximately a three-fold higher risk of death across pooled datasets. These findings support the independent prognostic value of circulating nucleic acids, even if their diagnostic performance for MRD remains insufficient to replace marrow-based assessment. Overall, ctDNA/cfDNA and CMMC analyses should therefore be regarded as investigational, complementary tools in MM: their analytical performance, standardization, and clinical thresholds for MRD-level detection require further validation before routine use in MRD-guided decision-making. The bone marrow microenvironment (BMME) provides critical survival and protective signals for residual myeloma cells. Malignant plasma cells engage in reciprocal interactions with stromal cells, endothelial cells, osteoblasts, immune components, and extracellular matrix proteins 13. These interactions enable subsets of plasma cells to enter a slow-cycling or dormant state, rendering them intrinsically less sensitive to cytotoxic therapies and potentially invisible to conventional MRD assays, which are optimized to detect actively proliferating cells. A critical aspect of BMME-mediated protection is cytokine signaling, notably through interleukin-6 (IL-6) produced by stromal and myeloma cells. IL-6 activates downstream JAK/STAT3, PI3K/Akt, and MAPK pathways, promoting anti-apoptotic signaling, metabolic adaptation, and survival under therapeutic stress 14. Alongside cytokine-driven survival, chemokine gradients—particularly the CXCL12/CXCR4 axis—mediate adhesion and retention of myeloma cells within protective niches. CXCL12 produced by stromal cells binds CXCR4 on plasma cells, activating PI3K/Akt and ERK signaling, enhancing drug resistance, and reducing susceptibility to apoptosis 15, 16. Interactions with the extracellular matrix via integrins, such as VLA-4 (α4β1), further protect myeloma cells through cell adhesion–mediated drug resistance (CAM-DR). Engagement of integrins not only anchors plasma cells to the stroma but also triggers intracellular signaling that promotes survival and suppresses apoptosis, contributing to therapy resistance 17. Moreover, the microenvironment promotes quiescence and dormancy. Dormant plasma cells evade cytotoxic therapies and may escape MRD detection, yet retain the capacity to re-enter the cell cycle under favorable conditions 18. Specific osteoblastic and vascular niches may sustain these low-proliferative states 19. Clinically, these microenvironmental and dormancy mechanisms imply that standard marrow-based MRD assays can underestimate true residual disease, particularly when resistant cells are sequestered within protective stromal and vascular niches 13, 14, 18, 19. They also provide a rationale for therapeutic strategies targeting stromal interactions (e.g., CXCR4 inhibitors, agents interfering with integrin/VLA-4–mediated adhesion), angiogenesis, or niche-specific signaling pathways to mobilize and sensitize residual cells to treatment 13, 15-17, 19. Integrating this biological understanding into MRD interpretation may help identify patients in whom apparent MRD negativity masks a persistent risk of late relapse driven by dormant or niche-protected clones 4, 18, 19. MRD assays quantify tumor burden but do not assess immune competence. Even in deep remission, residual cells may evade immune surveillance through upregulation of immune checkpoints, expansion of regulatory T cells, and effector cell exhaustion. Functional impairment of cytotoxic T cells and natural killer cells correlates with relapse risk 20. Regulatory T-cell depletion enhances anti-myeloma immune responses in preclinical and early clinical studies 21. Moreover, longitudinal immunophenotyping of patients receiving lenalidomide maintenance has revealed distinct peripheral blood and marrow immune signatures that differentiate those who maintain MRD negativity from those who relapse 22. Thus, durable disease control requires both minimal tumor burden and functional immune surveillance. The ability of immune-modulating strategies—including immunomodulatory drugs, anti-CD38 monoclonal antibodies, and T-cell–redirecting therapies such as bispecific antibodies and CAR-T cells—to induce deep and sometimes sustained MRD negativity further underscores the importance of immune control, while the heterogeneity in durability of these responses highlights that restoration and maintenance of effective immune surveillance are crucial for preventing relapse. In conclusion, MRD negativity has transformed response assessment in MM and serves as a validated surrogate for improved survival (Figure 1A). However, the observation of relapse after MRD negativity underscores that residual disease reflects more than what is captured by current marrow-based assays of tumor burden. It arises from the interplay of subclonal evolution, spatial heterogeneity, microenvironmental protection, immune dysfunction, and limitations of current detection technologies (Figure 1B). An integrated strategy combining serial MRD monitoring, functional imaging, systemic liquid biopsy, and immune profiling offers the most comprehensive framework for identifying patients at risk of relapse (Figure 1C). Rather than a static endpoint, MRD negativity should be viewed as a dynamic biological state within a broader ecosystem of tumor–host interactions. Only through such a multidimensional assessment can we move toward truly individualized, pre-emptive therapeutic strategies aimed at sustained, durable disease control in MM. Future research priorities include prospective validation of multimodal risk-prediction algorithms that integrate serial MRD measurements with imaging, circulating biomarkers, genomic risk, and immune profiling; harmonization of assay methodologies and thresholds; and incorporation of MRD-adapted, biology-informed strategies into clinical trials. The ultimate goal is not only to achieve MRD negativity but to translate it into durable, treatment-free remission for a larger proportion of patients with MM. All authors contributed to the manuscript and were involved in revisions and proofreading. All authors approved the submitted version. The authors have nothing to report. The authors have nothing to report. The authors declare no conflicts of interest. Data sharing not applicable to this article as no datasets were generated or analysed during the current study.