Прогнозирующее значение динамического мониторинга состояния минимальных остаточных заболеваний у пациентов с множественной миеломой: исследование реального мира.

1. Introduction

Multiple myeloma (MM) is a hematologic malignancy characterized by the clonal accumulation of malignant plasma cells. It accounts for approximately 10% of all hematologic malignancies and is the second most common type of hematological malignancy. Autologous stem cell transplantation (ASCT) remains a standard component of frontline therapy and has substantially improved survival outcomes [ 1 , 2 ]. However, relapse or disease progression after ASCT remains frequent, underscoring the need for close post-transplant monitoring and timely intervention [ 3 , 4 ].

Increasing evidence indicates that minimal residual disease (MRD) assessment was a known surrogate marker for survival in MM [ 3 , 5 , 6 , 7 ]. Specifically, previous studies have consistently demonstrated that MRD negativity was associated with prolonged progression-free survival (PFS) and overall survival (OS) [ 2 , 8 , 9 ], irrespective of treatment regimen, cytogenetic risk profile, or International Staging System (ISS) stage at diagnosis [ 6 , 10 , 11 ]. Moreover, longer durations of persistent MRD negativity were associated with progressively improved clinical outcomes [ 2 , 12 , 13 ].

However, MRD status is dynamic, and its prognostic value is substantially enhanced by longitudinal assessments compared with single-time-point evaluation [ 14 , 15 , 16 , 17 , 18 ]. Most previous studies, nevertheless, have assessed MRD at isolated time points or focused on either pre-ASCT or post-ASCT status [ 19 ]. Notably, MRD positivity at either the pre- [ 11 ] or post-transplant [ 14 ] stage has been identified as an independent prognostic factor for disease outcomes [ 20 ], although pre-ASCT MRD status was shown to be associated with PFS but not OS [ 11 ]. In real-world clinical practice, MRD status frequently evolved during treatment, with more than half of patients experiencing MRD conversion between pre- and post-ASCT assessments [ 21 ]. Increasing evidence suggests that quantitative MRD burden might provide additional prognostic information, as deeper MRD responses and progressive reductions in MRD burden were associated with improved clinical outcomes [ 17 , 22 , 23 ]. Despite these advances, comprehensive evaluations that integrated MRD status, MRD burden, and their dynamics across both treatment phases remained incompletely characterized.

Moreover, because MRD conversion could occur at variable time points after transplantation, a clinically meaningful landmark is needed to distinguish early from late conversion. The 3-month post-ASCT time point has been widely adopted as a key milestone for MRD assessment [ 13 , 19 ]. However, whether MRD conversion occurred before or after this landmark carried distinct prognostic implications and remained unclear.

In this context, we conducted a real-world study of 335 transplant-eligible patients with MM treated at the First Affiliated Hospital of Sun Yat-sen University between 2010 and 2025. This study aimed to (1) evaluate the impact of dynamic MRD patterns on PFS and OS; (2) determine the prognostic relevance of the timing of MRD conversion (early vs. late) among patients who experienced MRD status changes between the pre- and post-ASCT assessments; and (3) assess whether quantitative MRD kinetics provide additional prognostic stratification among patients with persistent MRD positivity.

2. Methods

2.1. Study Design and Patients

Patients with newly diagnosed multiple myeloma who received induction therapy followed by upfront autologous stem cell transplantation at the First Affiliated Hospital of Sun Yat-sen University between August 2010 and July 2025 were retrospectively screened for eligibility. A total of 370 patients underwent induction therapy followed by upfront ASCT during the study period. Two patients were excluded because of a prior or concurrent active malignancy or severe comorbidities that could substantially affect treatment delivery or survival outcomes. Among the remaining 368 patients, five were excluded because pre-ASCT MRD status was unavailable, and seven were excluded because no evaluable post-ASCT MRD assessment was available. Of the 356 patients with evaluable MRD data both before and after ASCT, 21 were further excluded because the available longitudinal MRD data were insufficient to reliably classify their post-ASCT MRD trajectory. Ultimately, 335 patients were included in the final analysis ( Figure 1 ).

MRD status was assessed longitudinally before and after ASCT. Based on longitudinal changes in MRD status, patients were categorized into four dynamic MRD patterns: persistent MRD negativity (MRD− to MRD−), conversion from MRD-negative to MRD-positive (MRD− to MRD+), MRD-positive to MRD-negative (MRD+ to MRD−), and persistent MRD positivity (MRD+ to MRD+).

All patients met the diagnostic criteria of the International Myeloma Working Group (IMWG). The inclusion criteria were: (1) newly diagnosed multiple myeloma with complete clinical data; (2) receipt of induction therapy followed by ASCT; and (3) availability of evaluable treatment response and serial MRD assessments before and after ASCT. The exclusion criteria were: (1) a prior history of malignancy or the presence of concurrent active malignancies; and (2) severe comorbidities or complications that could substantially affect treatment delivery or survival outcomes. This study followed the principles of the Declaration of Helsinki and was approved by the Ethics Committee of The First Affiliated Hospital of Sun Yat-sen University (IRB [2026]303). The requirement for written informed consent was waived due to the retrospective nature of the study.

2.2. MRD Assessment

MRD assessment before ASCT was performed after completion of induction therapy. After transplantation, MRD evaluations were conducted as part of routine clinical practice, scheduled at approximately 3, 6, 9, and 12 months, and every 6 months thereafter. In this real-world retrospective cohort, the actual timing and frequency of post-ASCT MRD assessments varied according to clinical practice and the availability of evaluable bone marrow samples ( Supplementary Table S1 ). Persistent MRD negativity was defined as the absence of flow cytometry-detectable disease across all evaluable MRD assessments performed before and after ASCT, with a minimum sensitivity of 5 × 10 −5 . Persistent MRD positivity was defined as the presence of flow-detectable disease at all corresponding time points using the same sensitivity threshold. MRD+ to MRD− conversion was defined as detectable MRD before ASCT, followed by the first documentation of MRD negativity during post-transplant follow-up. Conversely, MRD− to MRD+ conversion was defined as undetectable MRD before ASCT, followed by the first detection of MRD positivity during longitudinal follow-up. Subsequent MRD fluctuations were not considered for reclassification.

Patients with MRD− to MRD+ conversion and MRD+ to MRD− conversion were further stratified into early (≤3 months after ASCT) and late (>3 months after ASCT) conversion groups. In addition, among patients with persistent MRD positivity, longitudinal MRD kinetics were categorized as decreasing versus stable or increasing MRD burden. MRD burden was quantified as the absolute percentage of phenotypically aberrant plasma cells. Decreasing MRD burden was defined as an overall downward trend in the absolute MRD percentage across serial assessments, whereas stable or increasing MRD burden was defined as a generally stable or upward trend over time.

Bone marrow MRD assessment was performed using multiparameter flow cytometry. The antibody panel included CD38, CD45, CD19, CD20, CD56, CD54, CD138, κ light chain, and λ light chain (BD Biosciences, San Jose, CA, USA). A minimum of 1 × 10 6 nucleated cells were analyzed per sample. Detection of ≥20 phenotypically aberrant plasma cells was considered MRD-positive, corresponding to a limit of detection of approximately 2 × 10 −5 to 1 × 10 −5 .

2.3. Efficacy Evaluation

Treatment response was assessed according to the International Myeloma Working Group (IMWG) criteria and categorized as complete response (CR), very good partial response (VGPR), partial response (PR), minimal response (MR), stable disease (SD), or progressive disease (PD). Response evaluations were performed after completion of induction therapy, every 3 months during the first year following ASCT and every 6 months thereafter as part of routine clinical follow-up.

2.4. Follow-Up

The primary endpoints of the study were PFS and OS. Patients were followed until 26 November 2025. OS was defined as the time from initiation of treatment to death from any cause or last follow-up. PFS was defined as the time from initiation of treatment to documented disease progression or last follow-up.

2.5. Statistical Analysis

Categorical variables were presented as frequencies and percentages and compared using the chi-square test, as appropriate. Survival distributions were estimated using the Kaplan–Meier method and compared with the log-rank test. To address potential immortal time bias, a 6-month landmark sensitivity analysis was performed among patients who were alive and progression-free at 6 months after ASCT. MRD groups were defined using assessments available up to the landmark, and survival was calculated from the landmark time point.

Cox proportional hazards regression models were used to evaluate associations between MRD dynamic patterns and survival outcomes in univariate and multivariable analyses. Variables with a p -value < 0.10 in univariate analyses, together with potential clinically relevant covariates with established prognostic significance, were included in the multivariable Cox regression models. The number of patients included in each Cox regression analysis is reported in the corresponding tables. Complete-case analysis was performed, and patients with missing values for any covariate included in a given model were excluded from that model. Given the limited extent of missingness, multiple imputation was not performed. In multivariable analyses, OS Cox models were adjusted for Revised International Staging System (R-ISS) staging, hemoglobin, serum lactate dehydrogenase (LDH) level, high-risk cytogenetic abnormalities (HRCAs), cytogenetic abnormalities, and maintenance therapy. PFS Cox models were adjusted for age, R-ISS staging, HRCAs, cytogenetic abnormality, maintenance treatment, and treatment response at ASCT. Hazard ratios (HRs) were reported with corresponding 95% confidence intervals (CIs).

The proportional hazards (PH) assumption was assessed using partial residual-based diagnostics combined with Spearman’s rank correlation analysis. Specifically, the correlations between the partial residuals of each MRD-group dummy variable and survival time were examined. When a significant correlation was identified, an extended Cox model incorporating an interaction term between the corresponding MRD-group dummy variable and time was fitted to further assess whether the effect of that variable changed during follow-up. A statistically significant group-by-time interaction term was considered evidence of a time-dependent effect and a potential violation of the PH assumption.

All statistical analyses were performed using SPSS software (version 27.0; IBM Corp., Armonk, NY, USA) and R software (version 4.5.2). A two-sided p -value < 0.05 was considered statistically significant.

3. Results

3.1. MRD Evaluation and Patient Characteristics

A total of 335 patients undergoing ASCT were included in the analysis ( Table 1 ). Based on dynamic changes in MRD status assessed longitudinally before and after ASCT, patients were classified into four groups: 82 patients (24.5%) with persistent MRD-negative disease, 40 patients (11.9%) who experienced MRD− to MRD+ conversion, 168 patients (50.1%) who converted from MRD+ to MRD−, and 45 patients (13.4%) with persistent MRD-positive disease. The median numbers of documented MRD assessments per patient were 6 (range, 2–10), 8 (range, 3–11), 7 (range, 2–11), and 4 (range, 2–10) in the persistent MRD-negative, MRD− to MRD+, MRD+ to MRD−, and persistent MRD-positive groups, respectively ( Supplementary Tables S1 and S2 ).

The proportion of patients achieving at least CR at the time of ASCT was significantly lower in the persistent MRD-positive group than in the other three MRD dynamic groups (61.0 vs. 47.5% vs. 39.9% vs. 11.1%; p < 0.001). In addition, female patients were more frequently represented in the persistent MRD-positive group, with a borderline difference observed among the four groups ( p = 0.050). In contrast, other baseline characteristics, including age, M-protein type, ISS and R-ISS, cytogenetic abnormalities, HRCAs, induction regimen, maintenance therapy, anti-CD38-based therapy, hemoglobin level, serum LDH level, and transplantation-related complications were well balanced across the four groups (all p > 0.05).

To better characterize dynamic changes in MRD, patients with MRD− to MRD+ conversion ( n = 40) were further stratified into early (≤3 months after ASCT; n = 12) and late (>3 months after ASCT; n = 28) conversion groups, and a similar early–late stratification was applied to patients with MRD+ to MRD− conversion ( n = 168; early, n = 96; late, n = 72) ( Table 2 ). In addition, among patients with persistent MRD positivity ( n = 45), subgroups were defined according to decreasing ( n = 20) versus stable or increasing ( n = 25) MRD burden ( Table 3 ). No significant differences in baseline demographic, disease-related, or treatment-related characteristics were observed across any of these subgroup analyses (all p > 0.05).

3.2. Overall Survival According to MRD Dynamic Patterns

Figure 2 A illustrates Kaplan–Meier estimates of OS according to MRD dynamic patterns. During follow-up, 99 deaths were recorded in the overall cohort. Median OS was not reached in the persistent-negative group and was 105 months in the MRD− to MRD+ group, 131 months in the MRD+ to MRD− group, and 87 months in the persistent MRD-positive group ( Supplementary Table S3 ) ( p < 0.001). Using the persistent MRD-negative group as the reference, patients with persistent MRD positivity had significantly worse OS ( p < 0.001). Patients with MRD− to MRD+ conversion also showed inferior OS compared with the persistent MRD-negative group ( p = 0.001). Although the Kaplan–Meier curves suggested a trend toward better OS in patients with persistent MRD negativity than in those with MRD+ to MRD− conversion, this difference did not reach statistical significance ( p = 0.244). Among the non-persistent MRD-negative groups, patients with persistent MRD positivity had significantly worse OS than those with MRD+ to MRD− conversion ( p = 0.002). OS appeared similar between the MRD− to MRD+ and persistent MRD-positive groups ( p = 1.000). In addition, OS tended to be poorer in patients with MRD− to MRD+ conversion than in those with MRD+ to MRD− conversion, with a borderline difference ( p = 0.080).

Median OS was 116 months in the early MRD− to MRD+ group and 84 months in the late MRD− to MRD+ group ( Figure 3 A; p = 0.840). Among patients with MRD+ to MRD− conversion, OS did not differ significantly between early and late converters, with median OS of 133 months in the early group and 119 months in the late group ( Figure 3 C; p = 0.677). In contrast, among patients with persistent MRD positivity, those with decreasing MRD burden had a longer median OS than those with stable or increasing MRD burden (106 months vs. 50 months; Figure 3 E; p = 0.033).

3.3. Progression-Free Survival According to MRD Dynamic Patterns

Figure 2 B shows Kaplan–Meier estimates of PFS according to MRD dynamic patterns. During follow-up, 161 progression events were recorded in the overall cohort. Median PFS was not reached in the persistent MRD-negative group and was 50 months in the MRD− to MRD+ group, 71 months in the MRD+ to MRD− group, and 38 months in the persistent MRD-positive group ( Supplementary Table S3 ) ( p < 0.001). Using the persistent MRD-negative group as the reference, patients with MRD− to MRD+ conversion had significantly shorter PFS ( p < 0.001). Patients with MRD+ to MRD− conversion also showed significantly inferior PFS compared with the persistent MRD-negative group ( p = 0.004). Persistent MRD positivity was likewise associated with significantly worse PFS than persistent MRD negativity ( p < 0.001). Among the non-persistent negative groups, PFS was significantly longer in patients with MRD+ to MRD− conversion than in those with MRD− to MRD+ conversion ( p = 0.011) and those with persistent MRD positivity ( p = 0.034). PFS appeared similar between the MRD− to MRD+ and persistent MRD-positive groups ( p = 1.000).

Among patients with MRD− to MRD+ conversion, Kaplan–Meier showed no significant difference in PFS between early and late conversion groups ( Figure 3 B; p = 0.520). In contrast, among patients with MRD+ to MRD− conversion, PFS was significantly longer in early converters than in late converters, with a median PFS of 81 months versus 58 months ( Figure 3 D; p = 0.035). Among patients with persistent MRD positivity, no statistically significant difference in PFS was observed between those with decreasing MRD burden and those with stable or increasing MRD burden, although a numerical separation of the Kaplan–Meier curves was noted. Median PFS was 59 months in patients with decreasing MRD burden and 20 months in those with stable or increasing MRD burden ( Figure 3 F; p = 0.150).

The 6-month landmark sensitivity analysis showed significant differences in OS and PFS among the four MRD dynamic groups ( p = 0.004 and p < 0.001, respectively), suggesting that the observed association was not solely attributable to immortal time bias ( Supplementary Figure S1 ).

3.4. Univariate and Multivariable Analyses of OS According to the MRD Dynamic Pattern

In univariate analysis, factors significantly associated with OS included R-ISS staging, hemoglobin, serum LDH levels, HRCAs, cytogenetic abnormality, and maintenance treatment ( Table 4 ). After adjustment for these prognostically relevant covariates, MRD dynamic status remained an independent predictor of OS. Compared with patients who remained MRD-negative, those who were persistently MRD-positive were associated with the highest risk of death (HR, 4.762; 95%CI, 2.151–10.546; p < 0.001), followed by conversion from MRD− to MRD+ (HR, 3.925; 95% CI, 1.791–8.598; p < 0.001), whereas conversion from MRD+ to MRD− showed a numerically increased but borderline-significant risk (HR, 1.973; 95%CI, 0.963–4.046; p = 0.063).

To assess the robustness of these findings, additional analyses were performed using alternative MRD categories as reference groups ( Supplementary Table S4 ). When persistent MRD positivity was used as the reference, patients who converted from MRD-negative to MRD-positive (MRD− to MRD+) exhibited a comparable OS (HR, 0.824; 95% CI, 0.428–1.586; p = 0.562).

In multivariable Cox regression analyses adjusted for predefined clinical covariates ( Table 5 ), the timing of MRD− to MRD+ conversion was not independently associated with OS. Late MRD− to MRD+ conversion showed no significant difference in OS compared with early conversion (HR, 0.962; 95% CI, 0.346–2.676; p = 0.942). Similarly, among patients with MRD+ to MRD− conversion, OS did not differ significantly between late and early converters after multivariable adjustment (HR, 0.968; 95% CI, 0.499–1.876; p = 0.923). In addition, compared with patients with decreasing MRD burden, those with stable or increasing MRD burden showed a higher risk of death after adjustment, with a borderline level of statistical significance (HR, 4.518; 95% CI, 0.956–21.340; p = 0.057).

3.5. Univariate and Multivariable Analyses of PFS According to MRD Dynamic Patterns

In univariate analysis for PFS, age, R-ISS staging, HRCAs, cytogenetic abnormality, maintenance treatment, and response at ASCT were significantly associated with PFS ( Table 6 ). After multivariable adjustment, MRD dynamic status remained independently associated with PFS. Compared with patients who remained MRD-negative, both MRD− to MRD+ conversion (HR, 4.432; 95% CI, 2.452–8.012; p < 0.001) and persistent MRD positivity (HR, 4.402; 95% CI, 2.293–8.454; p < 0.001) were associated with markedly increased risks of disease progression, whereas MRD+ to MRD− conversion was associated with a more moderate increase in risk (HR, 2.319; 95% CI, 1.377–3.906; p = 0.002).

To further evaluate the robustness of these associations, additional analyses were conducted using alternative MRD categories as reference groups ( Supplementary Table S4 ). When persistent MRD positivity was used as the reference, patients who converted from MRD− to MRD+ demonstrated a comparable risk of progression (HR, 1.007; 95% CI, 0.574–1.765; p = 0.981), indicating similar PFS outcomes between these two unfavorable MRD trajectories.

For PFS, late MRD− to MRD+ conversion was not independently associated with PFS compared with early conversion (HR, 1.070; 95% CI, 0.415–2.758; p = 0.888). However, among patients with MRD+ to MRD− conversion, late conversion was independently associated with inferior PFS compared with early conversion (HR, 1.857; 95% CI, 1.136–3.034; p = 0.014). Patients with stable or increasing MRD burden had a significantly higher risk of disease progression than those with decreasing MRD burden after multivariable adjustment (HR, 4.397; 95% CI, 1.190–16.240; p = 0.026).

In addition, we tested the proportional hazards assumption for the multivariable Cox models of OS and PFS according to dynamic MRD patterns. No significant time-varying effects were identified ( Supplementary Table S5 ).

4. Discussion

In recent years, research has demonstrated that dynamic monitoring of MRD not only helped identify the progression trends of MM but also provided crucial insights for personalized treatment, enabling dynamic risk-adapted management based on MRD [ 12 , 22 ]. This shift allowed treatment strategies to evolve from traditional binary MRD endpoints to more refined and individualized approaches [ 15 , 24 ]. Previous studies have demonstrated the prognostic value of longitudinal MRD assessment. Gu et al. identified distinct MRD evolution patterns associated with outcomes in transplant-eligible patients [ 18 ]. de Tute et al. and Ferrero et al. further demonstrated the prognostic relevance of serial post-ASCT MRD assessment and MRD kinetics [ 19 , 23 ]. Against this context, we performed a real-world longitudinal analysis of 335 transplant-eligible MM patients treated over a 15-year period, with a median follow-up of 127 months. In addition to evaluating longitudinal MRD trajectories, we specifically examined the prognostic implications of the timing of post-ASCT MRD conversion and quantitative MRD burden changes. To the best of our knowledge, this was one of the longest follow-up studies to date on MRD status in MM patients. The primary finding of this study was that persistent MRD negativity conferred the most favorable prognosis, followed by MRD+ to MRD− conversion, whereas MRD− to MRD+ conversion and persistent MRD positivity were associated with the poorest outcomes. This further illustrated that MRD trajectory, rather than static MRD status, provided biologically and clinically meaningful prognostic information.

In our study, patients with MRD− to MRD+ demonstrated survival outcomes similar to those with persistent MRD positivity. However, a previous study suggested that persistent MRD-positive patients might have worse outcomes than those who convert from MRD− to MRD+ [ 20 ]. One possible explanation for this discrepancy lies in differences in response depth within the MRD-positive subgroup. In our study, a relatively high proportion of persistent MRD-positive patients achieved CR. Emerging evidence indicated that, even among MRD-positive patients, those achieving CR had superior survival compared with those who failed to reach CR [ 25 ]. Therefore, the prognostic impact of persistent MRD positivity might be partially modulated by the depth of conventional response, which could attenuate the survival disadvantage in this subgroup and contribute to the observed pattern.

In our study, patients who converted from MRD+ to MRD− had inferior survival outcomes compared with those with persistent MRD negativity. This observation is consistent with accumulating evidence suggesting that the timing and durability of MRD negativity are critical determinants of prognosis in multiple myeloma [ 13 ]. Taken together, these findings suggest that earlier achievement and longer persistence of MRD negativity reflect deeper disease eradication and a more favorable tumor biology, whereas delayed MRD clearance might indicate residual clonal populations that remain sensitive to therapy but are not fully eliminated.

In subgroup analysis, early conversion from MRD+ to MRD− was associated with superior PFS compared with delayed clearance, suggesting that rapid MRD eradication might reflect greater treatment sensitivity and more durable disease control. In the TOURMALINE-MM3 and -MM4 trials, MRD conversion from positive to negative within the first 14 months after ASCT was associated with significantly improved PFS compared with persistent MRD positivity, whereas this difference was no longer observed at 28 months [ 14 ]. These findings are consistent with our results, suggesting that earlier MRD clearance was associated with more favorable clinical outcomes. Similar conclusions have also been reported in studies by Lucia Y. Chen [ 13 ] and Joaquin Martinez-Lopez [ 4 ], which demonstrated that early achievement of MRD negativity is associated with improved survival outcomes in multiple myeloma. Furthermore, some studies have suggested that the prognosis associated with MRD conversion to negativity might be particularly pronounced in patients with R-ISS stage III disease [ 26 ], highlighting the potential importance of early MRD clearance in biologically high-risk populations.

In contrast, MRD conversion to positivity was associated with inferior outcomes irrespective of timing in our cohort. The similarly poor outcomes observed in patients with MRD-negative to MRD-positive conversion and those with persistent MRD positivity may reflect several biological and methodological factors. Re-emergence of MRD after an initially negative assessment may indicate clonal evolution or expansion of treatment-resistant subclones under therapeutic pressure, representing a biologically aggressive disease state similar to persistent MRD positivity. In addition, spatial heterogeneity of bone marrow involvement and sampling variability may lead to transient false-negative MRD results, particularly when residual disease is patchy or located outside the sampled site. Finally, MRD reappearance may precede biochemical or clinical relapse by several months, suggesting that conversion to MRD positivity identifies an early phase of renewed clonal expansion [ 20 ]. These mechanisms may partly explain why patients with MRD-negative to MRD-positive conversion had outcomes comparable to those with persistent MRD positivity. Notably, prior studies have reported that early MRD re-emergence (<10 years from diagnosis) is associated with significantly shorter PFS and OS, whereas late conversion (>10 years) might not compromise overall survival compared with persistent MRD negativity [ 20 ]. The discrepancy between these observations and our findings might be explained by differences in landmark definitions, and the operationalization of “early” versus “late” conversion could substantially influence risk stratification. Moreover, when adopting a 10-year landmark from diagnosis, non-myeloma-related mortality becomes increasingly relevant, including age-related frailty, comorbid conditions, and secondary malignancies, which might substantially influence overall survival and even PFS estimates. Long-term data suggested that MRD conversion might occur even after prolonged remission. Notably, the median interval between MRD re-emergence and clinical relapse was approximately one year. In high-risk patients, the interval between MRD conversion to positivity and overt relapse might be substantially shorter [ 20 ], indicating the existence of a potentially actionable preclinical window.

Quantitative MRD kinetics refine prognostication beyond categorical MRD status. Among persistently MRD-positive patients, stable or increasing MRD burden identified a biologically aggressive subset. These findings supported the concept that MRD kinetics capture clonal fitness and therapeutic pressure. Decreasing MRD burden might reflect continued responsiveness to maintenance therapy and potentially precede eventual conversion to MRD negativity [ 27 , 28 ]. Conversely, rising MRD burden likely represents early clonal re-expansion before overt biochemical relapse. Our finding is consistent with the results of a study by Ferrero et al., which found that patients with a pattern of MRD that was stable or increasing over time had a significantly shorter OS than patients with a pattern of MRD that was decreasing [ 23 ]. Using NGS-based quantitative MRD assessment, Bal et al. demonstrated a further reduction in myeloma disease burden following ASCT after quadruplet induction therapy, highlighting the ability of serial quantitative MRD measurements to capture treatment-related changes in residual disease burden [ 29 ]. Additionally, different MRD kinetics could be more useful for comparing the efficacy of different treatment strategies and for implementing individualized therapy-monitoring strategies.

Our findings regarding the prognostic importance of persistent MRD negativity are consistent with recent studies conducted using more sensitive MRD techniques and contemporary anti-CD38-based regimens. In analyses of the MAIA and ALCYONE trials [ 30 ], daratumumab-containing regimens substantially increased the rates and durability of MRD negativity, as assessed by next-generation sequencing. MRD negativity sustained for at least 6 or 12 months was associated with superior progression-free survival, supporting the prognostic importance of the durability of MRD response rather than a single negative assessment. Similar findings have been reported in transplant-eligible patients receiving anti-CD38-based quadruplet therapy. In the MASTER trial [ 28 ], patients received daratumumab, carfilzomib, lenalidomide, and dexamethasone followed by ASCT and MRD response-adapted consolidation. Serial MRD assessments were performed using NGS, and patients who achieved MRD negativity in two consecutive treatment phases were permitted to discontinue therapy and enter MRD surveillance. These findings indicate that sustained MRD negativity provides greater prognostic information than a single MRD-negative assessment. Although MRD was assessed using conventional multiparameter flow cytometry in the present cohort and only a small proportion of patients received anti-CD38 monoclonal antibodies, our results similarly showed that persistent MRD negativity was associated with the most favorable long-term outcomes. Nevertheless, direct comparisons with contemporary trials should be interpreted cautiously because of differences in assay sensitivity, treatment intensity, MRD assessment schedules, and patient characteristics.

This study is strengthened by prolonged follow-up, comprehensive serial MRD integration, and multivariable adjustment. Nevertheless, several limitations warrant consideration. First, the study spanned a long inclusion period of 15 years, during which substantial changes occurred in induction and maintenance regimens, supportive care, transplant practice, and MRD technologies. Anti-CD38-based therapy has been shown to improve treatment response and survival outcomes in multiple myeloma [ 31 ]; however, only a small proportion of patients in our cohort received such therapy, precluding a reliable subgroup analysis. Although treatment-related factors were included in the multivariable models, residual confounding related to treatment era, treatment intensity, and evolving clinical practice cannot be completely excluded. In addition, although MRD assessments were performed in the same institutional laboratory using broadly consistent procedures, potential variability related to reagent lots, instrument performance, and assay implementation over the 15-year period cannot be completely excluded and may have introduced measurement variability. Second, next-generation flow (NGF) and next-generation sequencing (NGS), which enable MRD detection at sensitivities approaching 10 −6 , are now routinely incorporated into contemporary clinical trials. However, MRD was assessed throughout the study using conventional multiparameter flow cytometry rather than NGS or NGF. Future studies employing deeper sequencing-based approaches might further refine longitudinal kinetic modeling. A third limitation is that we acknowledge this study is a retrospective observational study. Despite adjusting for multiple factors, there remain unavoidable confounding factors. Future prospective studies should be conducted to validate the findings of this study. Fourth, although the 6-month landmark analysis yielded results consistent with the primary analyses, immortal time bias cannot be completely excluded, particularly for MRD conversions occurring beyond the landmark period. In addition, the 3-month cutoff used to distinguish early from late conversion, although clinically based and prespecified, remains somewhat arbitrary. In addition, subsequent MRD assessment intervals were not completely uniform in this real-world cohort. Therefore, the observed survival differences should be interpreted as prognostic associations rather than causal effects. Moreover, a small proportion of patients were excluded because of sporadic missing MRD data rather than early progression or mortality. Although selection bias was likely limited, it cannot be completely excluded. Fifth, several subgroup analyses involved small sample sizes and limited numbers of outcome events, resulting in wide confidence intervals and reduced statistical power. Multiple subgroup and pairwise comparisons may also increase the risk of Type I error and false-positive findings. Accordingly, these analyses are considered exploratory and should be interpreted with caution until validated in larger independent cohorts.

5. Conclusions

In conclusion, longitudinal MRD trajectories provided independent and incremental prognostic information beyond static MRD assessment in patients with MM. Persistent MRD negativity conferred the best prognosis, followed by MRD+ to MRD− conversion, while persistent MRD positivity and MRD− to MRD+ conversion were associated with the worst outcomes. Furthermore, for patients who transitioned from MRD+ to MRD−, earlier conversion to MRD negativity correlated with better PFS. These findings suggest that longitudinal MRD trajectories might provide additional prognostic information beyond binary MRD endpoints and help refine risk stratification.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/cancers18182958/s1 , Figure S1: Six-month landmark analysis of overall survival (OS) and progression-free survival (PFS) according to dynamic MRD patterns. (A) OS stratified by four dynamic MRD patterns: MRD− to MRD−, MRD− to MRD+, MRD+ to MRD−, and MRD+ to MRD+. (B) PFS stratified by the same dynamic MRD patterns. Numbers at risk at selected time points are displayed below each panel. Overall differences among groups were assessed using the log-rank test; Table S1: Individual-level longitudinal MRD status in the 335 included patients; Table S2: Summary table of MRD assessment frequency; Table S3: Median Overall Survival (OS) and Progression-Free Survival (PFS) according to different MRD dynamic patterns; Table S4: Univariate and multivariate analyses of OS and PFS using different MRD statuses as reference; Table S5: Proportional hazards assumption testing for MRD dynamic groups in the Cox regression model.

Author Contributions

X.Z. and J.L. (Juan Li) conceived and designed the study and drafted the analysis plan. X.Z. and X.L. collected the data. X.Z. performed the data analysis and drafted the manuscript. X.L. supervised the writing and participated in manuscript revision. J.G. directed the statistical analysis. B.H. reviewed the original data. J.L. (Junru Liu), M.C. and L.K. reviewed and revised the manuscript. J.L. (Juan Li) reviewed the study results and critically revised the manuscript. All authors contributed to the article and approved the submitted version. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

This study was approved by the Ethics Committee of The First Affiliated Hospital of Sun Yat-sen University (protocol code IRB [2026]303 and date of approval is 18 June 2026) and conducted in accordance with the Declaration of Helsinki.

Informed Consent Statement

The requirement for written informed consent was waived due to the retrospective nature of the study.

Data Availability Statement

The original contributions presented in this study are included in the article and its Supplementary Materials . Further inquiries can be directed to the corresponding author.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This study was supported by the National Natural Science Foundation of China (grant numbers 82270209 and 82470197).

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

References

1. Dimopoulos M.A., Terpos E., Boccadoro M., Moreau P., Mateos M.-V., Zweegman S., Cook G., Engelhardt M., Delforge M., Hajek R., et al. EHA–EMN Evidence-Based Guidelines for diagnosis, treatment and follow-up of patients with multiple myeloma. Nat. Rev. Clin. Oncol. 2025;22:680–700. doi: 10.1038/s41571-025-01041-x.

2. Fan H., Wang W., Zhang Y., Wang J., Cheng T., Qiu L., Wang X., Xia Z., An G. Current treatment paradigm and survival outcomes among patients with newly diagnosed multiple myeloma in China: A retrospective multicenter study. Cancer Biol. Med. 2023;20:77–87. doi: 10.20892/j.issn.2095-3941.2022.0612.

3. Meseha M., Hoffman J., Kazandjian D., Landgren O., Diamond B. Minimal Residual Disease-Adapted Therapy in Multiple Myeloma: Current Evidence and Opinions. Curr. Oncol. Rep. 2024;26:679–690. doi: 10.1007/s11912-024-01537-2.

4. Martinez-Lopez J., Alonso R., Wong S.W., Rios R., Shah N., Ruiz-Heredia Y., Sanchez-Pina J.M., Sanchez R., Bahri N., Zamanillo I., et al. Making clinical decisions based on measurable residual disease improves the outcome in multiple myeloma. J. Hematol. Oncol. 2021;14:1–4. doi: 10.1186/s13045-021-01135-w.

5. Munshi N.C., Avet-Loiseau H., Rawstron A.C., Owen R.G., Child J.A., Thakurta A., Sherrington P., Samur M.K., Georgieva A., Anderson K.C., et al. Association of Minimal Residual Disease with Superior Survival Outcomes in Patients with Multiple Myeloma: A meta-analysis. JAMA Oncol. 2017;3:28–35. doi: 10.1001/jamaoncol.2016.3160.

6. Munshi N.C., Avet-Loiseau H., Anderson K.C., Neri P., Paiva B., Samur M., Dimopoulos M., Kulakova M., Lam A., Hashim M., et al. A large meta-analysis establishes the role of MRD negativity in long-term survival outcomes in patients with multiple myeloma. Blood Adv. 2020;4:5988–5999. doi: 10.1182/bloodadvances.2020002827.

7. Rodriguez-Otero P., Paiva B., San-Miguel J.F. Roadmap to cure multiple myeloma. Cancer Treat. Rev. 2021;100:102284. doi: 10.1016/j.ctrv.2021.102284.

8. Rawstron A.C., Child J.A., de Tute R.M., Davies F.E., Gregory W.M., Bell S.E., Szubert A.J., Navarro-Coy N., Drayson M.T., Feyler S., et al. Minimal Residual Disease Assessed by Multiparameter Flow Cytometry in Multiple Myeloma: Impact on Outcome in the Medical Research Council Myeloma IX Study. J. Clin. Oncol. 2013;31:2540–2547. doi: 10.1200/jco.2012.46.2119.

9. Paiva B., Vidriales M.B., Cerveró J., Mateo G., Pérez J.J., Montalbán M.A., Sureda A., Montejano L., Gutiérrez N.C., García de Coca A., et al. Multiparameter flow cytometric remission is the most relevant prognostic factor for multiple myeloma patients who undergo autologous stem cell transplantation. Blood. 2008;112:4017–4023. doi: 10.1182/blood-2008-05-159624.

10. de Tute R.M., Rawstron A.C., Gregory W.M., Child J.A., Davies F.E., Bell S.E., Cook G., Szubert A.J., Drayson M.T., Jackson G.H., et al. Minimal residual disease following autologous stem cell transplant in myeloma: Impact on outcome is independent of induction regimen. Haematologica. 2015;101:e69–e71. doi: 10.3324/haematol.2015.128215.

11. Pasvolsky O., Pasyar S., Bassett R.L., Khan H.N., Tanner M.R., Bashir Q., Srour S., Saini N., Lin P., Ramdial J., et al. Impact of pretransplant minimal residual disease in patients with multiple myeloma and a very good partial response or better receiving autologous hematopoietic stem cell transplantation. Cancer. 2023;130:1663–1672. doi: 10.1002/cncr.35171.

12. Terpos E., Malandrakis P., Ntanasis-Stathopoulos I., Kostopoulos I.V., Eleutherakis-Papaiakovou E., Kanellias N., Spiliopoulou V., Migkou M., Fotiou D., Theodorakakou F., et al. Sustained bone marrow and imaging MRD negativity for 3 years drives discontinuation of maintenance post-ASCT in myeloma. Blood. 2025;145:2353–2360. doi: 10.1182/blood.2024027686.

13. Chen L.Y., Thibaud S., Bodnar S., Chari A., Richter J., Cho H.J., Sanchez L.J., Rodriguez C., Rossi A.C., Richard S., et al. MRD-negative duration following latest line of therapy predicts long-term PFS in real-world patients with multiple myeloma. Blood Adv. 2025;9:176–179. doi: 10.1182/bloodadvances.2024014097.

14. Paiva B., Manrique I., Dimopoulos M.A., Gay F., Min C.-K., Zweegman S., Špička I., Teipel R., Mateos M.-V., Giuliani N., et al. MRD dynamics during maintenance for improved prognostication of 1280 patients with myeloma in the TOURMALINE-MM3 and -MM4 trials. Blood. 2023;141:579–591. doi: 10.1182/blood.2022016782.

15. Firestone R.S., Simhal A.K., McAvoy D., Jurgens E.M., Garcés J.-J., Nemirovsky D., Derkach A., Maclachlan K.H., Hultcrantz M., Mailankody S., et al. Immune MRD Status Informs Tumor MRD Outcome Prognostication in Patients with Multiple Myeloma on Lenalidomide Maintenance. Clin. Cancer Res. 2025;31:4446–4456. doi: 10.1158/1078-0432.CCR-25-0572.

16. Diamond B., Korde N., Lesokhin A.M., Smith E.L., Shah U., Mailankody S., Hultcrantz M., Hassoun H., Lu S.X., Tan C., et al. Dynamics of minimal residual disease in patients with multiple myeloma on continuous lenalidomide maintenance: A single-arm, single-centre, phase 2 trial. Lancet Haematol. 2021;8:e422–e432. doi: 10.1016/s2352-3026(21)00130-7.

17. Martinez-Lopez J., Lopez-Muñoz N., Chari A., Dorado S., Barrio S., Arora S., Kumar A., Chung A., Martin T., Wolf J. Measurable residual disease (MRD) dynamics in multiple myeloma and the influence of clonal diversity analyzed by artificial intelligence. Blood Cancer J. 2024;14:131. doi: 10.1038/s41408-024-01102-x.

18. Gu J., Liu J., Chen M., Huang B., Li J. Longitudinal Flow Cytometry Identified “Minimal Residual Disease” (MRD) Evolution Patterns for Predicting the Prognosis of Patients with Transplant-Eligible Multiple Myeloma. Biol. Blood Marrow Transpl. 2018;24:2568–2574. doi: 10.1016/j.bbmt.2018.07.040.

19. de Tute R.M., Pawlyn C., Cairns D.A., Davies F.E., Menzies T., Rawstron A., Jones J.R., Hockaday A., Henderson R., Cook G., et al. Minimal Residual Disease After Autologous Stem-Cell Transplant for Patients with Myeloma: Prognostic Significance and the Impact of Lenalidomide Maintenance and Molecular Risk. J. Clin. Oncol. 2022;40:2889–2900. doi: 10.1200/JCO.21.02228.

20. Mohan M., Kendrick S., Szabo A., Yarlagadda N.K., Atwal D., Pandey Y., Roy A.M., Parikh R., Lopez J., Thanendrarajan S., et al. Clinical implications of loss of bone marrow minimal residual disease negativity in multiple myeloma. Blood Adv. 2022;6:808–817. doi: 10.1182/bloodadvances.2021005822.

21. Badros A., Foster L., Anderson L.D., Chaulagain C.P., Pettijohn E., Cowan A.J., Costello C., Larson S., Sborov D.W., Shain K.H., et al. Daratumumab with lenalidomide as maintenance after transplant in newly diagnosed multiple myeloma: The AURIGA study. Blood. 2025;145:300–310. doi: 10.1182/blood.2024025746.

22. Giri S., Dhakal B., Callander N.S., Medvedova E., Godby K., Dholaria B.R., Bal S., Ravi G., Chhabra S., Silbermann R.W., et al. Optimal MRD-based end point to support response-adapted treatment cessation in newly diagnosed multiple myeloma. Blood. 2025;146:707–716. doi: 10.1182/blood.2024027674.

23. Ferrero S., Ladetto M., Drandi D., Cavallo F., Genuardi E., Urbano M., Caltagirone S., Grasso M., Rossini F., Guglielmelli T., et al. Long-term results of the GIMEMA VEL-03-096 trial in MM patients receiving VTD consolidation after ASCT: MRD kinetics’ impact on survival. Leukemia. 2014;29:689–695. doi: 10.1038/leu.2014.219.

24. Paiva B., Shi Q., Puig N., Cedena M.-T., Orfao A., Durie B.G.M., Munshi N.C., San-Miguel J. Opportunities and challenges for MRD assessment in the clinical management of multiple myeloma. Nat. Rev. Clin. Oncol. 2025;22:424–438. doi: 10.1038/s41571-025-01017-x.

25. Mao J., Xue L., Wang H., Zhou H., Zhu Y., Jia T., Cai Z., Zhao L., Zhao L., Wang Y., et al. Cytogenetic risk stratification combined with minimal residual disease status influences the therapeutic outcome and prognosis of multiple myelomas. Sci. Rep. 2025;15:12545. doi: 10.1038/s41598-025-97125-w.

26. Sun Q., Li X., Gu J., Huang B., Liu J., Chen M., Li J. Prognostic Significance of the Stage at Which an MRD-Negative Status Is Achieved for Patients with Multiple Myeloma Who Received ASCT. Front. Oncol. 2022;12:776920. doi: 10.3389/fonc.2022.776920.

27. Costa L.J., Chhabra S., Medvedova E., Dholaria B.R., Schmidt T.M., Godby K.N., Silbermann R., Dhakal B., Bal S., Giri S., et al. Daratumumab, Carfilzomib, Lenalidomide, and Dexamethasone with Minimal Residual Disease Response-Adapted Therapy in Newly Diagnosed Multiple Myeloma. J. Clin. Oncol. 2021;40:2901–2912. doi: 10.1200/jco.21.01935.

28. Costa L.J., Chhabra S., Medvedova E., Dholaria B.R., Schmidt T.M., Godby K.N., Silbermann R., Dhakal B., Bal S., Giri S., et al. Minimal residual disease response-adapted therapy in newly diagnosed multiple myeloma (MASTER): Final report of the multicentre, single-arm, phase 2 trial. Lancet Haematol. 2023;10:e890–e901. doi: 10.1016/s2352-3026(23)00236-3.

29. Bal S., Dhakal B., Silbermann R.W., Schmidt T.M., Dholaria B., Giri S., Chhabra S., Medvedova E., Godby K.N., D’Souza A., et al. Impact of autologous hematopoietic cell transplantation on disease burden quantified by next-generation sequencing in multiple myeloma treated with quadruplet therapy. Am. J. Hematol. 2022;97:1170–1177. doi: 10.1002/ajh.26640.

30. San-Miguel J., Avet-Loiseau H., Paiva B., Kumar S., Dimopoulos M.A., Facon T., Mateos M.-V., Touzeau C., Jakubowiak A., Usmani S.Z., et al. Sustained minimal residual disease negativity in newly diagnosed multiple myeloma and the impact of daratumumab in MAIA and ALCYONE. Blood. 2022;139:492–501. doi: 10.1182/blood.2020010439.

31. Corre J., Vincent L., Moreau P., Hebraud B., Hulin C., Béné M.C., Broijl A., Caillot D., Delforge M., Dejoie T., et al. Daratumumab-bortezomib-thalidomide-dexamethasone for newly diagnosed myeloma: CASSIOPEIA minimal residual disease results. Blood. 2025;146:679–692. doi: 10.1182/blood.2024027620.

Associated Data

Supplementary Materials

Data Availability Statement

The original contributions presented in this study are included in the article and its Supplementary Materials . Further inquiries can be directed to the corresponding author.