Динамическая стратификация риска в мягкой миеломе: интеграция эволюционирующих биомаркеров с моделью 2/20/20.

INTRODUCTION

Smoldering multiple myeloma (SMM) is an asymptomatic precursor condition to multiple myeloma (MM), characterized by 10%–59% bone marrow plasma cells (BMPC) and/or a monoclonal (M) protein ≥3 g/dL, in the absence of myeloma‐defining events. 1 Based on the nationwide Icelandic screening study (iStopMM), the prevalence of SMM among individuals aged ≥40 years is approximately 0.53%. 2 Additionally, data from cancer registries in the United States and Sweden indicate that around 14% of MM cases were detected and reported at the SMM stage. 3 , 4 SMM is a biologically heterogeneous condition with highly variable risks of progression to MM. 5 This variability suggests that SMM may encompass various subtypes with diverse biological behaviors and genomic profiles. 6 , 7 In some individuals, SMM represents an early form of MM, with ongoing transformation but not yet meeting diagnostic criteria for symptomatic disease. 8 In others, SMM is more indolent and resembles monoclonal gammopathy of undetermined significance (MGUS), with a proportion of patients never progressing to MM. Moreover, disease trajectory can shift with time due to factors such as genomic alterations or changes in the bone marrow microenvironment. 9

It is critical to distinguish high‐risk SMM patients likely to have rapid progression from those with a more stable disease course, particularly as early treatment strategies are increasingly being explored. 10 , 11 , 12 Several risk stratification models have been developed and are now widely adopted in clinical management. 13 , 14 , 15 These models rely on biomarkers reflecting tumor burden at initial SMM diagnosis. The model most widely used currently is the 2/20/20 model published by Lakshman et al., which stratifies patients based on BMPC >20%, M‐protein >2 g/dL, and a free light chain ratio (FLCr) >20, into low (0 points), intermediate (1 points), and high‐risk (2–3 points) groups, associated with 2‐year progression rates of 10%, 26%, and 47%, respectively. 13 However, baseline risk stratification models are hampered by several limitations. The risk classification of individual patients can vary considerably across commonly used models, with patients classified as high‐risk by one model sometimes classified as low‐ or intermediate‐risk by another, highlighting the imprecision of static baseline assessment. 2 , 16 Furthermore, these models are based on disease biomarkers at initial diagnosis, without considering how the disease evolves during follow‐up. Since SMM is an asymptomatic condition, diagnosis is typically incidental. Initial biomarkers may not fully capture the risk of progression, particularly in patients who have a low tumor burden at initial diagnosis but an aggressive underlying biology, with high risk of rapid progression. While rising parameters are considered high risk by clinical guidelines, 17 there are no uniformly adopted criteria for what constitutes evolving disease.

To overcome these limitations, we were motivated to evaluate the dynamic changes in the M‐protein and FLCr, define criteria for evolving disease, and assess their impact on the risk of progression to MM. We further aimed to integrate these dynamic markers into existing risk stratification models to improve prognostic accuracy in SMM.

METHODS

Study cohort

We included all patients who were diagnosed with SMM at Memorial Sloan Kettering Cancer Center (MSK) between January 2002 and December 2019, with follow‐up through May 2023. Patients were identified through an automated search of electronic medical records using a combination of free text and ICD‐10 codes (C90.0 and later D47.2). All patient electronic medical records were manually reviewed to confirm the diagnosis of SMM, defined by the IMWG 2014 criteria. 1 Dates of diagnosis and progression, as well as baseline laboratory, pathology, and imaging data, were manually reviewed. All patients had baseline data within four months of SMM diagnosis. Advanced imaging was defined as positron emission tomography–computed tomography (PET‐CT), whole‐body or total spine magnetic resonance imaging (MRI), or whole‐body CT. Serial M‐protein and FLC levels between the date of diagnosis and progression were obtained through computerized data extraction. All free light chain assessments were performed using the Freelite© assay (The Binding Site). The FLCr was defined as the ratio of the involved to the uninvolved free light chain level.

For the cohort for the evolving biomarker analysis, we only included patients who had one or more laboratory evaluations around 12 months (between 10 and 14 months) from diagnosis. Patients with missing data for either M‐protein or FLCr at this timepoint were excluded. The study was conducted according to the Declaration of Helsinki and was approved by the MSK Institutional Review Board. Participant informed consent was waived, given the retrospective nature of the study.

Follow‐up and outcomes

Per NCCN guidelines and institutional practice at MSK, patients with SMM are monitored with laboratory evaluations, including serum protein electrophoresis and FLC assay approximately every 3 months, and advanced imaging annually or as needed. 17 Monitoring frequency may be escalated or de‐escalated at the treating physician's discretion based on clinical risk assessment.

Time to progression (TTP) was defined as the time from SMM diagnosis to progression to MM or AL amyloidosis as defined by the IMWG 2014 criteria. 1 Patients who had not progressed at the end of follow‐up were censored at the date of the last clinic visit. Patients were also censored at the time of death if there was no evidence of progression of SMM. Patients who started treatment on a clinical trial for SMM were censored at the date of treatment initiation.

Statistical analysis

Patients were stratified according to the baseline Mayo 2/20/20 risk score, which assigns one point each for bone marrow plasma cell infiltration >20%, M‐protein >2 g/dL, and FLCr >20, classifying patients as low‐ (0 points), intermediate‐ (1 point), or high‐risk (2–3 points). 13 The cumulative incidence of progression from SMM to MM was estimated using the Kaplan−Meier method, and progression risk was compared across risk groups using the log‐rank test.

To evaluate the prognostic value of longitudinal changes in serum biomarkers, we performed a 1‐year landmark analysis, restricting the analysis to patients who had not progressed within the first year and had available biomarker measurements at approximately 12 months (±2 months) from diagnosis. We quantified change in each biomarker (Δ) by calculating the average monthly change (difference between the 1‐year and baseline values, divided by the number of months between those measurements). This approach accounts for variation in follow‐up timing (e.g., patients measured at 10 vs. 14 months). These values were then scaled to reflect the change over a full year, and these annualized values were used for subsequent analyses. Changes in M‐protein were modeled as absolute differences (g/dL). Changes in FLCr were log‐transformed prior to analysis given the skewed distribution of values, and are expressed as percentage increases for interpretability. Patients were then divided into quintiles based on the magnitude of change in M‐protein and FLCr (ΔQ1–ΔQ5) during the first year, and the Kaplan–Meier method was used to estimate progression risk from the 1‐year landmark across quintiles, with comparisons made using the log‐rank test.

We used Cox proportional hazards regression to assess the association between the biomarker change and risk of progression from the 1‐year landmark, adjusting for age at diagnosis, sex, and the baseline Mayo 2/20/20 risk score. 13 Results are presented as hazard ratios (HRs) with 95% confidence intervals (CIs). We then performed a systematic grid search to define dichotomous thresholds for evolving M‐protein (eMP) and evolving FLCr (eFLCr). For each candidate cut‐off from the grid of possible values, we fitted the Cox proportional hazards model while adjusting for the same set of covariates and selected the pair with the highest log‐likelihood. Model performance was assessed using five‐fold cross‐validation repeated ten times, comparing the baseline Mayo 2/20/20 model with the dynamic model. Predictive accuracy was assessed using both the area under the receiver operating characteristic curve (AUC) for 2‐year progression from the 1‐year landmark, and the concordance index (C‐index).

We employed a two‐tailed P value with an alpha level of 0.05 in all analyses that assessed associations. R 4.3.2 was used for all statistical analyses.

RESULTS

Baseline patient characteristics

A total of 323 patients were included in the study (Table 1 and Supporting Information S1: Figure 1 ). The median age was 63 years (interquartile range [IQR], 55–72; Table 1 ), with 186 (58%) men. Among all patients, 65% had IgG SMM, 21% had IgA SMM, 6% had free light chain SMM, 7.4% had biclonal SMM, and 0.6% had IgM SMM. All SMM patients had undergone a bone marrow biopsy; the median BMPC was 18% (IQR, 13–25). Median M‐protein level and FLCr were 1.3 g/dL (IQR, 0.7–1.9) and 8 (IQR, 2.9–22), respectively. At baseline, 24% of patients were classified as high‐risk, 30% as intermediate‐risk, and 46% as low‐risk according to the 2/20/20 criteria. Advanced imaging, including PET‐CT, whole‐body MRI or total spine MRI, or whole‐body CT, was performed at baseline in 80% of patients, and prior to progression in 87% of all patients, while 20% underwent skeletal surveys (Table 1 ).

Follow‐up and outcomes

The median number of laboratory evaluations during the first year was 5 across all baseline risk groups, with a median interval of approximately 3 months between assessments. Higher‐risk patients were monitored slightly more frequently (median time between assessments 2.4 vs. 2.9 months, P = 0.006), consistent with risk‐adapted clinical practice.

With a median follow‐up period of 53 months (95% CI, 48–62), 115 SMM patients progressed to MM or AL amyloidosis, with an overall median TTP of 115 months (95% CI, 102–151). At progression, 39% of patients developed bone lesions, 30% anemia, 8.7% renal insufficiency, 27% BMPC ≥60%, 32% FLC‐ratio ≥100, and 4.3% AL amyloidosis. One patient developed hypercalcemia. Twenty‐three percent of patients progressed by FLCr ≥100 and/or BMPC ≥60% alone without evidence of CRAB features (i.e., hypercalcemia, renal insufficiency, anemia, or bone lesions). Nineteen patients (5.9%) were started on clinical trials and were censored at the date of trial treatment initiation. No SMM patients received MM‐directed therapy or therapy for SMM outside of clinical trials.

Among the 115 patients that progressed, 30 (26%) were low‐risk, 36 (31%) intermediate‐risk, and 49 (43%) high‐risk at baseline according to the 2/20/20 model. Progression rates differed significantly by baseline 2/20/20 risk group (P < 0.001); high‐risk patients had a median TTP of 30 months (95% CI, 22–49), compared to 115 months (95% CI, 70–NR) and 213 months (95% CI, 151–NR) for intermediate‐ and low‐risk patients, respectively (Figure 1 ).

We next examined patient trajectories during the first year, stratified by baseline M‐protein and FLCr quintiles (Figure 2 ). We noted heterogeneous trajectories based on baseline biomarker stratification. Among patients who progressed within the first year, some came from the highest baseline quintiles, while a significant proportion of patients had low biomarker burden at diagnosis. This heterogeneity suggests that baseline risk alone does not fully capture disease trajectory, motivating the evaluation of dynamic biomarker changes as an additional tool for risk stratification.

Dynamic biomarker analysis

Among the 227 patients who did not progress within the first year and had available biomarker measurements at the 1‐year timepoint (Supporting Information S1: Figure 1 ), we estimated changes in M‐protein and FLCr from baseline to 1 year. Patients were then stratified into quintiles based on the magnitude of the change (ΔQs) in M‐protein and FLCr over the first year. Patients in the highest quintile of M‐protein increase (ΔQ5, >0.27 g/dL) had an HR for progression of 4.0 (95% CI, 1.8–8.9) compared to ΔQ1, and those in the highest quintile of FLCr increase (ΔQ5, >58%) had an HR of 2.7 (95% CI, 1.25–6.0) compared to ΔQ1, both adjusted for baseline 2/20/20 risk score, age, and sex (Table 2A ). Progression curves and median TTP, stratified by the quintiles of the biomarker change, are presented in Supporting Information S1: Figure 2 .

After observing independent significant associations with the levels of change in M‐protein and FLCr during the first year, we next sought to develop dichotomous variables for the evolving M‐protein and evolving FLCr. Using a systematic grid search, the optimal dichotomous threshold for evolving M‐protein (eMP) was an increase of >0.38 g/dL, and for FLCr an increase of >43%. We selected these values after observing that these candidates were most frequently selected in our 50 cross‐validations. For simplicity, the numbers were rounded to eMP ≥0.4 g/dL and eFLCr ≥40%. These thresholds were independently associated with an increased risk of progression to MM, with HRs of 2.8 (95% CI, 1.5–5.5) and 2.9 (95% CI, 1.8–4.8), for eMP and eFLCr, respectively, adjusting for baseline 2/20/20 risk score, compared to patients without eMP and eFLCr (Table 2B ). A total of 31% ( N = 71) of all patients had either eMP, eFLCr, or both. The median TTP was 18 and 41 months for patients with eMP and eFLCr, respectively, compared to 130 and 201 months in patients without eMP or eFLCr. The 1‐ and 4‐year progression rate after the 1‐year landmark was 33% (95% CI, 13–49) and 75% (85% CI, 46–88) for patients with eMP compared to 6% (95% CI, 3–10) and 21% (95% CI, 14–27) for those without eMP, and 25% (95% CI, 12–35) and 54% (95% CI, 38–66) for patients with eFLCr compared to 4% (95% CI, 1–7) and 18% (95% CI, 11–24) for those without eFLCr (Figure 3 ).

Incorporating eMP and eFLCr into the 2/20/20 Model

We applied our dynamic model, incorporating eMP and eFLCr into the 2/20/20 baseline risk score to assess whether evolving biomarkers further refined risk stratification within each baseline risk category. Even among patients with the same baseline risk score, those with evolving biomarkers experienced a significantly shorter TTP, demonstrating that the dynamic model identifies meaningful heterogeneity within each 2/20/20 risk group (Figure 4 ).

Calculating from the 1‐year landmark, high‐risk patients (i.e., 2/20/20 score = 2–3) with evolving changes ( N = 20) had a 1‐ and 4‐year progression rate of 41% (95% CI, 14–59; Figure 4 ) and 83% (95% CI, 45–95), compared to 10% (95% CI, 0–22) and 45% (95% CI, 15–64) for those without evolving markers ( N = 21). For intermediate‐risk patients (i.e., 2/20/20 score = 1), 1‐year and 4‐year progression rates were 16% (95% CI, 0–29) and 57% (95% CI, 30–73) with evolving markers ( N = 26), compared to 2% (95% CI, 0–6) and 17% (95% CI, 3–28) without ( N = 46). For low‐risk patients (2/20/20 risk score = 0), the 1‐ and 4‐year progression rates were 17% (95% CI, 0–31) and 29% (95% CI, 6–47) with evolving markers ( N = 25), compared to 2% (95% CI, 0–5) and 8% (95% CI, 2–14) for patients without evolving markers ( N = 89).

The incorporation of evolving biomarkers into the 2/20/20 risk score improved the predictive performance of the model. The C‐index increased from 0.70 for baseline 2/20/20 alone, to 0.74 for 2/20/20 applied at 1 year, and to 0.79 when combining baseline 2/20/20 with evolving biomarkers. In five‐fold cross‐validation repeated ten times, incorporating the evolving biomarkers led to improved performance of the dynamic model compared to the baseline 2/20/20 model (AUC 0.782 [IQR, 0.681–0.883] vs. 0.718 [IQR, 0.619–0.814]; C‐index 0.735 [IQR, 0.698–0.776] vs. 0.700 [IQR, 0.654–0.749], Supporting Information S1: Figure 3 ).

As a sensitivity analysis, we re‐ran the model treating FLCr ≥100 as a censoring event rather than a progression event, given prior evidence from our group that this threshold may not reliably indicate imminent symptomatic progression. 18 In this analysis, eMP and eFLCr remained independently associated with progression (HR 3.2 [95% CI, 1.7–6.2] and HR 3.0 [95% CI, 1.8–5.0], respectively; C‐index 0.781).

Comparison with published evolving disease models

We applied two previously published evolving disease models to our cohort and compared their performance to our dynamic model; the Barcelona model 19 and the Gran et al. model. 20 Both showed lower discrimination than our dynamic 1‐year risk score (C‐index 0.716 and 0.72 vs. 0.788). When each model was combined with our dynamic score in a multivariable analysis, the Gran classification lost independent prognostic significance, while the Barcelona model retained borderline significance but with attenuated effect sizes. In both cases, our dynamic 1‐year risk score remained strongly and independently associated with progression across all risk levels (Supporting Information S1: Table 1 ).

DISCUSSION

In this large longitudinal study of patients with SMM, we found that while baseline risk stratification remains important, prediction of progression to MM can be further improved by incorporating evolving risk factors. We defined evolving biomarkers within the first year after diagnosis, that is, eMP as an increase of ≥0.4 g/dL, and eFLCr as an increase of ≥40%, and found that these biomarkers are independent predictors of progression from SMM to MM. When integrated with the baseline 2/20/20 model, the presence of eMP and/or eFLCr further differentiated patients at higher risk and improved predictive performance compared to the 2/20/20 model alone.

The risk of SMM progression has been assessed through several risk models, including the 2/20/20 model as well as the IMWG and the PANGEA models, which include additional cytogenetic risk factors and laboratory parameters. 13 , 15 , 21 , 22 While the baseline disease burden and progression rates in the original 2/20/20 and IMWG cohorts were higher, our cohort was similar to more recently published contemporary cohorts, with a larger proportion of patients being classified as low‐risk. 23 , 24 , 25 Furthermore, these contemporary studies have demonstrated that high‐risk by 2/20/20 is associated with lower progression rates in contemporary cohorts than originally reported (2‐year progression rates ranging 26–35% compared to ~50%), likely reflecting a wider use of advanced imaging to rule out occult bone lesions. 24 , 25 , 26 Importantly, incorporating evolving biomarkers substantially refined risk within the 2/20/20 high‐risk group: high‐risk patients with eMP and/or eFLCr had 2‐ and 5‐year progression rates from diagnosis of 41% and 83%, while high‐risk patients without evolving markers had considerably lower rates of 10% and 45%. Additionally, only 43% of the 115 patients who progressed were high‐risk at baseline by the 2/20/20 model, further underscoring that baseline risk stratification models alone are imperfect tools for accurately identifying patients with truly high‐risk disease. 2 , 16

In the seminal paper by Kyle et al. introducing MGUS, the authors observed that a subset of patients experienced a gradual rise in M‐protein levels prior to progression to MM, laying the groundwork for the “evolving” phenotype. 27 Inspired by this observation, several investigators have subsequently explored the prognostic value of evolving biomarkers on disease progression in SMM, evaluating both direct indicators of disease burden, such as M‐protein and FLCr, as well as indirect markers like hemoglobin and creatinine. 19 , 20 , 21 , 28 , 29 , 30 , 31 While evolving disease is recognized as a high‐risk marker in clinical guidelines and is occasionally used as inclusion criteria for high‐risk SMM treatment trials, 32 , 33 there is currently no consensus definition for evolving markers. Existing models vary considerably in both the threshold used to define evolution (using relative percentage increases, absolute changes, or a combination of both, with varying cut‐offs) and in the timeframes over which evolution is assessed (ranging from 6 months to 3 years) (Supporting Information S1: Table 2 ). This heterogeneity has likely limited the adoption of evolving biomarkers into routine clinical practice. Moreover, many earlier models were developed based on SMM cohorts diagnosed before the widespread use of advanced imaging to rule out osteolytic bone disease, and based on pre‐IMWG 2014 diagnosis criteria, further limiting their applicability to contemporary SMM patients. Our model builds on these prior efforts, using a large, contemporary cohort in which most patients underwent advanced imaging and were diagnosed and monitored according to the IMWG 2014 criteria. We propose clearly defined, practical definitions for evolving biomarkers (eMP ≥0.4 g/dL and eFLCr ≥40% within the first 12 months of SMM diagnosis), offering a simple, accessible, and clinically relevant tool for improved risk stratification in routine clinical practice.

When we applied two of these prior dynamic models to our cohort 19 , 20 both showed lower discrimination than our dynamic risk score, and when combining each model with our dynamic score in a multivariable model, they either lost independent prognostic significance or had significantly attenuated effect sizes, while our model remained strongly significant. While we did not directly apply the PANGEA 2.0 model 22 to our cohort, our model achieved a comparable C‐index of 0.788 using only two routinely measured serum markers assessed at a single timepoint, without requiring an online calculator or additional clinical data. Furthermore, a recent independent study by Werly et al. in a German cohort of 447 SMM patients, using a similar 12‐month assessment window, arrived at the same eMP threshold of ≥0.4 g/dL, 34 providing external validation for this cut‐off for evolving M‐protein.

Several clinical trials have investigated early treatment strategies for high‐risk SMM, including the recent AQUILA trial. 10 , 11 , 12 , 35 , 36 There are ongoing studies to evaluate more intensive approaches, such as triplet therapy followed by autologous stem cell transplantation in GEM‐CESAR ( NCT02415413 ) 37 or quadruple therapy in ASCENT ( NCT03289299 ), 38 similar to frontline strategies for newly diagnosed MM. Additional trials are also exploring novel immunotherapies, including bispecific T cell engagers ( NCT05469893 , NCT06100237 , NCT05955508 ) 32 and chimeric antigen receptor (CAR) T cells ( NCT05767359 ). 39 Despite these published and ongoing trials, the optimal approach and timing for initiating therapy in SMM remains a matter of debate. For instance, in the AQUILA trial, approximately 40% of patients in the active monitoring arm remained progression‐free at 5 years, highlighting the risk of potential overtreatment. 10 These developments underscore the critical need for models that more accurately identify high‐risk patients who are most likely to benefit from early intervention, while sparing those for whom the benefit remains uncertain. The 2/20/20 model is currently the most widely used tool to define high‐risk SMM; however, as demonstrated in several contemporary cohorts, the 2/20/20 baseline model alone is insufficient to identify truly high‐risk disease, with 2‐year progression rates below the 50% threshold. 25 , 26 In our study, incorporating evolving biomarkers into the baseline 2/20/20 model improved predictive performance, with higher C‐index and AUC. Patients classified as high‐risk by 2/20/20 who also had eMP and/or eFLCr had a 2‐year and 5‐year progression rate from diagnosis of 41% and 83%, respectively, and thus a high risk of progression to MM. On the other hand, patients with high‐risk by 2/20/20 without evolving markers exhibited a more indolent disease course with 2‐ and 5‐year progression risks of 10% and 45%. This suggests that, despite their high‐risk classification, these patients may have limited benefits from early treatment interventions.

Furthermore, we found that patients classified as low risk by the 2/20/20 model and without evidence of evolving disease exhibited a stable disease course, more resembling the clinical behavior seen in MGUS. This group may represent a substantial proportion of the SMM population. In the screened cohort from the iStopMM study, 65% met low‐risk criteria based on the 2/20/20 model. 2 Our findings suggest that further stratifying this group using evolving biomarkers can help identify patients with an indolent trajectory where de‐escalation of follow‐up intensity can be warranted, potentially reducing the burden on patients as well as healthcare resource utilization, while maintaining appropriate monitoring.

Interestingly, in the PLCO Cancer Screening Trial, analysis of serial pre‐diagnostic samples from patients who progressed from MGUS to MM revealed heterogeneous patterns of evolution: some patients showed steadily increasing M‐protein and FLC levels over five years or more before progressing to MM, while others had low‐risk MGUS as recently as 1 year prior to MM diagnosis, suggesting rapid conversion without a prolonged period of rising biomarkers. 40 This variability in progression kinetics likely reflects underlying biological differences. Through whole‐genome and next‐generation sequencing, genomic aberrations and mutational signatures have been linked to a higher risk of progression to MM. 6 , 7 , 41 , 42 , 43 , 44 However, reliable biological markers for clinical practice to clearly distinguish monoclonal gammopathy from MM remain elusive. Until genomic profiling becomes more widely available in clinical practice, dynamic changes in serum markers may serve as a surrogate for biologically active disease and underlying genomic risk.

Strengths of this study include the large, well‐defined SMM cohort with longitudinal markers and long‐term clinical follow‐up. However, the study's retrospective design and inclusion of patients from a single academic institution may lead to a certain level of selection bias. That said, this cohort was similar to several recently published cohorts in terms of demographics and baseline disease burden. 23 , 24 , 25 Furthermore, the dynamic model can only be applied after 1 year of follow‐up and is therefore unable to identify very early progressors (those who progress within the first 12 months) who may represent a biologically distinct subgroup. Although we divided the dataset into training and validation cohorts, external validation in an independent population will be important to confirm broader applicability.

In conclusion, we demonstrate that evolving M‐protein (eMP ≥0.4 g/dL) and FLC ratio (eFLCr ≥40%) within the first year of SMM diagnosis are independent predictors of progression to active MM. These practical and accessible markers improve risk assessment when integrated with the 2/20/20 model, more accurately identifying patients at high risk of progression as well as those with a low‐risk, non‐evolving disease course who may be candidates for de‐escalated monitoring. Until more precise biologic markers become widely available in clinical practice, complementing baseline risk stratification with evolving markers provides a more comprehensive assessment of SMM and helps overcome the limitations of static baseline models.

AUTHOR CONTRIBUTIONS

Theresia Akhlaghi : Conceptualization; investigation; methodology; validation; software; formal analysis; data curation; supervision; writing—review and editing; writing—original draft; visualization; project administration. David Nemirovsky : Methodology; conceptualization; software; data curation; formal analysis; visualization; validation; investigation; writing—review and editing. Kylee H. Maclachlan : Data curation; writing—review and editing. Ross S. Firestone : Writing—review and editing. Neha Korde : Writing—review and editing. Sham Mailankody : Writing—review and editing. Alexander M. Lesokhin : Writing—review and editing. Hani Hassoun : Writing—review and editing. Dhwani Patel : Writing—review and editing. Urvi A. Shah : Writing—review and editing. Carlyn R. Tan : Writing—review and editing. Heather J. Landau : Writing—review and editing. Gunjan L. Shah : Writing—review and editing. Michael Scordo : Writing—review and editing. Anish K. Simhal : Writing—review and editing. Ola Landgren : Writing—review and editing. Sergio A. Giralt : Writing—review and editing. Saad Z. Usmani : Writing—review and editing; funding acquisition; resources. Andriy Derkach : Writing—review and editing; conceptualization; methodology; software; formal analysis; validation; supervision; resources; visualization. Malin Hultcrantz : Conceptualization; methodology; supervision; resources; project administration; writing—review and editing; writing—original draft; funding acquisition; investigation.

CONFLICT OF INTEREST STATEMENT

K.M. reports grant support from ASH, MMRF, and IMS. N.K. reports research funding from Amgen, Janssen, Epizyme, AbbVie; and participates in advisory board for Janssen and receives speaking fees from Dava Oncology. S.M. received consulting fees from Evicore, Optum, BioAscend, Janssen Oncology, Bristol Myers Squibb, AbbVie, ECor1, Galapagos, and Legend Biotech. A.M.L. reports grants from Bristol Myers Squibb; personal fees, grants, and nonfinancial support from Pfizer; personal fees from Janssen and Arcellx. A.M.L. also has a patent US20150037346A1 with royalties paid. HH reports grants from Takeda and Janssen. U.A.S. reports personal fees from Physicians' Education Resources; grants and other from Celgene/Bristol Myers Squibb; other from Janssen; and grants from Parker Institute for Cancer Immunotherapy and HealthTree Foundation. C.R.T. reports researching funding from Janssen and Takeda, and consultancy fees from Janssen and Sanofi. H.L. reports research funding from Alexion Pharmaceuticals, Janssen Scientific Affairs LLC, Prothena, Protego, Nexcella, and has advisory roles for Nexcella, AbbVie, and Prothena. G.L.S. reports research funding from Janssen, Amgen, BMS, Beyond Spring, and GPCR, and serves on the Data Safety Monitoring Board for ArcellX. M.S. served as a paid consultant for McKinsey & Company, Angiocrine Bioscience, Inc., and Omeros Corporation; received research funding from Angiocrine Bioscience, Inc., Omeros Corporation, Amgen Inc., Bristol Myers Squibb, and Sanofi; served on ad hoc advisory boards for Kite–A Gilead Company and Miltenyi Biotec; and received honoraria from i3Health, Medscape, CancerNetwork for CME‐related activity, and honoraria from IDEOlogy. OL acknowledges funding from: NCI/NIH, FDA, LLS, Rising Tide Foundation, MMRF, IMF, Paula and Rodger Riney Foundation, Perelman Family Foundation, Amgen, Celgene, Janssen, Takeda, Glenmark, Seattle Genetics, Karyopharm; has received honoraria for scientific talks/participated in advisory boards for Adaptive, Amgen, Binding Site, BMS, Celgene, Cellectis, Glenmark, Janssen, Juno, Pfizer; and served on Independent Data Monitoring Committees (IDMC) for international randomized trials by: Takeda, Merck, Janssen, Theradex. S.G. reports personal fees and advisory role (scientific advisory board) from Actinium, Celgene, Bristol Myers Squibb, Sanofi, Amgen, Pfizer, GlaxoSmithKline, JAZZ, Janssen, Omeros, Takeda, and Kite. S.Z.U. reports grants and personal fees from AbbVie, Amgen, BMS, Celgene, GSK, Janssen, Merck, MundiPharma, Oncopeptides, Pharmacyclics, Sanofi, Seattle Genetics, SkylineDX, and Takeda. M.H. reports research funding from Abbvie, BeOne, BMS, Cosette Pharmaceuticals, Daiichi Sankyo, GSK, J&J, The Binding Site; and participation in scientific advisory boards for AstraZeneca, BMS, J&J, GSK, and Pfizer. The remaining authors declare no competing interests.

FUNDING

This work was supported in part by the Memorial Sloan Kettering Core Grant (P30 CA008748).

Supporting information

Supplementary Information Revised.

Contributor Information

Theresia Akhlaghi, Email: theresia.akhlaghi@gmail.com.

Malin Hultcrantz, Email: hultcram@mskcc.org.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

REFERENCES

1. Rajkumar SV, Dimopoulos MA, Palumbo A, et al. International Myeloma Working Group updated criteria for the diagnosis of multiple myeloma. Lancet Oncol. 2014;15(12):e538‐e548. 10.1016/S1470-2045(14)70442-5

2. Thorsteinsdóttir S, Gíslason GK, Aspelund T, et al. Prevalence of smoldering multiple myeloma based on nationwide screening. Nat Med. 2023;29(2):467‐472. 10.1038/s41591-022-02183-6

3. Ravindran A, Bartley AC, Holton SJ, et al. Prevalence, incidence and survival of smoldering multiple myeloma in the United States. Blood Cancer J. 2016;6(10):e486. 10.1038/bcj.2016.100

4. Kristinsson S, Holmberg E, Blimark C. Treatment for high‐risk smoldering myeloma. N Engl J Med. 2013;369(18):1762‐1765. doi:10.1056/NEJMc1310911

5. Kyle RA, Remstein ED, Therneau TM, et al. Clinical course and prognosis of smoldering (asymptomatic) multiple myeloma. N Engl J Med. 2007;356(25):2582‐2590. 10.1056/NEJMoa070389

6. Oben B, Froyen G, Maclachlan KH, et al. Whole‐genome sequencing reveals progressive versus stable myeloma precursor conditions as two distinct entities. Nat Commun. 2021;12:1861. 10.1038/s41467-021-22140-0

7. Boyle EM, Deshpande S, Tytarenko R, et al. The molecular make up of smoldering myeloma highlights the evolutionary pathways leading to multiple myeloma. Nat Commun. 2021;12:293. 10.1038/s41467-020-20524-2

8. Maura F, Bergsagel PL, Ziccheddu B, et al. Genomics Define Malignant Transformation in Myeloma Precursor Conditions. J Clin Oncol. 2026;44(3):188‐199. 10.1200/JCO-25-01733

9. Maura F, Bolli N, Rustad EH, Hultcrantz M, Munshi N, Landgren O. Moving from cancer burden to cancer genomics for smoldering myeloma: a review. JAMA Oncol. 2020;6(3):425‐432. 10.1001/jamaoncol.2019.4659

10. Dimopoulos MA, Voorhees PM, Schjesvold F, et al. Daratumumab or active monitoring for high‐risk smoldering multiple myeloma. N Engl J Med. 2025;392:1777‐1788. 10.1056/NEJMoa2409029

11. Lonial S, Jacobus S, Fonseca R, et al. Randomized trial of lenalidomide versus observation in smoldering multiple myeloma. J Clin Oncol. 2020;38(11):1126‐1137. 10.1200/JCO.19.01740

12. Mateos MV, Hernández MT, Giraldo P, et al. Lenalidomide plus dexamethasone for high‐risk smoldering multiple myeloma. N Engl J Med. 2013;369(5):438‐447. 10.1056/NEJMoa1300439

13. Lakshman A, Vincent Rajkumar S, Buadi FK, et al. Risk stratification of smoldering multiple myeloma incorporating revised IMWG diagnostic criteria. Blood Cancer J. 2018;8:59. 10.1038/s41408-018-0077-4

14. Pérez‐Persona E, Vidriales MB, Mateo G, et al. New criteria to identify risk of progression in monoclonal gammopathy of uncertain significance and smoldering multiple myeloma based on multiparameter flow cytometry analysis of bone marrow plasma cells. Blood. 2007;110:2586‐2592. 10.1182/blood-2007-05-088443

15. Mateos MV, Kumar S, Dimopoulos MA, et al. International Myeloma Working Group risk stratification model for smoldering multiple myeloma (SMM). Blood Cancer J. 2020;10:102. 10.1038/s41408-020-00366-3

16. Hill E, Dew A, Morrison C, et al. Assessment of discordance among smoldering multiple myeloma risk models. JAMA Oncol. 2021;7(1):132. 10.1001/jamaoncol.2020.5585

17. Kumar SK, Callander NS, Adekola K, et al. Multiple myeloma, Version 2.2024, NCCN clinical practice guidelines in oncology. J Natl Compr Cancer Netw. 2023;21(12):1281‐1301. 10.6004/jnccn.2023.0061

18. Akhlaghi T, Maclachlan K, Korde N, et al. Evaluating serum free light chain ratio as a biomarker in multiple myeloma. Haematologica. 2025;110(2):493‐497. 10.3324/haematol.2024.285531

19. de Daniel A, Rodríguez‐Lobato LG, Tovar N, et al. The evolving pattern of the monoclonal protein improves the IMWG 2/20/20 classification for patients with smoldering multiple myeloma. HemaSphere. 2024;8(5):e76. 10.1002/hem3.76

20. Gran C, Luong V, Bruchfeld JB, et al. Dynamic follow‐up of smoldering multiple myeloma identifies a subset of patients at high risk of progression. Am J Hematol. 2021;96(3):E63‐E65. 10.1002/ajh.26062

21. Cowan A, Ferrari F, Freeman SS, et al. Personalised progression prediction in patients with monoclonal gammopathy of undetermined significance or smouldering multiple myeloma (PANGEA): a retrospective, multicohort study. Lancet Haematol. 2023;10(3):e203‐e212. 10.1016/S2352-3026(22)00386-6

22. Chabrun DE, Schwartz F, Gentile S, et al. Enhanced dynamic risk stratification of smoldering multiple myeloma. Nat Med. 2026;32:1745‐1753. 10.1038/s41591-026-04304-x

23. Werly A, Hampel M, Hielscher T, et al. Patterns of progression in a contemporary cohort of 447 patients with smoldering multiple myeloma. Blood Cancer J. 2024;14(1):176. 10.1038/s41408-024-01159-8

24. Mellgard G, Gilligan M, Cliff ERS, et al. Risk stratification models overestimate progression risk in contemporary patients with smoldering multiple myeloma. HemaSphere. 2024;8(3):e61. 10.1002/hem3.61

25. Kastritis E, Solia I, Malandrakis P, et al. Patterns of progression among 427 smoldering myeloma patients diagnosed after 2014: importance of monitoring. Blood Adv. 2025:4444‐4447. 10.1182/bloodadvances.2025016083

26. Akhlaghi T, Maclachlan KH, Wiggins A, Usmani SZ, Derkach A, Hultcrantz M. Prognosis and patterns of progression in smoldering multiple myeloma. Blood Adv. 2026;10(4):1344‐1347. 10.1182/bloodadvances.2025018021

27. Kyle RA. Monoclonal gammopathy of undetermined significance. Natural history in 241 cases. Am J Med. 1978;64(5):814‐826. 10.1016/0002-9343(78)90522-3

28. Rosiñol L, Bladé J, Esteve J, et al. Smoldering multiple myeloma: natural history and recognition of an evolving type. Br J Haematol. 2003;123(4):631‐636. 10.1046/j.1365-2141.2003.04654.x

29. Ravi P, Kumar S, Larsen JT, et al. Evolving changes in disease biomarkers and risk of early progression in smoldering multiple myeloma. Blood Cancer J. 2016;6(7):e454. 10.1038/BCJ.2016.65

30. Fernández de Larrea C, Isola I, Pereira A, et al. Evolving M‐protein pattern in patients with smoldering multiple myeloma: Impact on early progression. Leukemia. 2018;32(6):1427‐1434. 10.1038/s41375-018-0013-4

31. Wu V, Moshier E, Leng S, et al. Risk stratification of smoldering multiple myeloma: predictive value of free light chains and group‐based trajectory modeling. Blood Adv. 2018;2(12):1470‐1479. 10.1182/bloodadvances.2018016998

32. Nadeem O, Magidson S, Midha S, et al. Immuno‐PRISM: a randomized phase II platform study of bispecific antibodies in high‐risk smoldering myeloma. Blood. 2023;142(Suppl 1):206. 10.1182/blood-2023-177954

33. Hill E, Roswarski JL, Bhaskarla A, et al. Fixed duration combination therapy with carfilzomib, lenalidomide, and dexamethasone followed by lenalidomide maintenance leads to high rates of sustained MRD negativity in patients with high‐risk smoldering multiple myeloma: long term follow up of an Investigator Initiated Phase 2 Trial. Blood. 2023;142(Suppl 1):337. 10.1182/blood-2023-181811

34. Werly A, Hielscher T, Schmidt S, et al. A dynamic risk stratification model for patients with smoldering multiple myeloma through integration of the evolving patterns of monoclonal protein and serum free light chains. Clin Lymphoma Myeloma Leuk . 2025;25:S250‐S251.

35. Mateos MV, Hernández MT, Giraldo P, et al. Lenalidomide plus dexamethasone versus observation in patients with high‐risk smouldering multiple myeloma (QuiRedex): long‐term follow‐up of a randomised, controlled, phase 3 trial. Lancet Oncol. 2016;17(8):1127‐1136. 10.1016/S1470-2045(16)30124-3

36. Landgren CO, Chari A, Cohen YC, et al. Daratumumab monotherapy for patients with intermediate‐risk or high‐risk smoldering multiple myeloma: a randomized, open‐label, multicenter, phase 2 study (CENTAURUS). Leukemia. 2020;34(7):1840‐1852. 10.1038/s41375-020-0718-z

37. Mateos MV, Martínez‐López J, Rodríguez‐Otero P, et al. Curative strategy (GEM‐CESAR) for high‐risk smoldering myeloma (SMM): post‐hoc analysis of sustained undetectable measurable residual disease (MRD). Blood. 2022;140(Suppl 1):292‐294. 10.1182/blood-2022-159606

38. Kumar SK, Alsina M, Laplant B, et al. Fixed duration therapy with daratumumab, carfilzomib, lenalidomide and dexamethasone for high risk smoldering multiple myeloma‐results of the ascent trial. Blood. 2022;140(Suppl 1):1830‐1832. 10.1182/blood-2022-168930

39. Nadeem O, Cordas dos Santos DM, Nikiforow S, et al. Ciltacabtagene autoleucel in high‐risk smoldering multiple myeloma: the CAR‐PRISM phase 2 trial. Nat Med. 2026;32:2420‐2429. 10.1038/s41591-026-04365-y

40. Landgren O, Hofmann JN, McShane CM, et al. Association of immune marker changes with progression of monoclonal gammopathy of undetermined significance to multiple myeloma. JAMA Oncol. 2019;5(9):1293‐1301. 10.1001/jamaoncol.2019.1568

41. Maura F, Petljak M, Lionetti M, et al. Biological and prognostic impact of APOBEC‐induced mutations in the spectrum of plasma cell dyscrasias and multiple myeloma cell lines. Leukemia. 2018;32(4):1043‐1047. 10.1038/leu.2017.345

42. Bustoros M, Sklavenitis‐Pistofidis R, Park J, et al. Genomic profiling of smoldering multiple myeloma identifies patients at a high risk of disease progression. J Clin Oncol. 2020;38:2380‐2389. 10.1200/JCO.20.00437

43. Bolli N, Maura F, Minvielle S, et al. Genomic patterns of progression in smoldering multiple myeloma. Nat Commun. 2018;9(1):3363. 10.1038/s41467-018-05058-y

44. Bustoros M, Anand S, Sklavenitis‐Pistofidis R, et al. Genetic subtypes of smoldering multiple myeloma are associated with distinct pathogenic phenotypes and clinical outcomes. Nat Commun. 2022;13(1):3449. 10.1038/s41467-022-30694-w

Associated Data

Supplementary Materials

Supplementary Information Revised.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.