Non‐transplant eligible patients with newly diagnosed multiple myeloma (NTE‐NDMM) who are categorized as frail according to the International Myeloma Working Group Frailty Index (IMWG‐FI) show pronounced heterogeneity in clinical characteristics, leading to differences in frailty scores—ranging from 2 to 5—based on the type and number of geriatric impairments and comorbidities. As a result, clinical outcomes may differ markedly within this subgroup. 1 Indeed, a post hoc analysis of the HOVON‐123 trial, in which patients were treated with nine cycles of dose‐adjusted melphalan, prednisone, and bortezomib, revealed that the overall survival (OS) of frail patients with an IMWG‐FI of 2 had an OS comparable to intermediate‐fit patients and superior to frail patients with scores of 3–5. In addition, treatment discontinuation after 9 months was considerably higher in patients with higher frailty scores. 2 Accordingly, in the IFM 2017‐03 trial, in which daratumumab–lenalidomide was compared with lenalidomide–dexamethasone, progression free survival (PFS) was significantly prolonged in patients with a frailty score of 2 compared to those with scores of 4–5. 3 Notably, a similar frailty score may capture distinct aging phenotypes: for example a score of 2 may solely result from advanced age or may reflect significant geriatric impairments in a younger individual. To address this heterogeneity, frail patients may be more rationally stratified based on the underlying drivers of frailty, yielding three frail subgroups: frail‐by‐age (>80 years, no comorbidities or impairments in [I]ADL), frail‐by‐impairments (≤80 years, with comorbidities and/or impairments in [I]ADL), and ultra‐frail (>80 years, with comorbidities and/or impairments in [I]ADL). The prognostic impact of these three distinct frailty subgroups on OS has been variably reported. In the HOVON‐143 trial, in which patients were treated with nine cycles of ixazomib, daratumumab, and dexamethasone (IDd) followed by a maintenance phase of IDd for a maximum of 2 years, patients classified frail‐by‐age‐alone demonstrated superior OS, although the difference did not reach statistical significance, likely due to limited sample size. 4 , 5 In contrast, a retrospective analysis of the original IMWG‐FI patient cohort reported comparable OS between patients classified frail‐by‐age‐alone and all other frail patients. 6 To resolve this inconsistency, we examined the impact of frailty subtype on OS in a larger group size by pooling data of two prospective HOVON trials (HOVON‐123 and HOVON‐143) in NTE‐NDMM patients. The studies were conducted in accordance with the Declaration of Helsinki and were approved by the institutional review board and ethics committees before study initiation. All patients provided written informed consent.
The subsets of frail NTE‐NDMM patients from the HOVON‐123 and HOVON‐143 studies were analyzed. Study details have been published before. 4 , 7 Patient‐ and disease characteristics were compared between the three different frail subgroups by the Wilcoxon rank‐sum tests. Survival outcomes were compared using Cox proportional hazard models and visualized with the Kaplan–Meier method. To account for potential confounding factors, PFS, PFS2, and OS were corrected for baseline disease characteristics that were identified as significant predictors for outcomes in multivariable analysis. Further details on statistical analyses are described in Supplemental Methods .
In total, 202 frail patients were included in this analysis: 33 (16%) were frail‐by‐age‐alone, 94 (47%) were frail‐by‐impairments, and 75 (37%) were ultra‐frail. Ultra‐frail patients exhibited significantly lower albumin levels (<35 g/L; 74%), compared to those frail‐by‐impairments (52%) and frail‐by‐age‐alone (55%) (P < 0.01). Compared to frail‐by‐age‐alone patients, both ultra‐frail and frail‐by‐impairments subgroups more frequently presented with poor performance score (WHO performance status [WHO] ≥ 2; 9%, 49%, and 43% respectively, P < 0.01), elevated beta‐2‐microglobuline (B2M) levels (≥5.5 mg/L; 24%, 55%, and 50%, respectively, P = 0.03), International Staging System (ISS) Stage III (24%, 55%, and 50%, respectively, P = 0.02), and Revised‐ISS (R‐ISS) Stage III (9%, 20%, and 20%, respectively, P = 0.02). Lactate dehydrogenase (LDH) levels did not differ significantly among frail subgroups (Table 1 ).
Table 1
Baseline patient and disease characteristics.
| Frail‐by‐age‐alone | Frail‐by‐impairments | Ultra‐frail | P‐value | |
| Number of patients | 33 (16) | 94 (47) | 75 (37) | |
| Age | <0.001 | |||
| ≤75 | ‐ | 10 (11) | ‐ | |
| 76 to ≤80 | ‐ | 84 (89) | ‐ | |
| >80 | 33 (100) | ‐ | 75 (100) | |
| ADL | <0.001 | |||
| Independent | 33 (100) | 67 (71) | 39 (52) | |
| Dependent | ‐ | 25 (27) | 32 (43) | |
| Unknown | ‐ | 2 (2) | 4 (5) | |
| IADL | <0.001 | |||
| Independent | 33 (100) | 32 (34) | 18 (24) | |
| Dependent | ‐ | 55 (59) | 48 (64) | |
| Unknown | ‐ | 7 (7) | 9 (12) | |
| WHO | <0.001 | |||
| 0–1 | 30 (91) | 54 (57) | 38 (51) | |
| ≥2 | 3 (9) | 40 (43) | 37 (49) | |
| CCI | <0.001 | |||
| ≤1 | 33 (100) | 37 (39) | 28 (37) | |
| ≥2 | ‐ | 57 (61) | 47 (63) | |
| Number of patients | 33 (16) | 94 (47) | 75 (37) | |
| Kreatinine (mmol/L) | 90 (78–129) | 98 (77–147) | 123 (82–152) | 0.11 |
| Albumine (g/L) | 34 (32–36) | 34 (29–38) | 32 (29–34) | 0.044 |
| Missing | ‐ | 1 | 3 | |
| Albumine | 0.006 | |||
| ≥35 g/L | 15 (45) | 44 (47) | 17 (23) | |
| <35 g/L | 18 (55) | 49 (52) | 44 (74) | |
| Missing | ‐ | 1 (1) | 3 (4) | |
| B2M (mg/L) | 4.5 (3.3–6.0) | 5.6 (3.9–8.4) | 6.2 (3.6–9.4) | 0.2 |
| Missing | 4 | 3 | 2 | |
| B2M | 0.032 | |||
| <3.5 mg/L | 9 (27) | 17 (18) | 17 (23) | |
| ≥3.5 and <5.5 mg/L | 12 (36) | 27 (29) | 15 (20) | |
| ≥5.5 mg/L | 8 (24) | 47 (50) | 41 (55) | |
| Missing | 4 (12) | 3 (3) | 2 (3) | |
| Calcium (mmol/L) | 2.45 (2.39–2.51) | 2.45 (2.31–2.60) | 2.51 (2.37–2.65) | 0.3 |
| Hemoglobin (mmol/L) | 6.1 (5.6–6.8) | 6.4 (5.7–7.0) | 6.0 (5.7–6.7) | 0.5 |
| LDH | >0.9 | |||
| Normal | 29 (88) | 78 (83) | 60 (80) | |
| Elevated | 4 (12) | 14 (15) | 13 (17) | |
| Missing | ‐ | 2 (2) | 2 (3) | |
| Cytogenetic risk | 0.8 | |||
| Normal | 23 (70) | 62 (66) | 50 (67) | |
| High risk | 4 (12) | 18 (19) | 11 (15) | |
| Missing | 6 (18) | 14 (15) | 14 (19) | |
| ISS stage | 0.022 | |||
| I | 8 (24) | 11 (12) | 6 (8) | |
| II | 13 (39) | 33 (35) | 26 (35) | |
| III | 8 (24) | 47 (50) | 41 (55) | |
| Missing | 4 (12) | 3 (3) | 2 (3) | |
| R‐ISS stage | 0.038 | |||
| I | 7 (21) | 7 (7) | 3 (4) | |
| II | 16 (48) | 60 (64) | 47 (63) | |
| III | 3 (9) | 19 (20) | 15 (20) | |
| Missing | 7 (21) | 8 (9) | 10 (13) |
No significant trial effect was observed for PFS (P = 0.96), PFS2 (P = 0.51), or OS (P = 0.99), allowing a pooled analysis. Median PFS was 12.7 months (95% CI: 11.5–17.2) in ultra‐frail patients, 16.5 months (95% CI: 12.9–21.9) in those frail‐by‐impairments, and 21.2 months (95% CI: 15.9–28.6) in the frail‐by‐age‐alone subgroup. In both uni‐ and multivariate analyses, a composite of LDH, B2M, and albumin emerged as the strongest predictor for PFS. Median PFS did not differ significantly between the three frail subgroups even after adjustment for LDH, B2M, and albumin (Figure S1 ; Tables S1–S3 ).
Median PFS2 was significantly shorter in ultra‐frail patients (22.4 months; 95% CI: 17.2–31.6), as compared to those frail‐by‐impairments (31.4 months; 95% CI: 24.5–39.5) and the frail‐by‐age‐alone subgroup (40.0 months; 95% CI: 31.8–53.2). In both uni‐ and multivariate analyses, B2M levels were the strongest predictor for PFS2. After adjusting for B2M levels, ultra‐frail patients—who had the highest B2M levels—continued to show significantly shorter PFS2, compared to patients frail‐by‐impairments (hazard ratio [HR] 1.19, 95% CI 1.01–1.41, P = 0.04), but not compared to the frail‐by‐age‐alone group. There was no difference in PFS2 between patients frail‐by‐age and patients frail‐by‐impairments (Figure S2 ; Tables S1, S2 , and S4 ).
Median OS was significantly shorter in ultra‐frail patients compared to both other subgroups: 23.7 months (95% CI: 19.1–35.5) in ultra‐frail patients, 38.2 months (95% CI: 30.7–51.5) in patients frail‐by‐impairments, and 49.0 months (95% CI: 38.4–62.1) in patients‐frail‐by‐age. In both uni‐ and multivariate analyses, B2M and albumin levels were the strongest predictors for OS. After adjusting for B2M and albumin levels, OS remained significantly shorter in ultra‐frail patients, compared to those frail‐by‐impairments (HR 1.28, 95% CI 1.07–1.52, P < 0.01) and patients frail‐by‐age (HR 1.92, 95%CI 1.14–3.22, P = 0.01), indicating that frailty subgroups are independently associated with OS. There was no difference in OS between patients frail‐by‐age and patients frail‐by‐impairments (Figure 1 ; Tables S1, S2 , and S5 ).
Notably, ultra‐frail patients experienced the highest early mortality (within 2 months of treatment initiation: 8/75; 11%), as compared to those frail‐by‐impairments (2/94; 2%) and those frail‐by‐age (1/33; 3%) (P < 0.01). Among ultra‐frail patients, early deaths were primarily due to progressive disease (3/8; 38%), followed by infection (2/8; 25%), toxicity (1/8, 13%), and multifactorial causes (2/8; 25%) (Table S6 ). In addition, ultra‐frail patients discontinued treatment within nine cycles (60%) more often, compared to patients frail‐by‐impairments (37%) and patients frail‐by‐age (36%) (P < 0.01) (Table S7 ). Treatment completion was lowest among ultra‐frail patients, with only 27% completing the full protocol, compared to 45%–47% in the other frail subgroups (Table S8 ). Lastly, ultra‐frail patients experienced the highest incidence of Grade ≥3 non‐hematologic and hematologic adverse events (Table S9 ).
Overall, 121 out of 202 (60%) patients received second‐line treatment. Ultra‐frail patients were less likely to receive second‐line treatment (33/75, 44%), compared to patients frail‐by‐impairments (64/94, 68%) and patients frail‐by‐age (24/33, 73%) (Table S10 ).
Finally, we assessed the dynamics of IMWG frailty scores across frail subgroups. Of note, this analysis is hampered by the fact that due to high rates of early treatment discontinuation, dynamic frailty could only be assessed in 20/75 (27%) of ultra‐frail patients, in 45/94 (48%) of patients frail‐by‐impairments, and in 17/33 (52%) of patients frail‐by‐age‐alone. Improvement of frailty score occurred in 10/20 (50%) of ultra‐frail patients and in 19/45 (42%) of patients frail‐by‐impairments; patients who were frail based on age alone could not improve (due to age > 80 being a constant factor). Deterioration in frailty occurred in 4/17 (24%) ultra‐frail patients (excluding those with a frailty score of 5), in 10/45 (22%) of patients frail‐by‐impairments, and in 3/17 (18%) of patients frail‐by‐age‐alone.
Direct comparison of the frailty subgroup analysis with available literature is limited by several factors. First, not all studies employed the IMWG‐FI but used the Simplified‐FI instead, which we previously showed, misclassifies over 20% of patients as frail rather than intermediate‐fit, complicating cross‐trial comparison. 1 , 8 , 9 Second, a standardized definition of “ultra‐frail” patients has not been established yet. In one study, ultra‐frailty was defined as an S‐FI score ≥ 3, based solely on cumulative impairments rather than a combination of age and functional deficits, and outcomes for this subgroup were not reported. 10 Third, to determine outcomes of frailty subgroups, comparisons should be harmonized. For example, in the original IMWG cohort used to define frailty, patients who were frail‐by‐impairments and ultra‐frail patients were grouped together, preventing specific evaluation of ultra‐frail patients. 6 To our knowledge, this is the first large‐scale study to report detailed clinical outcomes for IMWG‐FI–defined frailty subgroups stratified by the underlying causes of frailty. Validation of the proposed definition of frailty subgroups in an external cohort is warranted.
The ultimate aim is to develop a frailty score that is both clinically informative and easy to implement, enabling identification of patient subgroups with distinct outcomes. Our subgroup classification, with superior prognostic value, builds upon the gold standard IMWG‐FI and requires no additional data, which enables easy implementation in the clinic. This does not account for several other scores, either implementing additional testing of organ function, laboratory parameters, or disease characteristics for which invasive bone marrow investigations are required. 11 , 12 , 13 Although further simplification would be desirable, we showed that the S‐FI, incorporating physician‐reported WHO‐PS instead of patient‐reported (i) ADL, overestimates frailty. It would be interesting to investigate whether subclassification could overcome these concerns. 9
Although frailty subgroup analysis revealed prognostic value, it remains unknown whether frailty subgroup‐adjusted treatment improves clinical outcome. At present, frailty‐adjusted treatment strategies are largely guided by expert opinions. 14 In oncology, a systematic review and meta‐analysis of clinical trials showed that in older patients, a “Start low, go slow” approach improved treatment completion, reduced toxicity, and led to a superior survival as compared to standard unadjusted treatment. 15 The FiTNEss trial, investigating NTE‐NDMM patients, incorporated a “Start low, go slow” approach in a randomized trial, comparing standard (reactive) treatment with ixazomib, lenalidomide, and dexamethasone (IRd) versus frailty‐adjusted (adaptive) treatment with IRd. 16 , 17 Surprisingly, early treatment discontinuation within 60 days of randomization was significantly reduced only in intermediate‐fit patients receiving adaptive treatment, but not in frail patients, where the greatest benefit had been anticipated. This may have been caused by pre‐emptive dose reductions evident in the reactive group. 16 However, there was an event‐free and OS benefit with frailty‐adjusted treatment. Whether further refinement of the frailty classification will improve treatment strategies remains to be shown.
In conclusion, our findings demonstrate that IMWG‐FI–defined frail patients represent a clinically heterogeneous population with markedly divergent outcomes. In particular, ultra‐frail patients experienced poor prognosis, with a median PFS of just over 1 year and a median OS of approximately 2 years. These results underscore the importance of incorporating frailty subgroups into the design and interpretation of future studies to pave the way for more tailored and effective treatment strategies for this vulnerable population.
AUTHOR CONTRIBUTIONS
Kazimierz Groen : Conceptualization; investigation; writing—original draft; formal analysis. Febe Smits : Conceptualization; investigation; writing—original draft; formal analysis; visualization. Kazem Nasserinejad : Conceptualization; investigation; methodology; formal analysis; writing—review and editing. Mark‐David Levin : Writing—review and editing; conceptualization. Josien C. Regelink : Writing—review and editing. Gert‐Jan Timmers : Writing—review and editing. Esther G. M. de Waal : Writing—review and editing. Matthijs Westerman : Writing—review and editing. Gerjo A. Velders : Writing—review and editing. Koen de Heer : Writing—review and editing. Rineke B. L. Leys : Writing—review and editing. Roel J. W. van Kampen : Writing—review and editing. Claudia A. M. Stege : Writing—review and editing. Maarten R. Seefat : Writing—review and editing. Inger S. Nijhof : Writing—review and editing. Ellen van der Spek : Writing—review and editing. Saskia K. Klein : Writing—review and editing; conceptualization. Niels W. C. J. van de Donk : Writing—review and editing. Paula F. Ypma : Writing—review and editing. Sonja Zweegman : Conceptualization; investigation; funding acquisition; writing—original draft; supervision; writing—review and editing; methodology; formal analysis.
CONFLICT OF INTEREST STATEMENT
K. Groen: payment/honoraria for presentations from Bristol Myers Squibb and BeiGene (no personal funding). F. Smits: no conflicts of interest. K. Nasserinejad: no conflicts of interest. M.‐D. Levin: support for attending meetings and/or travel: Janssen, Takeda. J. C. Regelink: no conflicts of interest. G.‐J. Timmers: participation on an advisory board: Novartis; travel, accommodations, expenses: Novartis, Janssen. E. G. M. de Waal: no conflicts of interest. M. Westerman: no conflicts of interest. G. A. Velders: no conflicts of interest. K. de Heer: no conflicts of interest. R. B. L. Leys: no conflicts of interest. R. J. W. van Kampen: no conflicts of interest. C. A. M. Stege: speaker's bureau: Sanofi, Celgene/Bristol Myers Squibb, and Takeda; consulting or advisory role: Sanofi, Janssen. M. R. Seefat: no conflicts of interest. I. S. Nijhof: payment or honoraria for lectures, presentations, or educational events: Janssen, Celgene/Bristol Myers Squibb, and Sanofi. E. van der Spek: payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events: Janssen. S. K. Klein: no conflicts of interest. N. W. C. J. van de Donk: consulting or advisory role: Janssen, Celgene, Bristol Myers Squibb, Novartis, Amgen, Servier, Takeda, Bayer, Roche, Pfizer, AbbVie, and Adaptive (no personal funding); research funding: Janssen, Celgene, Amgen, Novartis, Bristol Myers Squibb, and Cellectis. P. F. Ypma: payment or honoraria for presentations: Janssen; support for attending meetings and/or travel: Janssen. S. Zweegman: consulting or advisory role: Janssen‐Cilag, Takeda, Celgene/Bristol Myers Squibb, Sanofi, and Oncopeptides (no personal funding); research funding: Janssen, Takeda.
ETHICS STATEMENT
The studies were conducted in accordance with the Declaration of Helsinki and were approved by the institutional review board and ethics committees before study initiation. All patients provided written informed consent.
FUNDING
Janssen Pharmaceuticals, H143 Grant (ICD contract record number 590049); KWF Kankerbestrijding, Grant H143 (project number VU 2013‐6411); Takeda Pharmaceutical Company, H143 Grant.
Supporting information
Supporting Information.
ACKNOWLEDGMENTS
The authors would like to thank all participating patients and their families, and the HOVON data center.
DATA AVAILABILITY STATEMENT
Data will be available on request. The request will be reviewed by the Principal Investigator, the HOVON working group, and the HOVON board.
REFERENCES
1 Palumbo A , Bringhen S , Mateos MV , et al. Geriatric assessment predicts survival and toxicities in elderly myeloma patients: an International Myeloma Working Group report . Blood . 2015 ; 125 ( 13 ): 2068 ‐ 2074 . 25628469 10.1182/blood-2014-12-615187 PMC4375104
2 Stege CAM , Nasserinejad K , Klein SK , et al. Improving the identification of frail elderly newly diagnosed multiple myeloma patients . Leukemia . 2021 ; 35 ( 9 ): 2715 ‐ 2719 . 33589752 10.1038/s41375-021-01162-z
3 Manier S . The IFM2017‐03 Phase 3 Trial: A Dexamethasone Sparing‐Regimen With Daratumumab and Lenalidomide for Frail Patients With Newly‐Diagnosed Multiple Myeloma . ASH ; 2024 . 10.1016/S1470-2045(25)00280-3 41038184
4 Stege CAM , Nasserinejad K , van der Spek E , et al. Ixazomib, daratumumab, and low‐dose dexamethasone in frail patients with newly diagnosed multiple myeloma: the Hovon 143 study . J Clin Oncol . 2021 ; 39 ( 25 ): 2758 ‐ 2767 . 33945289 10.1200/JCO.20.03143
5 Smits F . Frailty subgroups determine heterogeneous outcomes in elderly patients with newly‐diagnosed multiple myeloma—long‐term follow‐up of the HOVON 143 trial . In: Zweegman S , ed. 66th ASH Annual Meeting Abstracts: Blood . 2024 : 4739 .
6 D'Agostino M , Larocca A , Offidani M , et al. Octogenarian newly diagnosed multiple myeloma patients without geriatric impairments: the role of age >80 in the IMWG frailty score . Blood Cancer J . 2021 ; 11 ( 4 ): 73 . 33846296 10.1038/s41408-021-00464-w PMC8041817
7 Zweegman S . Feasibility and efficacy of dose adjusted melphalan–prednisone–bortezomib (MPV) in elderly patients ≥ 75 years of age with newly diagnosed multiple myeloma; the non‐randomised phase II HOVON 123 study . Blood . 2016 .
8 Facon T , Dimopoulos MA , Meuleman N , et al. A simplified frailty scale predicts outcomes in transplant‐ineligible patients with newly diagnosed multiple myeloma treated in the FIRST (MM‐020) trial . Leukemia . 2020 ; 34 ( 1 ): 224 ‐ 233 . 31427722 10.1038/s41375-019-0539-0 PMC7214253
9 Groen K , Smits F , Nasserinejad K , et al. Assessing frailty in myeloma: the pursuit of simplicity may sacrifice precision of predicting clinical outcomes . HemaSphere . 2024 ; 8 ( 7 ): e85 . 38966208 10.1002/hem3.85 PMC11223651
10 Mian H , Martin TG , Pond GR , et al. Outcomes of frailty subgroups of older adults (age >/= 70) treated with teclistamab: an International Myeloma Foundation immunotherapy database real‐world analysis . Leukemia . 2025 ; 39 : 1252 ‐ 1255 . 40113913 10.1038/s41375-025-02552-3
11 Engelhardt M , Domm AS , Dold SM , et al. A concise revised Myeloma Comorbidity Index as a valid prognostic instrument in a large cohort of 801 multiple myeloma patients . Haematologica . 2017 ; 102 ( 5 ): 910 ‐ 921 . 28154088 10.3324/haematol.2016.162693 PMC5477610
12 Cook G , Royle KL , Pawlyn C , et al. A clinical prediction model for outcome and therapy delivery in transplant‐ineligible patients with myeloma (UK Myeloma Research Alliance Risk Profile): a development and validation study . Lancet Haematol . 2019 ; 6 ( 3 ): e154 ‐ e166 . 30738834 10.1016/S2352-3026(18)30220-5 PMC6391517
13 Scheubeck S , Ihorst G , Schoeller K , et al. Comparison of the prognostic significance of 5 comorbidity scores and 12 functional tests in a prospective multiple myeloma patient cohort . Cancer . 2021 ; 127 ( 18 ): 3422 ‐ 3436 . 34061991 10.1002/cncr.33658
14 Zweegman S , Engelhardt M , Larocca A . Elderly patients with multiple myeloma: towards a frailty approach? Curr Opin Oncol . 2017 ; 29 ( 5 ): 315 ‐ 321 . 28763310 10.1097/CCO.0000000000000395
15 Aleixo G , Patel T , Ani J , et al. “Start low, go slow,” a strategy to tailor treatment dosing in older or vulnerable adults with advanced solid cancer: a systematic review and meta‐analysis . J Geriatr Oncol . 2025 ; 16 ( 2 ): 102153 . 39537457 10.1016/j.jgo.2024.102153
16 Coulson AB , Royle KL , Pawlyn C , et al. Frailty‐adjusted therapy in Transplant Non‐Eligible patients with newly diagnosed Multiple Myeloma (FiTNEss (UK‐MRA Myeloma XIV Trial)): a study protocol for a randomised phase III trial . BMJ Open . 2022 ; 12 ( 6 ): e056147 . 10.1136/bmjopen-2021-056147 PMC9163533 35654466
17 Cook G . IMWG Frailty Score‐Adjusted Therapy Delivery Reduces the Early Mortality Risk in Newly Diagnosed Multiple Myeloma: Results of the UK Myeloma Research Alliance (UK‐MRA) Myeloma XIV Fitness Trial . ASH ; 2024 .