Разработка и проверка новой хрупкой модели для пациентов с недавно поставленным диагнозом множественной миэломы.

Introduction

Multiple myeloma (MM) predominantly affects elderly individuals, and its incidence increases with age. Data from the 2017 Surveillance, Epidemiology, and End Results database indicated an incidence rate of 8.2 per 100,000 among individuals aged 50–54 years, rising to 50.6 per 100,000 among those aged 80–84 [ 1 , 2 ]. As the demographic shifts towards an older population, the proportion of elderly patients with MM increases, with an estimated 75% of newly diagnosed cases expected to occur in individuals aged 64–84 years between 2032 and 2034 [ 3 ]. The predominance of older adults within the MM population contributes to a heightened prevalence of frailty, with at least 30% of patients exhibiting frailty at diagnosis [ 4 ]. Frailty is a non-specific clinical condition characterised by a decline in physiological reserve, resulting in increased vulnerability and diminished stress resistance. Frailty assessment involves multiple dimensions of factors, including patient-related factors, disease-related factors and socioeconomic factors [ 5 – 7 ]. Furthermore, the frailty status of patients exhibits significant dynamic changes, influenced by disease progression, therapeutic responses, functional recovery, and supportive care interventions. Clinical assessments must be conducted throughout the disease trajectory to provide timely opportunities for personalised interventions [ 8 ]. Compared with their healthier counterparts, frail patients demonstrate reduced drug tolerance, increased susceptibility to adverse drug reactions, and higher rates of treatment interruption and discontinuation, culminating in a diminished survival duration [ 9 , 10 ].

In recent years, considerable efforts have been made to develop comprehensive models for assessing frailty in MM patients. A significant contribution is the frailty assessment model established by the International Myeloma Working Group (IMWG) in 2015, which incorporates age, activities of daily living (ADL), instrumental activities of daily living, and the Charlson Comorbidity Index (CCI) [ 11 ]. The Mayo frailty model (2016) included three parameters: Eastern Cooperative Oncology Group performance status (ECOG-PS), age, and N-terminal natriuretic peptide type B [ 12 ]. The simplified frailty scale (2019) comprises three indicators: age, ECOG-PS, and CCI [ 13 ]. In 2021, Lee et al. developed a novel simplified frailty model (Smodel) for the Asian population, selecting indicators consistent with the simplified frailty scale. This model categorises patients into three groups: fit (15.0%), intermediate fit (51.6%), and frail (33.4%), with significant differences observed in OS and event-free survival across these groups [ 14 ]. These models have effectively identified patients with poor prognoses within their respective cohorts, with both the IMWG frailty model and the simplified frailty scale also predicting adverse events (AEs) in patients.

Despite the availability of several frailty models capable of predicting prognosis and AEs, there remains no consensus on the standard assessment of frailty in patients with MM or how clinicians should adjust drug regimens and dosages based on frailty indices [ 15 ]. A comprehensive, multi-dimensional assessment of frailty in patients with MM is essential [ 16 ]. Furthermore, existing frailty models require further optimisation by incorporating additional factors to assist clinicians in making personalised treatment decisions and dose adjustments [ 17 , 18 ], ultimately enhancing treatment outcomes and improving patients’ quality of life [ 19 ]. To supplement the frailty assessment of MM patients, this study integrated multiple indicators related to frailty, including economic factors, pain, smoke, sleep disorders, and other previously reported factors [ 20 – 23 ]. Objective nutritional assessment indicators such as the prognostic nutritional index (PNI), modified body mass index (mBMI), and neutrophil-to-lymphocyte ratio (NLR) were also included [ 24 – 26 ]. Utilising LASSO regression, random survival forest, univariate, and multivariate Cox regression analyses, five indicators—age, HCT-CI, ECOG-PS, ISS, and PNI—were ultimately selected to construct a novel MM frailty model. This study aims to supplement the existing framework for assessing frailty in patients with MM.

Methods

Cohort selection and follow-up

This single-centre retrospective study excluded patients who did not receive treatment and those with large amounts of missing data. A total of 606 patients with newly diagnosed MM at Xijing Hospital from May 2006 to August 2022 were ultimately included, with the diagnostic criteria according to the IMWG guidelines [ 27 ]. The entire dataset was randomly divided into a development set ( N = 424) and a validation set ( N = 182) at a 7:3 ratio using the Sample function in R software. The enrolment flowchart of patients is depicted in Figure S1.

Overall survival (OS) was defined as the time from MM diagnosis to either death from any cause or the last follow-up. Progression-free survival (PFS) was defined as the time from MM diagnosis to disease progression, death from any cause, or the last follow-up, which occurred on December 1, 2023. For both PFS and OS, patients who were lost to follow-up were censored at the date of their last confirmed follow-up. Conventional or adjusted dosages and grade ≥ 2 nonhematologic AEs experienced during induction therapy were recorded. AEs were defined according to the General Terminology Standard for Adverse Events 5.0 (CTCAE_V 5.0, 2017).

Variable selection and definitions

Variables were collected at diagnosis, including sex, age, Hematopoietic Cell Transplantation-Comorbidity Index (HCT-CI), ADL, ECOG-PS, Norton score [ 28 ], loss of appetite [ 29 ], weight loss during the previous three months [ 29 ], body mass index (BMI) [ 29 ], pain, sleep disorder [ 30 ], smoke, insurance, educational level, polypharmacy [ 31 ], DS stage, ISS stage, RISS stage, PNI, NLR, and mBMI. The definitions and calculation methods for these variables are described in the Supplemental Methods. Continuous variables were categorised into discrete groups to establish a practical frailty model for clinical calculation. Variables with established reference ranges were categorised accordingly, while those lacking clear or consensus reference ranges were classified using cut-off values calculated using X-tile software [ 32 ]. Patients with missing data were further excluded based on the final variables included in the model. Ultimately, data from 412 patients were used for model construction, and data from 179 patients were used for validation. The numbers of data points used for model comparison are shown in Figure S1.

Construction and validation of the novel model

Univariate Cox regression analysis was used to screen variables within the development set and identify those with p < 0.1 as potential candidates. To mitigate the risk of model overfitting, we conducted LASSO regression analysis on these candidate variables [ 33 ]. Variable importance was then assessed using the random survival forest method [ 34 ], and variables that reduced the model’s predictive performance were excluded. Consequently, five variables were ultimately selected to construct a novel frailty model.

The predictive ability of the novel model was evaluated in the development and validation sets from four aspects: (1) the discriminative ability of the model was assessed using the time-dependent Harrell’s concordance index (C-index) and area under the curve (AUC) of the receiver operating characteristic (ROC) curve. (2) The degree of alignment between the actual and predicted incidence rates was assessed using a calibration curve. A good fit between the curve and the 45° diagonal line indicated that the predicted incidence rate aligned well with the actual incidence rate. (3) The model’s clinical applicability, risk threshold range, and net benefit were evaluated using decision curve analysis (DCA). (4) The predictive ability of the novel model was assessed via integrated discrimination improvement (IDI) and continuous net reclassification index (NRI) tests to determine whether it was improved.

Statistical methods

All the statistical analyses were performed using R 4.1.0 and SPSS 26.0. A two-sided p value < 0.05 was considered statistically significant. Continuous variables are presented as medians (interquartile ranges), and categorical variables are summarised by the frequency and percentage of patients in each group. Categorical variables were assessed using the Chi-square test or Fisher’s exact test. Survival rates were estimated using the Kaplan–Meier method, and comparisons between risk groups were performed using the log-rank test. The R packages utilised in these analyses are described in the Supplemental Methods.

Results

Patient characteristics

According to the inclusion criteria, data from 682 patients were collected. Seventy-six patients were excluded due to a lack of treatment or extensive missing baseline data (exceeding 30%). Consequently, 606 patients met the eligibility criteria and were included in further analyses. The baseline characteristics are presented in Table S1, revealing no statistically significant differences between the development and validation cohorts. Among the patients aged over 55 years, 61.3% were in the development set, and 59.4% were in the validation set. The median follow-up durations for the development and validation sets were 45.6 months (3.2 to 194.6 months) and 39.5 months (6.1 to 149.0 months), respectively. The corresponding median OS times were 57.1 months and 53.0 months, and median PFS times were 25.1 months and 23.6 months, respectively.

Construction and evaluation of the novel model

Univariate Cox regression analysis was initially conducted on all variables in the development set, identifying nine variables with p < 0.1, including age, HCT-CI, ECOG-PS, ADL, Norton scale score, loss of appetite, ISS, PNI, and mBMI (Table 1 ). LASSO regression analysis was subsequently applied to these variables, yielding six selected variables: age, HCT-CI, ECOG-PS, loss of appetite, ISS, and PNI (Figure S2A, B). These six variables were then incorporated into a random survival forest regression analysis, resulting in the final selection of five variables: age, HCT-CI, ECOG-PS, ISS, and PNI (Figure S2C). A novel frailty model (Fmodel) was constructed using these five predictors as a nomogram (Fig. 1 A) through a multivariate Cox proportional hazards model. The p values for the three tests (likelihood ratio, Wald, and score) were all less than 0.001 (Table S2), indicating a robust fit of the novel model. In the development set, 126 samples were used, with 1000 resamplings performed to plot the calibration curve and evaluate the calibration degree of Fmodel. The 3 year OS calibration curve demonstrated that the predicted probability agreed with the observed probability (Fig. 1 B). Regarding discrimination, the ROC curves for the 6 month, 1 year, 3 year, and 5 year OS rates of the Fmodel are shown in Fig. 1 C and were all greater than 0.7, indicating that the Fmodel had good discrimination ability. In addition, the calibration curves of the 6 month, 1 year, and 5 year OS rates are shown in Figure S2D–F.

The Fmodel was compared with the Smodel and the RISS to evaluate the predictability. The time-dependent AUC and C-index of the Fmodel consistently exhibited values superior to those of the Smodel and RISS (Table S3; Fig. 2 A–D). Additionally, the DCAs for Fmodel, Smodel, and RISS at 3 years are illustrated in Fig. 2 E, F. We also calculated the IDI and continuous NRI for Fmodel (Table S3) to assess whether there was an enhancement in the predictive efficiency compared with Smodel and RISS. Compared with the Smodel, the Fmodel demonstrated enhanced predictive efficiency for the OS of patients with MM at 6 months, 1 year, 3 years, and 5 years. The IDIs of the 6 month, 1 year, 3 year, and 5 year OS rates were 2.1% ( P = 0.02), 2.9% ( P = 0.00), 6.6% ( P = 0.00, Fig. 2 G), and 11.5% ( P = 0.00), respectively. The continuous NRIs for these time points were 35.3% ( P = 0.03), 22.3% ( P = 0.02), 24.5% ( P = 0.00, Fig. 2 G), and 37.0% ( P = 0.00), respectively. Furthermore, the predictive efficiency of the Fmodel for 5 year OS was significantly greater than that of the RISS, as evidenced by a significant increase in the IDI and continuous NRI of 12.9% ( P = 0.00) and 37.3% ( P = 0.02), respectively. Compared with the RISS, the Fmodel was not inferior for the prediction of 6 month, 1 year, and 3 year OS (Fig. 2 G).

Construction of a new frailty scoring system

The newly developed frailty scoring system is presented in Table 2 . Patients were categorised into three categories: fit (0–2), intermediate fit (3–5), and frail (≥ 6). In the development set ( N = 412), 87 patients (21.1%) were classified as fit, 224(54.4%) as intermediate fit, and 101(24.5%) as frail. The median OS times for these three groups were 106.0, 58.4, and 37.0 months, respectively. Compared with the Smodel (Fig. 3 B) and the RISS (Fig. 3 C), the Fmodel (Fig. 3 A) demonstrated superior ability to differentiate among the three patient groups. Additionally, median PFS across various models were assessed (Fig. 3 D‒F), revealing that both the Fmodel and the RISS exhibited robust prognostic stratification capabilities (Fig. 3 D, F).

Validation of the Fmodel

In the validation cohort, a multivariable Cox proportional hazards model was constructed using the five variables: age, HCT-CI, ECOG-PS, ISS, and PNI. The p values for this model’s three tests (likelihood ratio, Wald, and score) were consistently less than 0.001. ROC curves for 6 month, 1 year, 3 year, and 5 year OS are shown in Figure S3E, demonstrating that the Fmodel also exhibited good discriminative ability in the validation set. Calibration curves aligned closely with the 45°diagonal line, indicating a high level of concordance between predicted and observed survival probabilities (Figure S3A‒D).

Table S4 presents additional comprehensive evaluation results of the Fmodel in the validation set, revealing that the time-dependent AUC and C-index were generally higher than those of the Smodel and RISS. In addition, compared with the other two scoring systems, the DCA of Fmodel had greater net profit over a broader range of risk thresholds. Compared with the Smodel, the Fmodel showed improved predictive efficacy of the IDI and continuous NRI for 6 month, 1 year, 3 year and 5 year OS, and it was not inferior to the RISS in predicting all timepoints.

In the validation set ( N = 179), 42 patients (23.5%) were classified as fit, 91 (50.8%) as intermediate fit, and 46 (25.7%) as frail; the median OS values were 107.6, 53.9, and 33.8 months, respectively. We plotted the OS curves of patients with MM in the validation set under three scoring systems (Figure S4A-C). As the number of cases decreased, the K–M survival curves of Smodel were not significantly different ( p = 0.45, Figure S4B). We also assessed the median PFS of the three patient cohorts using various models. We plotted PFS curves (Figure S4D-F), demonstrating that the Fmodel and RISS exhibited robust prognostic stratification capabilities (Figure S4D, F).

To assess the generalisability of the Fmodel in different treatment contexts, the dataset was chronologically divided into two periods: May 2006–December 2015 and January 2016–August 2022, comprising 260 and 346 patients, respectively. K–M survival analyses of OS and PFS revealed statistically significant differences among the three patient stratifications in both periods (Figure S4G, H, J, K). Additionally, the median OS and PFS times for the fit and intermediate fit groups exhibited notable improvements over time, whereas no such trend was observed in the frail subgroup. These findings underscore that, despite advancements in MM therapeutics, frail patients continue to experience limited clinical benefit.

Furthermore, model validation was evaluated in patients over 65 years ( N = 153). Using the Fmodel stratification system, the OS curves among the three groups showed statistically significant differences ( p = 0.0088). Although the PFS curves did not reach statistical significance ( p = 0.053), a trend towards improved survival was observed in the non-frail group (Figure S4I, L), demonstrating that the Fmodel maintains robust predictive stability in this elderly patient population.

Analyses of AEs and early mortality

Based on the Fmodel, there were significant differences in baseline characteristics among the fit ( N = 129), intermediate fit ( N = 315), and frail ( N = 147). The frail group was characterised by older age, more comorbidities, functional impairment, worse nutritional status, and poorer treatment tolerance (data not shown). In the frail patient cohort, data on treatment dosages and AEs were available for 112 patients (76.2%) during the induction therapy period. Among the 69 patients who received conventional-dose treatment, 14 (20.3%) experienced grade ≥ 2 nonhematologic AEs (Fig. 4 A, Table S5). In contrast, among the 43 patients who received reduced-dose treatment, only 2(4.7%) experienced grade ≥ 2 nonhematologic AEs (Fig. 4 B, p = 0.026), suggesting that frail patients in the Fmodel group were more likely to experience grade ≥ 2 nonhematologic AEs when conventional-dose treatment was administered. However, the two groups had no significant difference in survival time (Fig. 4 C).

Across the whole dataset (N = 591), early mortality rates were analysed based on frailty stratification by the Fmodel (Fig. 4 D). At 2 months, the early mortality rates were 0/129 (0.0%), 5/315 (1.6%), and 15/147 (10.2%), respectively. At 6 months, the rates were 0/129 (0.0%), 11/315 (3.5%), and 23/147 (15.6%), respectively. By 12 months, the rates were 1/129 (0.8%), 22/315 (7.0%), and 33/147 (22.4%), respectively. Logistic regression models were constructed for each time point using variables from the Fmodel, all of which demonstrated statistical significance ( p < 0.05). The corresponding AUC values were 0.8303 ( data not shown ), 0.8350 (Fig. 4 E), and 0.7973 (Fig. 4 F), indicating that the Fmodel effectively predicted early mortality in MM patients.

Discussion

MM predominantly affects the elderly population, and its clinical manifestations, including bone pain, anaemia, and susceptibility to infections, contribute to a greater likelihood of patients experiencing debilitating conditions. Frailty is closely linked to the overall outcome of MM patients; however, accurately identifying frailty subgroups remains a critical area of ongoing research. In this study, we recruited the MM population in Northwest China. We systematically identified potential prognostic factors affecting MM patients, including age, comorbidities, disease-related risk stratification, physical condition, functional capacity, nutritional status, and socioeconomic factors. Utilising statistical methodologies, we incorporated five variables—age, HCT-CI, ECOG-PS, ISS, and PNI—to develop a novel frailty model for MM patients. Compared with the Smodel and RISS, this Fmodel exhibited robust discrimination and stability in predicting OS and PFS in both the development and validation sets. It demonstrated mostly superior or noninferior calibration, discrimination, clinical applicability, and predictive ability. Frail patients within the Fmodel were found to be at a higher risk for grade ≥ 2 nonhematologic AEs when receiving conventional-dose treatment. Furthermore, the Fmodel provided enhanced predictions of early mortality in MM patients.

Research has shown that nutritional indicators correlate with MM patient prognosis [ 35 – 37 ]. Kim et al. performed a retrospective analysis on the association between pre-treatment dietary status and survival in 216 MM patients, demonstrating that those experiencing malnutrition prior to treatment presented significantly lower OS rates than their well-nourished counterparts [ 38 ]. The MNA-SF questionnaire is a prevalent tool for nutritional evaluation in MM research, with findings indicating that the MNA-SF serves as an independent predictor of OS in MM patients [ 17 ]. While it simplifies the assessment process of the MNA questionnaire [ 29 ], it is still hindered by low sensitivity and a considerable rate of missed diagnoses. Due to the limitations of retrospective data, we were unable to gather information on nutritional status as evaluated by the aforementioned scales. Therefore, we chose to utilise the objective indicator PNI to represent the nutritional status of MM patients, which has been widely applied in nutritional assessments across various diseases [ 39 – 43 ]. Low PNI scores have been linked to poor prognosis in MM patients, prostate cancer patients, breast cancer patients, oesophageal cancer patients, and other malignancies, and our model further validated its prognostic significance.

The age and clinical characteristics of patients with MM contribute to a higher prevalence of comorbidities [ 44 ]. Numerous studies have indicated that comorbidities significantly influence the prognosis of MM patients [ 45 – 47 ], particularly in predicting early mortality [ 48 ]. Patients with MM frequently experience a high incidence of infections during active disease phases. The HCT-CI includes infection and various organ comorbidities, offering a more comprehensive representation of the comorbidity status in patients with MM than does the CCI. Our study also revealed that the Fmodel, incorporating the HCT-CI, effectively predicted early mortality in MM patients. Our model integrated objective indicators, age, physical performance, comorbidities, and nutritional status, all of which were routinely obtainable in clinical settings. Compared with existing frailty models, this approach encompasses both disease-specific indicators and a broader range of patient-related factors. This study potentially addressed limitations such as the lack of disease-specific indicators in the IMWG frailty model, insufficient evaluation of comorbidities and disease-related risks in the Mayo frailty model, the overly simplistic binary classification method of the simplified frailty scale, and the absence of statistical screening in the Smodel [ 49 ]. The Fmodel provides a relatively thorough assessment of MM patients’ frailty status, demonstrating strong predictive power for prognosis. Further validation in larger cohorts and prospective clinical trials is warranted.

Establishing a frailty assessment model is intended to assist clinicians in making informed treatment decisions. Currently, there is a paucity of research examining the relationship between frailty status and adjustments in drug dosing for MM patients. Our study revealed that frail patients identified via the Fmodel were at a greater risk of experiencing grade ≥ 2 nonhematologic AEs with conventional-dose treatment than with reduced-dose chemotherapy; however, no significant difference in survival was observed between the two groups. These findings suggest that conventional-dose treatment may not confer superior survival benefits for frail patients, given the increased incidence of AEs leading to subsequent dose reduction or discontinuation [ 50 , 51 ]. Our model offered a new alternative for the clinical evaluation of frailty status. However, clinical decision-making represents a sophisticated art form; the current disease stage, multi-dimensional frailty assessment, and the clinician's clinical experience should guide individual medical decision-making [ 52 , 53 ]. This study offers a preliminary exploration into how patients with frailty characteristics weigh the trade-offs between treatment risks and benefits, emphasising the need for further prospective research to validate these findings.

However, our study still has certain limitations. First, our data were derived solely from a single centre in the Chinese population. Further validation is necessary to ascertain its applicability through multi-centre studies and diverse ethnic populations. Second, the retrospective design may introduce bias despite implementing strategies such as randomisation and training of data collectors to mitigate potential selection and information bias; completely avoiding these biases remains challenging. Third, the real-time assessment of frailty in patients with MM is critically important. However, due to the limitations inherent in retrospective studies, most clinical parameters were not dynamically monitored throughout the treatment course. Consequently, this research did not focus on dynamic assessments. And additional exploration is warranted regarding the timely adjustment of treatment plans based on the longitudinal and evolving frailty status of patients.

In conclusion, we developed and validated a novel frailty model for the first time within a cohort of MM in Northwest China, thereby enhancing the current framework for assessing frailty in this population. Our findings suggest that frail patients might not benefit from conventional therapeutic dosages. Therefore, a thorough evaluation of frailty status is essential for meticulously determining treatment alternatives and modifying medication dosages. Balancing therapeutic efficacy, potential adverse effects, and quality of life is crucial in clinical decision-making. Unlike existing models, our framework integrates a broader spectrum of factors and anticipates further prospective validation.

Supplementary Information

Supplementary Material 1: Figure S1 Flowchart of patient enrolment

Supplementary Material 2: Figure S2 Variable selection and calibration curves in the development cohort.Coefficients of different variables from Lasso analysis.1000 bootstrap resamples with Lasso analysis. Variable importance ranking from random survival forest analysis.Calibration curves of the Fmodel for predicting 6 month, 1 year and 5 year OS

Supplementary Material 3: Figure S3 Calibration and ROC curves in the validation cohort.Calibration curves of the Fmodel for predicting 6 month, 1 year, 3 year and 5 year OS.The 6 month, 1 year, 3 year and 5 year ROC curves of the Fmodel

Supplementary Material 4: Figure S4 K‒M survival curves in the validation cohort.OS of patients stratified by the Fmodel, Smodel and RISS.PFS of patients stratified by the Fmodel, Smodel and RISS.OS and PFS of patients stratified by the Fmodel between May 2006 and December 2015.OS and PFS of patients stratified by the Fmodel between January 2016 and August 2022.OS and PFS of patients older than 65 years stratified by the Fmodel

Supplementary Material 5

Supplementary Material 6

Supplementary Material 7

Supplementary Material 8

Supplementary Material 9

Supplementary Material 10

Acknowledgements

We want to acknowledge the support from all the medical and nursing staff, Department of Hematology, Xijing Hospital, Air Force Medical University, Xi'an, Shaanxi, China.

Author contributions

Biao Tian, Li Xu and Shuangshuang Jia contributed equally to this work as co-first authors. Hailong Tang and Guangxun Gao supervised the study and are corresponding authors. Biao Tian, Li Xu and Shuangshuang Jia were responsible for the study design, data analysis and manuscript writing. Wenrui Sun, Chunyan Zhang, Juan Feng, Juan Hui, Miaoyu Li, Wanting Xiao, Lulu Wang, Ruochen Wu, Yanxia Weng, Zhenyang Li, Jiao Jia, Zeyi Ai participated in data collection, processing, and interpretation. All authors contributed to manuscript preparation and approved the final version of the manuscript for submission. Hailong Tang and Guangxun Gao supervised the study work flow, modified the manuscript and provided fundings.

Funding

The authors would like to acknowledge the financial support from the National Natural Science Foundation of China (No. 82470203), Shaanxi Province Health Blood Tumor Disease Scientific Research Innovation Project (No. 2023TD-03).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study adhered to the tenets of the Declaration of Helsinki and was approved by the Ethics Committee of Xijing hospital. All patients provided informed consent.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Biao Tian, Li Xu and Shuangshuang Jia have contributed equally to this work.

Contributor Information

Hailong Tang, Email: thl19880713@163.com.

Guangxun Gao, Email: gaoguangxun@fmmu.edu.cn.

References

1. Cowan AJ, Green DJ, Kwok M, Lee S, Coffey DG, Holmberg LA, et al. Diagnosis and management of multiple myeloma: a review. JAMA. 2022;327(5):464–77. 10.1001/jama.2022.0003.

2. Facon T, Leleu X, Manier S. How i treat multiple myeloma in geriatric patients. Blood. 2024;143(3):224–32. 10.1182/blood.2022017635.

3. Rosenberg PS, Barker KA, Anderson WF. Future distribution of multiple myeloma in the United States by sex, age, and race/ethnicity. Blood. 2015;125(2):410–2. 10.1182/blood-2014-10-609461.

4. Zweegman S, Engelhardt M, Larocca A. Elderly patients with multiple myeloma: towards a frailty approach? Curr Opin Oncol. 2017;29(5):315–21. 10.1097/CCO.0000000000000395.

5. Tang H, Jia S, Bi L, Jia W, Gao G. Treatment options for older unfit patients with acute myeloid leukemia. Future Oncol. 2021;17(7):837–51. 10.2217/fon-2020-0615.

6. Wildes TM, Tuchman SA, Klepin HD, Mikhael J, Trinkaus K, Stockerl Goldstein K, et al. Geriatric assessment in older adults with multiple myeloma. J Am Geriatr Soc. 2019;67(5):987–91. 10.1111/jgs.15715.

7. Gordon EH, Hubbard RE. Frailty: understanding the difference between age and ageing. Age Ageing. 2022. 10.1093/ageing/afac185.

8. Kim DH, Rockwood K. Frailty in older adults. N Engl J Med. 2024;391(6):538–48. 10.1056/NEJMra2301292.

9. Choi JY, Kim KI. Assessing frailty using comprehensive geriatric assessment in older patients with hematologic malignancy. Blood Res. 2022;57(S1):1–5. 10.5045/br.2022.2021218.

10. Wildes TM, Campagnaro E. Management of multiple myeloma in older adults: gaining ground with geriatric assessment. J Geriatr Oncol. 2017;8(1):1–7. 10.1016/j.jgo.2016.04.001.

11. Palumbo A, Bringhen S, Mateos MV, Larocca A, Facon T, Kumar SK, et al. Geriatric assessment predicts survival and toxicities in elderly myeloma patients: an international myeloma working group report. Blood. 2015;125(13):2068–74. 10.1182/blood-2014-12-615187.

12. Milani P, Vincent Rajkumar S, Merlini G, Kumar S, Gertz MA, Palladini G, et al. N-terminal fragment of the type-B natriuretic peptide (NT-proBNP) contributes to a simple new frailty score in patients with newly diagnosed multiple myeloma. Am J Hematol. 2016;91(11):1129–34. 10.1002/ajh.24532.

13. Facon T, Dimopoulos MA, Meuleman N, Belch A, Mohty M, Chen W, 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):224–33. 10.1038/s41375-019-0539-0.

14. Lee HS, Lee J, Jo JC, Jung SH, Lee JJ, Kim D, et al. Development of a new clinical index to easily assess frailty of elderly patients with multiple myeloma in Asian population. Sci Rep. 2021;11(1):22907. 10.1038/s41598-021-02433-6.

15. Grant SJ, Freeman CL, Rosko AE. Treatment of older adult or frail patients with multiple myeloma. Hematology Am Soc Hematol Educ Program. 2021;2021(1):46–54. 10.1182/hematology.2021000231.

16. Wildiers H, Heeren P, Puts M, Topinkova E, Janssen-Heijnen MLG, Extermann M, et al. International society of geriatric oncology consensus on geriatric assessment in older patients with cancer. J Clin Oncol. 2014;32(24):2595–603. 10.1200/JCO.2013.54.8347.

17. Chen Y, Gu J, Huang B, Liu J, Li X, Li J. Development and validation of a chemotherapy tolerance prediction model for Chinese multiple myeloma patients: the tm frailty score. Front Oncol. 2023;13:1103687. 10.3389/fonc.2023.1103687.

18. Yao Y, Sui WW, Liao AJ, Wang W, Chen LJ, Chu XX, et al. Comprehensive geriatric assessment in newly diagnosed older myeloma patients: a multicentre, prospective, non-interventional study. Age Ageing. 2022. 10.1093/ageing/afab211.

19. Möller M, Gengenbach L, Graziani G, Greil C, Wäsch R, Engelhardt M. Geriatric assessments and frailty scores in multiple myeloma patients: a needed tool for individualized treatment? Curr Opin Oncol. 2021;33(6):648–57. 10.1097/CCO.0000000000000792.

20. Doody P, Asamane EA, Aunger JA, Swales B, Lord JM, Greig CA, et al. The prevalence of frailty and pre-frailty among geriatric hospital inpatients and its association with economic prosperity and healthcare expenditure: a systematic review and meta-analysis of 467,779 geriatric hospital inpatients. Ageing Res Rev. 2022;80:101666. 10.1016/j.arr.2022.101666.

21. Ardoino I, Franchi C, Nobili A, Mannucci PM, Corli O, Reposi I. Pain and frailty in hospitalized older adults. Pain Ther. 2020;9(2):727–40. 10.1007/s40122-020-00202-3.

22. García-Esquinas E, Navas-Acien A, Rodríguez-Artalejo F. Exposure to secondhand tobacco smoke and the frailty syndrome in US older adults. Age. 2015. 10.1007/s11357-015-9762-4.

23. Morley JE. Frailty and sarcopenia in elderly. Wien Klin Wochenschr. 2016;128(Suppl 7):439–45. 10.1007/s00508-016-1087-5.

24. Li L, Wu M, Yu Z, Niu T. Nutritional status indices and monoclonal gammopathy of undetermined significance risk in the elderly population: findings from the national health and nutrition examination survey. Nutrients. 2023;15(19):4210. 10.3390/nu15194210.

25. Kittleson MM, Ruberg FL, Ambardekar AV, Brannagan TH, Cheng RK, Clarke JO, et al. 2023 ACC expert consensus decision pathway on comprehensive multidisciplinary care for the patient with cardiac amyloidosis. J Am Coll Cardiol. 2023;81(11):1076–126. 10.1016/j.jacc.2022.11.022.

26. Wang X, Feng H, Hui Y, Yu Z, Zhao T, Mao L, et al. Neutrophil-to-lymphocyte ratio is associated with malnutrition risk estimated by the royal free hospital–nutritional prioritizing tool in hospitalized cirrhosis. JPEN J Parenter Enteral Nutr. 2022;46(1):123–9. 10.1002/jpen.2097.

27. Rajkumar SVP, Dimopoulos MAP, Palumbo AP, Blade JM, Merlini GPM, Mateos MM, et al. International myeloma working group updated criteria for the diagnosis of multiple myeloma. Lancet Oncol. 2014;15(12):e538–48. 10.1016/S1470-2045(14)70442-5.

28. Shimoni Z, Dusseldorp N, Cohen Y, Barnisan I, Froom P. The Norton scale is an important predictor of in-hospital mortality in internal medicine patients. Ir J Med Sci (1971 -). 2023;192(4):1947–52. 10.1007/s11845-022-03250-0.

29. Rubenstein LZ, Harker JO, Salva A, Guigoz Y, Vellas B. Screening for undernutrition in geriatric practice: developing the short-form mini-nutritional assessment (mna-sf). J Gerontol A Biol Sci Med Sci. 2001;56(6):M366–72. 10.1093/gerona/56.6.m366.

30. Mogavero MP, Delrosso LM, Fanfulla F, Bruni O, Ferri R. Sleep disorders and cancer: state of the art and future perspectives. Sleep Med Rev. 2021;56:101409. 10.1016/j.smrv.2020.101409.

31. Masnoon N, Shakib S, Kalisch-Ellett L, Caughey GE. What is polypharmacy? A systematic review of definitions. BMC Geriatr. 2017. 10.1186/s12877-017-0621-2.

32. Camp RL, Dolled-Filhart M, Rimm DL. X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Clin Cancer Res. 2004;10(21):7252–9. 10.1158/1078-0432.CCR-04-0713.

33. Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 2010;33(1):1–22.

34. Zhao P, Zhen H, Zhao H, Huang Y, Cao B. Identification of hub genes and potential molecular mechanisms related to radiotherapy sensitivity in rectal cancer based on multiple datasets. J Transl Med. 2023;21(1):176. 10.1186/s12967-023-04029-2.

35. Zhang Y, Chen Q, Lu C, Yu L. Prognostic role of controlling nutritional status score in hematological malignancies. Hematology. 2022;27(1):653–8. 10.1080/16078454.2022.2078040.

36. Zhang L, Chen S, Wang W, Wang Y, Liang Y. Inflammatory and nutritional scoring system for predicting prognosis in patients with newly diagnosed multiple myeloma. J Inflamm Res. 2023;16:7–17. 10.2147/JIR.S390279.

37. Zhang L, Chen S, Huang M, Wang W, Liang Y, Wang Y. Prognostic value of the nutritional risk index in patients with newly diagnosed multiple myeloma. Ann Hematol. 2023;102(1):125–32. 10.1007/s00277-022-05059-4.

38. Kim HS, Lee JY, Lim SH, Cho J, Kim SJ, Jang JH, et al. Patient-generated subjective global assessment as a prognosis tool in patients with multiple myeloma. Nutrition. 2017;36:67–71. 10.1016/j.nut.2016.06.009.

39. Shi J, Liu T, Ge Y, Liu C, Zhang Q, Xie H, et al. Cholesterol-modified prognostic nutritional index (cpni) as an effective tool for assessing the nutrition status and predicting survival in patients with breast cancer. BMC Med. 2023. 10.1186/s12916-023-03225-7.

40. Jabłońska B, Mrowiec S. Nutritional status and its detection in patients with inflammatory bowel diseases. Nutrients. 2023;15(8):1991. 10.3390/nu15081991.

41. Liang F, Dong XY, Tang GF, Qi KM, Chen W, Sang W, et al. Influence of prognostic nutritional index and controlling nutritional status on the prognosis of patients with multiple myeloma. Zhonghua Xue Ye Xue Za Zhi. 2021;42(4):332–7. 10.3760/cma.j.issn.0253-2727.2021.04.011.

42. Ellez HI, Keskinkilic M, Semiz HS, Arayici ME, Kısa E, Oztop I. The prognostic nutritional index (PNI): a new biomarker for determining prognosis in metastatic castration-sensitive prostate carcinoma. J Clin Med. 2023;12(17):5434. 10.3390/jcm12175434.

43. Okadome K, Baba Y, Yagi T, Kiyozumi Y, Ishimoto T, Iwatsuki M, et al. Prognostic nutritional index, tumor-infiltrating lymphocytes, and prognosis in patients with esophageal cancer. Ann Surg. 2020;271(4):693–700. 10.1097/SLA.0000000000002985.

44. Gregersen H, Vangsted AJ, Abildgaard N, Andersen NF, Pedersen RS, Frolund UC, et al. The impact of comorbidity on mortality in multiple myeloma: a Danish nationwide population-based study. Cancer Med. 2017;6(7):1807–16. 10.1002/cam4.1128.

45. Bila J, Jelicic J, Djurasinovic V, Vukovic V, Sretenovic A, Andjelic B, et al. Prognostic effect of comorbidity indices in elderly patients with multiple myeloma. Clin Lymphoma Myeloma Leuk. 2015;15(7):416–9. 10.1016/j.clml.2015.03.004.

46. Onec B, Okutan H, Albayrak M, Can ES, Aslan V, Soyer Kosemehmetoglu O, et al. Comparative evaluation of common comorbidity scores and Freiburger comorbidity index as prognostic variables in a real life multiple myeloma population. Indian J Hematol Blood Transfus. 2016;32(4):424–30. 10.1007/s12288-015-0618-y.

47. Hou Z, Kang Y, Yang G, Wang Z, Wang F, Yu Y, et al. Pleural effusion-based nomogram to predict outcomes in unselected patients with multiple myeloma: a large single center experience. Ann Hematol. 2021;100(7):1789–801. 10.1007/s00277-021-04484-1.

48. Jung S, Cho M, Kim HK, Kim SJ, Kim K, Cheong J, et al. Risk factors associated with early mortality in patients with multiple myeloma who were treated upfront with a novel agents containing regimen. BMC Cancer. 2016. 10.1186/s12885-016-2645-y.

49. Sim S, Kalff A, Tuch G, Mollee P, Ho PJ, Harrison S, et al. The importance of frailty assessment in multiple myeloma: a position statement from the myeloma scientific advisory group to myeloma Australia. Intern Med J. 2023;53(5):819–24. 10.1111/imj.16049.

50. Oriol A, Dimopoulos M, Schjesvold F, Beksac M, Facon T, Dhanasiri S, et al. Pomalidomide, bortezomib, and dexamethasone in lenalidomide-pretreated multiple myeloma: a subanalysis of optimismm by frailty and bortezomib dose adjustment. Clin Lymphoma Myeloma Leuk. 2024;24(3):165–76. 10.1016/j.clml.2023.10.009.

51. Cho S, Shin H, Jung KS, Kim DY. Dose adjustment helps obtain better outcomes in multiple myeloma patients with bortezomib, melphalan, and prednisolone (vmp) treatment. Turk J Hematol. 2019;36(2):106–11. 10.4274/tjh.galenos.2019.2019.0306.

52. Zweegman S, Larocca A. Frailty in multiple myeloma: the need for harmony to prevent doing harm. Lancet Haematol. 2019;6(3):e117–8. 10.1016/S2352-3026(19)30011-0.

53. Kaweme NM, Changwe GJ, Zhou F. Approaches and challenges in the management of multiple myeloma in the very old: future treatment prospects. Front Med. 2021;8:612696. 10.3389/fmed.2021.612696.

Associated Data

Supplementary Materials

Supplementary Material 1: Figure S1 Flowchart of patient enrolment

Supplementary Material 2: Figure S2 Variable selection and calibration curves in the development cohort.Coefficients of different variables from Lasso analysis.1000 bootstrap resamples with Lasso analysis. Variable importance ranking from random survival forest analysis.Calibration curves of the Fmodel for predicting 6 month, 1 year and 5 year OS

Supplementary Material 3: Figure S3 Calibration and ROC curves in the validation cohort.Calibration curves of the Fmodel for predicting 6 month, 1 year, 3 year and 5 year OS.The 6 month, 1 year, 3 year and 5 year ROC curves of the Fmodel

Supplementary Material 4: Figure S4 K‒M survival curves in the validation cohort.OS of patients stratified by the Fmodel, Smodel and RISS.PFS of patients stratified by the Fmodel, Smodel and RISS.OS and PFS of patients stratified by the Fmodel between May 2006 and December 2015.OS and PFS of patients stratified by the Fmodel between January 2016 and August 2022.OS and PFS of patients older than 65 years stratified by the Fmodel

Supplementary Material 5

Supplementary Material 6

Supplementary Material 7

Supplementary Material 8

Supplementary Material 9

Supplementary Material 10

Data Availability Statement

No datasets were generated or analysed during the current study.