In the era of new drugs, age and the ECOG-PS are no longer sufficient for describing the heterogeneity of elderly MM patients, and the influence of patients' own prognostics characteristics gradually increases with age, while a relative detriment of the impact of prognostic factors related to tumor biology, such as cytogenetics, can occur.A frailty assessment of a patient's functional status, comorbidities, psychology, and cognition can be used to identify the best treatment options for elderly MM patients and facilitate clinical research in elderly MM patients [19, 20]. Comprehensive geriatric assessment (CGA) is now widely used in oncology [21], and the National Comprehensive Cancer Network (NCCN) guidelines also recommend CGA to assess frailty in patients with malignant haematological diseases. However, CGA is time-consuming and complex and requires a team of specialists, and the disease characteristics of MM itself dictate the need for a specific frailty assessment tool, thus contributing to the development of a frailty tool for MM patients.Although most clinicians believe that routine frailty assessment is time-consuming and costly, research has shown that, in fact, not only is frailty assessment feasible,it is also likely less time-consuming and costly than many other prognostic tests done in oncology [22,23,24].

The IMWG-FI is the first assessment tool applied to MM patients and was developed in a multicentre retrospective study with a large sample size (n = 869) and external validation [25]. The score consists of 34 factors and needs to be completed by both the doctor and patient, which is not only cumbersome but also inevitably subjective. In addition, individuals aged ≥ 80 years are directly assigned to the frailty group regardless of their physical status or the presence or absence of comorbidities, which is clearly unreasonable [26, 27]. The IFM reduces the subjectivity and complexity of the IMWG-FI by replacing ADL and IADL in the IMWG-FI with the ECOG-PS. However, because the ECOG-PS is a dichotomous variable, it does not reflect the heterogeneity of patient frailty in a dynamic manner [28]. The Mayo score includes only 3 factors (age, ECOG-PS, and NT-proBNP level) compared with the IMWG-FI, increasing the convenience and objectivity, but the difficulty of standardising NT-proBNP levels across centres limits its clinical application. The MRP, which is mainly based on biochemical and haematological indices, is simpler and easier to use than the other indices and can significantly improve the efficiency of risk stratification in elderly patients. The MRP is based on biochemical and haematological indices and is simple, significantly improves the efficiency and objectivity of risk stratification in elderly patients, and avoids the direct inclusion of patients aged ≥ 80 years into the frail group.

In this study, although the assessment results of the IMWG-FI with the MRP and Mayo score all had moderate concordance, the concordance between the assessment results of any other two tools was moderate, in which only 14 patients (21.8%) were identified as frail by all four tools, which suggests that treatment decisions based on the assessment results of the different frailty tools may result in similar patients receiving different treatments, thus affecting their prognosis and quality of life [29].

The proportion of frail patients was greater in the present study than in the IMWG-FI development study (53.6% and 30%, respectively) [13]. This difference may be because the patient cohort in this study was derived from the "real world", including patients with severe renal insufficiency (creatinine level > 2 mg/dl, 46.7% and 5%), an ECOG-PS score of 2–3 (84.4% and 31%), and an ISS score of III (77.8% and 31%), caused by a higher proportion of patients at high cytogenetic risk (46.7% and 24%) (Table 1). When comparing the present study with the Mayo score development study [14], the proportion of frail patients was higher (23.8% and 7%), and analysis of the demographic and baseline characteristics of the two studies revealed that the proportion of patients with an ECOG-PS ≥ 2 (45.0% and 19.0%) and the median NT-proBNP level (1225 ng/L and 109 ng/L) were higher in the present study than in the Mayo development study (Table 1). The differences between the two studies may have originated from the inclusion of the ECOG-PS and NT-proBNP level in the Mayo score, and the difficulty in harmonising the criteria for NT-proBNP levels is the main reason for the difficulty in externally validating this score. The proportion of frail patients was consistent when comparing the present study with an external validation study of MRP scores (46.4% and 46.4%) [16], whereas the proportion of frail patients was greater when comparing the present study with the IMWG-FI development study (67.9% and 49%), which may be related to the fact that the proportion of patients with an ISS of III (77.2% and 38.4%) and ECOG-PS ≥ 2 (40.4% and 21.5%) was greater (Table 1).

In this study, the assessment of frailty in patients with a survival period longer than 3 years is not static, and it can either improve or deteriorate over time (Supplementary Materials 3), which is analogous to the conclusion of Mian H et al. [30], indicating that dynamic assessment is a superior predictor of outcome compared to frailty at diagnosis. Some patients in this study have decided on chemotherapy regimen based on dynamic assessment, but unfortunately, there are few data in this part.

When frailty was assessed using the IMWG-FI, MRP and Mayo score, the median OS (Fig. 3) and PFS (Fig. 4) were significantly different between the nonfrail group and the frail group[25]; however, there was no statistically significant difference in the median PFS or OS of patients with frailty assessed by the four models (Fig. 5), which suggests that the survival outcomes of the patients who were recognised as frail by the different models were similar in the present study. The efficacy of stratification of the prognosis of the patients by the IMWG-FI, the MRP and the Mayo score was better than that of the IFM. The IMWG-FI is still the geriatric assessment model recommended by the EMN, the European Society for Medical Oncology (ESMO) and other guidelines, and all other frailty models developed need to be compared and validated with the IMWG-FI. In this study, the IMWG-FI showed moderate agreement with the MRP and IFM and only fair agreement with the Mayo score, indicating that the MRP is more effective than the Mayo score and IFM in stratifying patient prognosis. There are fewer prospective studies applying the MRP, Mayo score and IFM scale to guide treatment, and the Myeloma XIV study [FiTNEss, NCT03720041] [31] conducted by the Myeloma UK Consortium used the IMWG-FI; these studies compared the MRP score, which is expected to provide additional evidence-based evidence on the validity of the IMWG-FI and the clinical feasibility of the MRP.

In this study, we found that the assessment results of the four tools varied greatly, and different assessment models may lead to similar patients being assigned to different frailty groups and receiving different treatments; however, the agreement between the MRP score and the IMWG-FI score was high, so combining the IMWG-FI score with the MRP score may reduce the subjectivity of assessment and improve the identification of frailty [32]. However, given the diverse circumstances of each center, it is advisable to assess frailty in daily clinical practice, and the selection of the score shall rely on the background and experience of each center until a standardized approach can be available. The present study still has limitations. First, the number of patients aged ≥ 80 years included in the study was small to verify whether this conclusion was equally effective in these patients. Second, this was a retrospective study with a small sample size. Fewer patients aged ≥ 80 years were included in the study to verify whether the result was equivalent for these patients. Third, Further analysis is required to examine drug tolerance and the underlying causes of death.Last, since this was a retrospective study with a small sample size, additional randomised controlled studies are needed to validate the findings.