Cumulative deficits frailty index and relationship status predict survival in multiple myeloma

Multiple myeloma (MM) is considered a disease of older adults, with 69 years being the median age at diagnosis in the United States.¹ However, therapeutic advances in the last decade have not affected survival in older patients to the same degree as younger patients.^2,3 Reduced physical performance and organ dysfunction are significant factors contributing to poor outcomes in older patients, leading to higher rates of treatment toxicity and premature treatment discontinuation.⁴ At the same time, older patients are often denied access to effective therapies due to perceived intolerance solely based on age.^5,6 Thus, to avoid overtreatment of frail patients and undertreatment of fit patients, the characterization of physiological age at the time of diagnosis in this widely heterogeneous group is essential. Frailty, defined as a state of decreased physiological reserve,⁷ has been shown to predict increased treatment toxicity and worse outcomes in older patients with MM.^8,9 Over the last 10 years, several tools have been proposed to identify frail patients using combinations of age, comorbidities, functional status, Eastern Cooperative Oncology Group (ECOG) performance status (PS), laboratory markers of organ function, and/or cytogenetics.^10-15 The International Myeloma Working Group (IMWG) index, developed in 2015, is the most commonly used tool in the clinical trial setting. This is based on age, functional status, and the Charlson Comorbidity Index.⁹ Most of the currently available frailty tools have been developed using data from clinical trials, in which stringent selection criteria are used for inclusion. In addition, most tools were developed in transplant-ineligible cohorts only. Importantly, the use of chronological age in many of these scores has the potential to classify patients as frail based on age alone. Another limitation of some of these scores is the use of physician-rated PS, which may not be a reliable measure of functional status.^16,17 Although studies have primarily focused on baseline frailty, it is important to recognize that frailty status is dynamic and can improve or worsen after treatment.¹⁸ 

One of the main approaches to characterize frailty is the cumulative deficit index, also known as the Rockwood frailty index (FI).¹⁹ This is measured by calculating deficits across a broad range of systems using routinely collected clinical data. The FI has been used in several studies to assess frailty and shown to have prognostic value in different populations including patients with cancer.^20-23 In this study, we adopt a cumulative deficit approach to measure frailty in patients with newly diagnosed MM seen in our institution, using comorbidities and patient-provided information on functional status. We then use this index to evaluate the association between frailty and baseline patient- and disease-related characteristics, symptom burden, and survival outcomes. We also calculate the FI at 3, 6, and 12 months from the start of treatment, when data are available, and assess its prognostic impact at these time points.

In addition to frailty, social support, and socioeconomic status (SES) have been suggested to influence outcomes in patients with cancer, including those with MM. However, the independent prognostic value of these factors has not been consistent across studies.^24-29 Our study examines the association between these factors and disease/symptom burden and survival in patients with newly diagnosed MM.

This is a single-institution retrospective study. We included patients with MM who were seen at Mayo Clinic at the time of their diagnosis between 2005 and 2018 and had available data to calculate the FI within 3 months before diagnosis. Patients were identified from a preexisting prospectively maintained database. Additional data on disease stage, laboratory values at diagnosis, cytogenetics by fluorescence in situ hybridization (FISH), FI variables, and relationship status were obtained by a review of electronic medical records. The study was approved by the Mayo Clinic Institutional Review Board. All included patients had authorized the use of their data for research purposes.

We calculated the FI using the cumulative deficits method based on established criteria.¹⁹ The selection of variables and scoring are based on previously published studies from our institution.^20,21 We included 32 deficits for the calculation of the FI (Table 1). The first 14 items were about patient-reported activities of daily living, instrumental activities of daily living, dependence for assistance, dependence for a device for normal breathing, and exercise tolerance. These items were obtained from an institutional form, which was administered to all patients for completion before an outpatient clinic visit; this form, called “Current Visit Information,” was adopted in all clinical departments at Mayo Clinic until 2018. It comprised the following sections: patient information, health care provider information, medications, allergies, systems review, social context (education, employment, and relationship status), substance use history, and self-care/home environment assessment. Patients were asked to complete the form using pen and paper, and this was scanned to their medical records. The items used to calculate the FI are shown in supplemental Figure 1. The estimated time to complete these items is 3 minutes. Patients who had not completed the form within 3 months before diagnosis were excluded. The other 16 items included comorbidities, as documented by the treating provider, and body mass index. These were abstracted from the electronic medical record for the time of diagnosis. Malignancy was not included among the comorbidities because this would apply to all patients in this study. Anemia was also not included among comorbidities because this is a myeloma-defining event. The individual items were scored as 0, 0.5, or 1, with 1 indicating deficit. Patients who had >2 missing items were excluded. The FI was defined as the sum of individual scores divided by the total number of nonmissing items.

Data on self-reported relationship status and living situation were obtained from the same institutional form used to collect the first 14 items. Patients were asked whether they were married, separated, single, widowed, in a committed relationship, or “other.” They were also asked whether they lived alone, with a spouse, domestic partner, family, or “other.”

To assess SES, we used the Mayo Clinic Housing-Based Socioeconomic Status (HOUSES) index, a validated individual-level SES measure that is derived from individual housing characteristics. The development and validation of this measure has been published.³⁰ Briefly, this index is based on 4 housing variables: housing value, square footage of the housing unit, number of bedrooms, and number of bathrooms. These are obtained by linking address information to enumerated real property data that are available from local government assessors' offices. Each property item corresponding to an individual's address is standardized into a z score and aggregated into an overall z score for the 4 variables, such that a higher HOUSES score indicates higher SES. The HOUSES z score is then converted to quartiles, with the first quartile (Q1) representing the lowest SES and Q4 representing the highest SES.

We abstracted data on fatigue, pain, and quality of life (QOL) at the time of diagnosis using a patient-completed 3-item questionnaire adopted in the hematology department at Mayo Clinic from 2010 to 2018, termed “Hematology Patient Reported Symptom Screen” (supplemental Figure 2).³¹ Patients were asked to rate their symptoms in the past 7 days on a scale from 0 to 10, with 10 being the worst for fatigue and pain but the best for QOL. Patients completed the “Hematology Patient Reported Symptom Screen” using pen and paper, and the scores were then entered into the electronic medical record by a clinical assistant.

The International Staging System (ISS)³² and Revised ISS (R-ISS)³³ were used for disease staging at diagnosis. We used the electronic medical record to abstract data on cytogenetic risk by interphase FISH obtained at diagnosis. The methods for FISH analysis in our institution have been previously described.³⁴ High-risk FISH was defined by the presence of ≥1 of the following: high-risk immunoglobulin (IgG) translocation (t[4;14], t[14;16], or t[14;20] translocation), deletion 17p, and 1q gain/amplification.^33,35 We also obtained data on first-line induction, categorizing this into proteasome inhibitor–based, immunomodulatory drug–based, or proteasome inhibitor + immunomodulatory drug–based induction. Early transplantation was defined as autologous stem cell transplantation within 1 year of diagnosis. Treatment response was defined in accordance with the IMWG consensus criteria.³⁶ 

Baseline characteristics were presented as frequencies (percentage) for categorical variables and median (interquartile range) for continuous variables. Variables were compared between groups using Fisher exact test and Wilcoxon signed-rank test for categorical and continuous variables, respectively. Progression-free survival (PFS) was defined as the time from diagnosis to first disease progression or death from any cause. Patients without an event at last follow-up were censored. Overall survival (OS) was defined as the time from diagnosis to death from any cause. PFS and OS were estimated using the Kaplan-Meier method and compared between groups using the log-rank test. We evaluated the impact of frailty, relationship status, and SES on OS using univariate Cox proportional regression models and in multivariate models including age, R-ISS, early transplantation, and best response to first-line induction. The associations were summarized using hazard ratios (HRs) and the corresponding 95% confidence intervals (CIs). For all tests, P values < .05 were considered statistically significant. All statistical analysis was performed using JMP statistical software.

A total of 626 patients had available data within 3 months before diagnosis. Among those, 111 had >2 missing items for the FI and were excluded. The final cohort included 515 patients diagnosed with MM between January 2005 and May 2018. The proportion of females was lower among excluded patients (30% vs 40%; P = .04). Otherwise, baseline characteristics were similar between the 2 groups (supplemental Table 1).

The median age was 66 years (interquartile range, 59-74). The mean FI was 0.15 (standard deviation, 0.13), and the median was 0.10 (range, 0-0.60). Patients were first classified into fit (FI ≤ 0.06), intermediate-fit (0.06 < FI < 0.15), and frail (FI ≥ 0.15), based on tertiles. The first 2 groups were combined into a "nonfrail" group (FI < 0.15). This FI cutoff was also adopted in a previous study from our institution, which included patients with ovarian cancer.²¹ Among all patients, 315 (61%) were classified as nonfrail, and 200 (39%) were classified as frail. The prevalence of frailty was 28%, 38%, and 62% among patients aged <65, 65 to 74, and ≥75 years, respectively. Baseline characteristics in nonfrail and frail patients are presented in Table 2. Patients who were frail were older (median age, 70 vs 64 years; P < .001) and more likely to have anemia, ISS III (vs I/II) disease, and ECOG PS ≥2 (vs 0-1) than nonfrail patients. There was a slightly higher rate of high-risk FISH abnormalities among frail than nonfrail patients (49% vs 39%; P = .046), but there was no difference in the proportion of R-ISS III between the 2 groups. Patients who were frail were less likely to undergo early transplantation compared to nonfrail patients. There was no association between frailty and response to first-line induction (≥partial response [PR] or ≥very good PR [VGPR]).

The median follow-up in nonfrail and frail patients was 7.2 years (95% CI, 6.6-8.4) and 8.2 years (95% CI, 7.1-9.1), respectively (P = .36). In the overall cohort, the median PFS after first-line therapy was 3.1 years (95% CI, 2.8-3.8) in nonfrail and 2.1 years (95% CI, 1.8-2.7) in frail patients (P = .03). OS was not reached (95% CI, 6.6 years to not reached) in nonfrail patients (not reached in fit and 6.3 years in intermediate-fit), and it was 3.7 years (95% CI, 3.0-5.0) in frail patients (P < .001; Figure 1). The OS in nonfrail vs frail patients in various subgroups is shown in supplemental Table 3. The physician-assessed ECOG PS was available for 243 patients. Among patients with ECOG PS 0 (n = 83), 70 (84%) were nonfrail, and 13 (16%) were frail. Among patients with ECOG PS 1 (n = 93), 55 (59%) were nonfrail, and 38 (41%) were frail. Among patients with ECOG PS 0 to 1 (n = 176), OS was 8.3 years and 4.2 years in nonfrail and frail patients, respectively (P = .01; Figure 2A). Among patients with ECOG PS ≥2 (n = 67), 25 (37%) were nonfrail, and 42 (63%) were frail. OS was 6.3 years and 3.3 years in the 2 groups, respectively (P = .11; Figure 2B). OS was shorter in frail than nonfrail patients with R-ISS I to II, but there was no difference in OS between the 2 groups among patients with R-ISS III. Among patients who did not undergo early transplantation, OS was significantly shorter in frail than nonfrail patients. However, there was no significant difference in OS between the 2 groups among those who underwent early transplant. Otherwise, frail patients had significantly shorter survival than nonfrail patients in all subgroups.

Data on relationship status were available for 509 patients. Among those, 410 (81%) reported being married. Eleven (2%) were in a committed relationship, 39 (8%) were widowed, 21 (21%) were single, and 27 (5%) were divorced or separated. One patient (<1%) answered “other.” Patients who were not married/in a relationship were older (median age, 71 vs 66 years; P = .005), more likely to have high ISS III disease (51% vs 32%; P = .002), ECOG PS ≥2 (40% vs 24%; P = .04), and to be frail (57% vs 35%; P < . 001), but less likely to be in the upper quartile for SES (26% vs 45%; P = .004) and to undergo early transplantation (22% vs 39%; P = .002). The proportion of patients with R-ISS III was not significantly different between the 2 groups (19% vs 14%, respectively; P = .29). There was no difference in the proportion with anemia, hypercalcemia, or renal failure between the 2 groups. There was also no difference in the proportion who achieved ≥PR or ≥VGPR between the 2 groups.

Patients who were not married/in a relationship had decreased survival compared with those who were married/in a relationship. The median OS was 3.7 (95% CI, 2.5-5.1) and 7.5 years (95% CI, 6.3-8.5) in the 2 groups, respectively (P < .001). Among females, OS was 3.3 years vs 8.3 years, respectively (P < .001; Figure 3A). Among males, OS was 4.0 years vs 6.4 years, respectively (P = .04; Figure 3B).

Data on living situation were available for 506 patients. Among those, 383 (76%) reported living with a spouse, 34 (7%) lived with family, 13 (3%) lived with a domestic partner, and 72 (14%) lived alone; 4 (1%) answered “other.” There was a high concordance between living situation and relationship status. Among those who were married/in a relationship, only 5 patients (1%) lived alone. For patients who were not married/in a relationship, 16 patients (19%) were living with someone (14 with family and 2 with a partner). Living alone was similarly associated with older age, frailty, higher stage disease, and lower probability of undergoing early transplantation, but the association with ECOG PS ≥2 and SES did not reach statistical significance.

Patients who lived alone had significantly decreased survival (median OS, 3.3 years; 95% CI, 2.3-5.3) compared to those who lived with family (median OS, 6.3 years; 95% CI, 3.9-8.3) and those who lived with a spouse or partner (median OS, 7.1 years; 95% CI, 56.3-8.6; P < .001), but there was no significant difference in OS between the latter 2 groups (P = .42).

The HOUSES index was available for 454 patients. Among those, 63 (14%), 71 (16%), 131 (29%), and 189 (42%) were in Q1, Q2, Q3, and Q4, respectively. Patients in Q1 to Q3 had decreased OS (median OS, 5.3 years; 95% CI, 4.5-6.3) compared to patients in Q4 (median OS, 8.3 years; 95% CI, 7.0-9.2; P = .003). However, there was no significant difference in OS among patients in Q1 to Q3.

Patients in Q1 to Q3 were more likely to be frail (44% vs 31%; P = .008) and to have ISS III (41% vs 28%; P = .005) and R-ISS III (19% vs 9%; P = .01) disease than patients in Q4, but there was no difference in age or proportion with anemia, hypercalcemia, or renal failure between the 2 groups.

On univariate analysis, age, frailty, SES, relationship status, R-ISS III stage, early transplantation, and ≥VGPR to first-line induction were all associated with OS. On multivariate analysis including all variables, older age, frailty, and R-ISS III were independently associated with decreased OS; early transplantation, achieving ≥VGPR to first-line induction, and married/committed relationship status were independently associated with improved OS. The HOUSES index quartile group (Q4 vs Q1-Q3) was not associated with survival. The OS HR for frailty was 1.5 (95% CI, 1.1-2.1; P = .008). When the analysis was restricted to patients aged ≥60 years, the HR was 1.6 (95% CI, 1.1-2.4; Table 3). The proportional hazards assumption was satisfied for both the multivariate model and individual variables.

Data on fatigue, pain, and QOL at the time of diagnosis were available for 207, 206, and 206 patients, respectively. Patients who were frail reported significantly higher scores for fatigue (median, 6 vs 3; P < .001) and pain (median, 5 vs 2; P < .001) and lower scores for QOL (median, 5 vs 7.5; P < .001).

Patients who were not married/in a relationship had higher pain scores (median, 5 vs 3; P = .03) than those who were, but there were no significant differences in fatigue or QOL scores between the 2 groups. Similarly, patients in Q1 to Q3 for SES had higher pain scores (median, 4.5 vs 3; P = .049) than patients in Q4, but there was no difference in fatigue or QOL scores between the 2 groups. There was no significant difference in fatigue, pain, or QOL scores at diagnosis between patients who lived alone vs those who lived with a partner or family.

The FI at 3, 6, and 12 months was available for 21, 99, and 109 patients, respectively. At 3 to 6 months, 70% of patients (n = 83) had no change in status, 25% of patients (n = 30) had a deterioration (nonfrail to frail), and 6% of patients (n = 7) had improvement in status (frail to nonfrail). At 12 months, 68% of patients (n = 74) had no change in status, 2% patients (n = 2) had improvement, and 30% of patients (n = 33) had deterioration in status (Figure 4). Frailty at 6 months and 12 months were both associated with decreased OS; the HRs were 3.4 (95% CI, 1.9-6.3; P < .001) and 1.7 (95% CI, 1.0-3.0; P = .04), respectively.

The conventional method for risk stratification in MM has primarily focused on disease-based characteristics. However, there has been a growing emphasis on the role of frailty as a predictive factor for adverse outcomes in older patients with MM with the development of several frailty scoring systems. In this study, we adopted the cumulative deficit approach to characterize frailty in a real-world cohort of patients with newly diagnosed MM, the majority of whom received novel therapies. Approximately 40% were classified as frail and had decreased PFS and OS compared with nonfrail patients. As expected, the prevalence of frailty increased with age, but 28% of those aged <65 years were classified as frail and had a trend toward decreased OS, although not reaching statistical significance. Equally important, among patients aged ≥75 years, 38% were nonfrail and had longer survival than frail patients, emphasizing that advanced age alone is inadequate to classify patients as frail. Indeed, patients classified as frail based solely on age (>80 years) using the IMWG index have been shown to have better outcomes than those aged >80 years who meet additional frailty criteria. This highlights the importance of distinguishing between these groups.³⁷ In addition, the IMWG system has not been shown to perform well in the subset of patients aged >75 years.³⁸ Although the incorporation of ECOG PS has the potential to increase uptake of frailty tools in practice, this measure does not always correlate with patient-reported functional status.^16,17 This was reflected in our findings, in which 29% of patients with ECOG PS <2 were frail and had inferior outcomes compared with nonfrail patients within this group.

The cumulative deficit approach has been adopted previously to define frailty in MM using population-based registries, classifying patients into ≥2 groups with different outcomes.^18,23,39 Given the nature of these studies, disease-specific characteristics, including stage and cytogenetic profile, were not available. Our study also spans a more recent treatment era (up to 2018), and we demonstrate the prognostic impact of frailty both before and after 2013, and in different groups based on induction regimen. In this study, we observed that frail patients were older and had higher disease stage, a slightly higher rate of high-risk FISH abnormalities, and lower probability of early transplantation than nonfrail patients. However, the prognostic impact of frailty was independent of these factors.

Although the impact of baseline frailty in MM has been previously studied, there are limited data on changes in frailty status after treatment. In this study, we observed that the majority of patients did not experience changes in frailty status at 3 to 12 months compared with baseline. Approximately a third of patients experienced a change in frailty status, predominantly a shift from nonfrail to frail. This is attributed to patient-reported deficits in activities of daily living and instrumental activities of daily living at those 2 time points, likely secondary to treatment side effects. A large population-based study by Mian et al previously evaluated changes in frailty status at 1 to 3 years from diagnosis using a cumulative deficit FI; at 1 year, only 16% had improvement in status, whereas a third experienced deterioration, which is consistent with our results. Changes in frailty status still occurred after 3 years. The updated status was more predictive of outcomes than baseline frailty status.¹⁸ 

In our study, data to calculate the FI at later time points were only available for a subset of patients, which can introduce a risk of selection bias. However, both studies together underscore the dynamic nature of frailty and highlight the importance of reassessing frailty during treatment.

Because studies assessing the impact of frailty in MM have largely focused on treatment toxicity and survival, the association with patient-reported outcomes is underexplored. Patients who were frail had higher symptom burden, reporting higher scores for fatigue and pain, and lower scores for QOL than nonfrail patients. These have been shown to be the most common symptoms reported in newly diagnosed patients with MM and associated with increased health care utilization, dose reduction, and premature treatment discontinuation.^40-45 This emphasizes the need for aggressive symptom management, especially in frail patients.

In addition to frailty, social support and SES are important factors that need to be considered when managing patients with MM. A large population-based study previously showed that marriage was associated with improved cancer-specific survival and OS in patients with MM independent of age, SES, and educational status, and the impact was more pronounced in males.²⁶ Our study also showed that having a partner (married or committed relationship) was associated with improved survival, but the prognostic impact was more pronounced in females. In addition, we observed that patients who did not have a partner were more likely to be frail, have higher disease stage, and increased pain, which may be, at least in part, related to delayed diagnosis. We also demonstrated that the prognostic impact of relationship status was independent of age, disease stage, treatment response, transplant status, SES, and frailty, which highlights the importance of social support. Given the low sample size, with 81% of patients being married, we could not compare outcomes between divorced and widowed individuals.

The impact of SES on outcomes of patients with MM has been explored in several studies, mainly using neighborhood characteristics and insurance type, with mixed results.^27-29,46 Using individual housing characteristics, we observed that patients belonging to the highest quartile for SES had improved survival and were less likely to be frail, but this association was not statistically significant when adjusting for age and disease stage. However, our cohort, which includes patients seen at a single tertiary care center, may not provide an adequate representation of underserved populations. Nevertheless, the management of all patients with MM should take into consideration both disease-specific and patient-specific factors and include an assessment of physiological status, financial resources, transportation, and social support.

This study is limited by its retrospective nature and potential for selection bias by including only those patients with available requisite data for calculating the FI. Our cohort also reflects patients who were able to obtain care at a tertiary institution. Therefore, it is likely that the proportion of frail patients is higher in the general population.

Despite these limitations, our study has several strengths: it is based on a real-world population, includes both transplant-eligible and -ineligible patients, and relies of patient-provided information for functional status. The availability of data on disease characteristics, including cytogenetics, transplant status, and treatment response, allowed us to evaluate the prognostic impact of frailty, social support, and SES after controlling for these variables. Importantly, we reported frailty status at different time points for a subset of patients. In addition to survival, this study demonstrates an association between frailty and symptom burden derived from patient-completed questionnaires.

As the prognostic impact of frailty is increasingly appreciated, its assessment in routine practice remains a major barrier to frailty-adapted care. When selecting a frailty tool, one size does not fit all, and the choice should take into consideration the available resources. However, it is important to use patient-provided information to assess functional status, which is more accurate than physician assessments. Our index relied on a simple and short questionnaire on functional status that can be adopted in routine practice. Online calculators can increase the efficiency and uptake of frailty assessment tools in clinical practice. If resources permit, automated scoring systems using electronic medical record-derived data on functional status and comorbidities can greatly increase the uptake of these tools in practice. It is important to highlight that although most frailty tools classify patients into ≥2 discrete groups, individuals may exhibit various degrees of frailty, and interventions should be tailored accordingly.

Frailty, defined by a cumulative deficit index, is associated with adverse disease characteristics and higher symptom burden in patients with newly diagnosed MM. It is also associated with increased mortality, independent of age, disease stage, transplant status, and treatment response. Although frailty is more prevalent in older patients, a subset of patients aged <65 years are frail at presentation. The physician-assessed PS is inadequate for measuring functional status and should be complemented with patient-provided information. In addition to frailty, social support is independently associated with inferior outcomes and should be considered when approaching newly diagnosed patients. Although the impact of SES on survival is not yet established, an evaluation of financial resources remains an integral aspect of holistic care.

This study was made possible using the resources of the HOUSES program of the Precision Population Science Laboratory of the Mayo Clinic.

The content of this article is solely the responsibility of the authors and does not represent the official views of the HOUSES program.

Contribution: N.A. and S.K.K. designed the study, collected and analyzed the data, and wrote the first draft of the manuscript; P.D. and B.L. performed the statistical analysis; G.S. contributed to data extraction; and A.K., T.M., S.A., M.W., A.D., F.K.B., R.W., J.C., M.Q.L., S.H., M.A.G., and S.V.R. provided critical revision and final approval of the manuscript version for publication.

Conflict-of-interest disclosure: A.D. received research funding from Celgene, Millennium Pharmaceuticals, Pfizer, and Janssen, and a travel grant from Pfizer. M.A.G. served as a consultant for Millennium Pharmaceuticals and received honoraria from Celgene, Millennium Pharmaceuticals, Onyx Pharmaceuticals, Novartis, GlaxoSmithKline, Prothena, Ionis Pharmaceuticals, and Amgen. M.Q.L. received research funding from Celgene. S.V.R. received grants from the National Institutes of Health and research funding from Celgene for clinical trials. S.K.K. served as a consultant for Celgene, Millennium Pharmaceuticals, Onyx Pharmaceuticals, Janssen, and Bristol Myers Squibb, and received research funding from Celgene, Millennium Pharmaceuticals, Novartis, Onyx Pharmaceuticals, AbbVie, Janssen, and Bristol Myers Squibb. The remaining authors declare no competing financial interests.

Correspondence: Shaji K. Kumar, Division of Hematology, Mayo Clinic, 200 First St SW, Rochester, MN 55905; email: kumar.shaji@mayo.edu.