Introduction

Multiple myeloma (MM) is the third most common hematologic malignancy worldwide [1], and primarily affects adults aged 65 years or older [2]. Patients with MM, including those with relapsed/refractory disease, often report significant impairment in health-related quality of life (HRQOL) due to disease-related symptoms like fatigue, pain, and reduced physical function [3,4,5]. Although many novel MM agents have shown improved efficacy and longer overall survival [6,7,8], evidence-based studies evaluating the impact of MM therapies on patients’ HRQOL have been limited [9]. HRQOL data has become increasingly important in guiding medical practice for MM [10, 11].

Carfilzomib is an irreversible proteasome inhibitor. The combination of carfilzomib (56 mg/m2) with dexamethasone has been approved for the treatment of relapsed or refractory MM (RRMM) [12]. Under this approval, carfilzomib is administered twice-weekly, on 2 consecutive days, intravenously (IV) over 30 min for 3 weeks of a 4-week cycle [12, 13]. Previous studies with bortezomib, a previous generation proteasome inhibitor, have demonstrated that MM patients treated with a once-weekly bortezomib dosing regimen were more adherent, had fewer dose reductions, and were treated for longer durations compared with patients on the twice-weekly schedule [14, 15]. Once-weekly dosing can substantially reduce patient travel and wait time, as well as caregiver time [16,17,18,19,20,21,22]. Improving the convenience of treatment is expected to improve HRQOL by allowing more time and opportunity for upholding social and familial roles [5]. Overall, a once-weekly Kd regimen may have value to patients beyond an improvement in treatment outcomes deriving from greater concordance with treatment regimens [23].

The phase III A.R.R.O.W. trial investigated the efficacy of once-weekly vs. twice-weekly Kd dosing schedules in patients with RRMM. A.R.R.O.W. demonstrated that once-weekly carfilzomib at 70 mg/m2 plus dexamethasone (Kd70 mg/m2) significantly improved progression-free survival (PFS) compared with twice-weekly carfilzomib at 27 mg/m2 plus dexamethasone (Kd27 mg/m2; median, 11.2 months vs. 7.6 months; hazard ratio, 0.69 (95% confidence interval (CI), 0.54–88); p = 0.0029) [23]. In addition, once-weekly Kd70 mg/m2 was associated with longer treatment exposure compared with twice-weekly Kd27 mg/m2, while exposure-adjusted discontinuation rates were similar between arms [23]. Thus, by increasing convenience with the utilization of the appropriate carfilzomib dose, A.R.R.O.W. demonstrated that patients stayed on once-weekly Kd70 mg/m2 longer and derived additional benefit without incremental risk compared with twice-weekly Kd27 mg/m2. Based on the results of the A.R.R.O.W. study, once-weekly Kd70 mg/m2 was recently approved for the treatment of patients with RRMM [12]. Patient-reported outcomes (PROs) were specified as exploratory endpoints in A.R.R.O.W.

Here we present patient-reported convenience and satisfaction and patient-reported HRQOL data of the once-weekly Kd70 mg/m2 and twice-weekly Kd27 mg/m2 treatment arms to assess from a patient perspective the impact of once-weekly Kd70 mg/m2 on convenience and quality of life (QOL). The following key subscales were prespecified for analysis: European Organization for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire Core Module (QLQ-C30) Global Health Status (GHS)/QOL, pain, fatigue, physical functioning, insomnia, and role functioning subscales; disease-specific MM questionnaire (QLQ-MY20) disease symptoms, side effects of treatment, and future perspective subscales; and European Quality of Life-5 Dimensions-5 Levels (EQ-5D-5L) index score and visual analog scale (VAS) score.

Materials and methods

Patients and study design

A.R.R.O.W. (Clinicaltrials.gov NCT02412878) was a randomized, open-label, phase III trial. Adult patients with RRMM who were previously treated with 2–3 prior lines of therapy, including a proteasome inhibitor and an immunomodulatory agent, were included. Patients were recruited from approximately 118 sites across North America, Europe, and Asia. Full trial details have been published previously [23].

Eligible patients were randomized (1:1) to either once-weekly Kd70 mg/m2 or twice-weekly Kd27 mg/m2. Randomization was stratified according to International Staging System stage (1 vs. 2 or 3), refractoriness to bortezomib treatment (yes vs. no), and age (<65 years vs. ≥65 years). The once-weekly Kd70 mg/m2 arm received carfilzomib (30-min intravenous (IV) infusion) on days 1, 8, and 15 of all cycles (20 mg/m2 on day 1, cycle 1; 70 mg/m2 thereafter). The twice-weekly Kd27 mg/m2 arm received carfilzomib (10-min IV infusion) on days 1, 2, 8, 9, 15, and 16 (20 mg/m2 on days 1 and 2 during cycle 1; 27 mg/m2 thereafter). Dexamethasone at 40 mg was given to all patients on days 1, 8, 15 (all cycles), and 22 (cycles 1–9 only). Study treatment was administered in 28-day cycles, and cycles were repeated until disease progression, withdrawal of consent, or unacceptable toxicity.

The study protocol was approved by the institutional review boards or ethics committees of all participating institutions, and all patients provided written informed consent.

HRQOL assessments and endpoints

PROs were assessed with patient-reported convenience and satisfaction questionnaires, the EORTC QLQ-C30 [24], the EORTC QLQ-MY20 [25, 26], and the EQ-5D-5L index score and its VAS score [27]. Clinical outcome assessments were performed to measure these PROs. Patient-reported convenience and satisfaction with the carfilzomib dosing schedule were measured with two single-item 5-point Likert scales, with higher scores indicating a higher level of convenience/satisfaction. The validity, reliability, and relevance of the QLQ-C30 and QLQ-MY20 have been specifically demonstrated for patients with MM [26,27,28,29,30]. Further details regarding the subscales and scoring for the QLQ-C30, QLQ-MY20, and EQ-5D-5L are provided in the Supplemental Methods.

The minimal important difference (MID) represents the smallest clinically meaningful, group-level difference in a PRO score [31]. For the QLQ-C30, MIDs were prespecified in the statistical analysis plan based on the evidence-based interpretation guidelines for comparison between groups [32] and in accordance with others in the myeloma population [33] (Table 1). The QLQ-MY20 currently has no robust published estimates for MIDs; therefore, the standard error of measurement using Cronbach’s alpha at baseline was used as a proxy [34]. For the EQ-5D-5L analysis, a MID of 0.037 for the index score [35] and 7 for the VAS score [36] were used.

Table 1 EORTC QLQ-C30, EORTC QLQ-MY20, and EQ-5D-5L minimally important differences and responder definitions
Full size table

The responder definition is a threshold for defining meaningful change over time within an individual [37]. For the QLQ-C30 and QLQ-MY20, these were defined based on the minimal possible change on each scale, and the next increment as sensitivity analysis, in line with Cocks and Buchanan [38] (Table 1). As there are currently no separate recommendations for EQ-5D-5L, the MID was applied as a responder definition.

The QLQ-C30, QLQ-MY20, and EQ-5D-5L questionnaires were administered prior to study treatment on day 1 of cycle 1 (i.e., baseline), then every other cycle during treatment. For patients who discontinued therapy prior to progression, these questionnaires were administered during the follow-up period before disease progression every 12 weeks. Patient-reported convenience and satisfaction questionnaires were collected on day 1 of cycle 2 and end of treatment only.

Statistical analyses

The intent-to-treat population was used for analyses of missing data. The safety population, defined as all randomized patients who received at least one dose of carfilzomib or dexamethasone, was used for analyses of patient-reported convenience and satisfaction. All other analyses were based on the HRQOL analyses set, which included all randomized patients who completed at least one QLQ-C30, QLQ-MY20, EQ-5D-5L, or patient-reported convenience and satisfaction questionnaire.

Completion rates were calculated for all randomized patients at all HRQOL assessment visits, and for all patients expected to have an assessment (including randomized patients who were still in the study at that visit). HRQOL scores were plotted for each treatment arm to assess trends by missing data pattern.

At each assessment, the proportion of patients improved, stable, or worsened relative to baseline according to responder definitions was summarized for the QLQ-C30, QLQ-MY20, and EQ-5D-5L. Patient-reported convenience and satisfaction was compared between once-weekly Kd70 mg/m2 and twice-weekly Kd27 mg/m2 dosing schedules using multivariable binary and ordinal logistic regression adjusting for randomization stratification factors (Supplemental Materials).

A mixed model for repeated measures was used to compare PRO subscales between treatment arms. This primary model included treatment, continuous time, and randomization stratification factors as fixed effects, and a random intercept and slope. Baseline PRO scores were constrained to a common mean between treatments [39]. Least squares mean differences between treatment arms were estimated from the model. A sensitivity analysis tested the robustness of the primary mixed model for repeated measures to deviation from the missing at random assumption by jointly modeling longitudinal scores and time until last PRO assessment using a random effects association structure [40, 41]. In an additional post-hoc analysis, a treatment-by-time interaction was added to the primary mixed model for repeated measures as a fixed effect, allowing treatment effect to vary over time rather than forced constant. Least squares means by time point were plotted.

Time to deterioration (TTD; i.e., time from randomization until the first deterioration in PRO score meeting the responder definition corresponding to that score) was analyzed using a Cox regression analysis that accounted for the randomization stratification factors and baseline PRO scores. Hazard ratios were estimated from the stratified Cox regression model, p values from a stratified log rank test, and median TTD from unstratified Kaplan–Meier. Patients with no data or no baseline data were censored at the randomization date, and patients with no post-baseline data were censored at the randomization date +1 day.

Data sharing

Qualified researchers may request data from Amgen clinical studies. Complete details are available at the following: http://www.amgen.com/datasharing.

Results

Patient population

In the randomized phase III A.R.R.O.W. study, a total of 478 patients were randomized to treatment (once-weekly Kd70 mg/m2, n = 240; twice-weekly Kd27 mg/m2, n = 238) from September 9, 2015 to June 15, 2017. Among the randomized patients, a total of 469 had at least one post-baseline PRO assessment before end of treatment and were included in the PRO analyses (Fig. 1); once-weekly Kd70 mg/m2, n = 235; twice-weekly Kd27 mg/m2, n = 234). Baseline characteristics and mean baseline scores for the QLQ-C30, QLQ-MY20, and EQ-5D-5L subscales were similar between the once-weekly Kd70 mg/m2 and twice-weekly Kd27 mg/m2 arms (Table 2).

Fig. 1
figure 1

Patient disposition (CONSORT diagram) (The trial profile for the A.R.R.O.W. intent-to-treat population has been previously published [24]). EORTC QLQ-C30 European Organization for Research and Treatment of Cancer Quality of Life Core Module, EORTC QLQ-MY20 EORTC Quality of Life Multiple Myeloma Module 20, EQ-5D-DL European Quality of Life–5 Dimensions–5 Levels, Kd carfilzomib and dexamethasone, PRO patient-reported outcome

Full size image
Table 2 Baseline patient characteristics and PROs
Full size table

Completion and missing data patterns

QLQ-C30 completion rates at baseline were similar between the once-weekly Kd70 mg/m2 (90.4%) and twice-weekly Kd27 mg/m2 (92.9%) treatment arms (Table 3). Completion rates were higher in the once-weekly vs. twice weekly arm from cycle 9 onwards except at end of treatment. Similar completion rates were observed for the QLQ-MY20 and EQ-5D-5L questionnaires. For the convenience and satisfaction questions, completion rates were similar at cycle 2 day 1 and at end of treatment, respectively (once-weekly Kd70 mg/m2, 82.1 and 36.3%; twice-weekly Kd27 mg/m2, 81.9 and 44.1%). The proportion of patients completing the questionnaire (of those expected) were also similar between the arms. For the QLQ-C30, in the once-weekly Kd70 mg/m2 and twice-weekly Kd27 mg/m2 treatment arms, respectively, 76.7 and 79.4% of patients had a baseline and at least one post-baseline assessment; 5.8 and 3.4% had missing baseline assessment. Similar completion rates were observed for the QLQ-MY20 and EQ-5D-5L. Across the EORTC QLQ-C30 and EORTC QLQ-MY20 subscales and the EQ-5D-5L scores, there was no clear trend prior to dropout in either treatment arm for patients to be improving or declining.

Table 3 Extent of missing EORTC QLQ-C30 questionnaires (intent-to-treat)
Full size table

QLQ-MY20 MID

Cronbach’s alpha for the QLQ-MY20 subscales were all above 0.7, indicating acceptable internal consistency [42]. The standard error of measurement was 9 for disease symptoms, 10 for future perspectives, and 7 for side effects of treatment. These standard errors of measurement were similar to those observed in previous studies [3, 43].

Descriptive analyses

The proportion of patients with improved scores from baseline was consistently higher (across most visits according to the two specified responder definitions) in the once-weekly Kd70 mg/m2 arm compared with the twice-weekly Kd27 mg/m2 arm in QLQ-C30 physical functioning, role functioning, and fatigue; QLQ-MY20 disease symptoms; and EQ-5D-5L VAS and index score. Other subscales exhibited no clear and consistent differences across the arms.

Patient-reported convenience and satisfaction

A greater proportion of patients in the once-weekly Kd70 mg/m2 vs. twice-weekly Kd27 mg/m2 arm reported convenience of the carfilzomib dosing schedule (very convenient/convenient/neutral; 82.4% vs. 65.8%). Additionally, patients in the once-weekly Kd70 mg/m2 arm reported greater convenience than patients in the twice-weekly Kd27 mg/m2 arm after adjustment for randomization stratification factors (OR, 4.98 (95% CI, 2.54–9.77); p < 0.001) for groupings of very convenient/convenient/neutral vs. inconvenient/very inconvenient. The once-weekly Kd70 mg/m2 regimen was also associated with higher odds of reporting convenience in ordinal logistic regression (OR, 3.03 (95% CI, 2.07–4.45); p < 0.001); results of other models are in Table 4.

Table 4 Logistic regression and ordinal logistic regression results for patient-reported convenience and satisfaction
Full size table

The proportion of patients reporting satisfaction (very satisfied/satisfied/neutral) was numerically higher for the once-weekly Kd70 mg/m2 arm than the twice-weekly Kd27 mg/m2 arm (84.7% vs. 79.6%). This trend was also observed in a logistic regression model adjusting for randomization stratification factors for once-weekly Kd70 mg/m2 vs. twice-weekly Kd27 mg/m2 (OR, 2.41 [95% CI, 0.97–6.01]) for groupings of very satisfied/satisfied/neutral vs. dissatisfied/very dissatisfied. However, the once-weekly Kd70 mg/m2 regimen was associated with higher odds of reporting satisfaction in ordinal logistic regression, in which satisfaction was considered a categorical variable with five ordered values (very satisfied/satisfied/neutral/dissatisfied/very dissatisfied) (OR, 1.61 (95% CI, 1.10–2.36); p = 0.014); results of other models are in Table 4.

Treatment arm differences and TTD for QLQ-C30, QLQ-MY20, and EQ-5D-5L subscales

The primary mixed model for repeated measures did not demonstrate clinically meaningful differences in any scores. These findings were corroborated by the joint model. For all subscales of the QLQ-C30, QLQ-MY20, and EQ-5D-5L, the treatment-by-time interaction term two-sided p values were >0.05, indicating that treatment effect did not vary over time. When estimating mean differences between arms at each time point based on the mixed model for repeated measures with treatment-by-time interaction, the once-weekly Kd70 mg/m2 treatment arm showed clinically meaningful improvement in QLQ-C30 GHS/QOL scores at cycle 15 (Supplemental Fig. 1) and in the EQ-5D-5L index score at cycles 7, 9, 11, 13, and 15 (Supplemental Fig. 2). Improvements with once-weekly Kd70 mg/m2 vs. twice-weekly Kd27 mg/m2 were also estimated with a p value < 0.05 at multiple time points for QLQ-C30 GHS/QOL, fatigue, pain, physical functioning, insomnia, and role functioning; QLQ-MY20 disease symptoms; and EQ-5D-5L index and VAS scores; however, these differences were not clinically meaningful as they did not reach the prespecified MIDs (Supplemental Table 1; Supplemental Fig. 3).

TTD was compared between treatment arms within each of the key subscales for QLQ-C30, QLQ-MY20, and EQ-5D-5L. There was a trend for longer median TTD for once-weekly Kd70 mg/m2 vs. twice-weekly Kd27 mg/m2 arms for QLQ-C30 fatigue (5.6 vs. 3.8 months; HR 0.79 (95% CI, 0.59–1.05), p = 0.035), QLQ-MY20 Disease Symptoms (14.8 vs. 7.4 months; HR 0.67 (95% CI, 0.48–94), p = 0.008), EQ-5D-5L index score (12.9 vs 5.6 months; HR 0.58 (95% CI, 0.42-80), p = 0.002), and EQ-5D-5L VAS score (median not reached vs. 7.4 months; HR, 0.75 (95% CI, 0.54–1.05); p = 0.031) (Fig. 2). In the once-weekly Kd70 mg/m2 and twice-weekly Kd27 mg/m2 arms, respectively, 34.2% and 40.3% of patients experienced a deterioration event of ≥5 points (responder definition) for the QLQ-C30 GHS/QOL subscale. Median (95% CI) TTD for GHS/QOL was 7.4 months (3.9–not estimable) in the once-weekly Kd70 mg/m2 arm vs. 3.8 months (3.7–5.8) in the twice-weekly Kd27 mg/m2 arm. TTD for GHS/QOL per Cox regression was similar between treatment arms (HR, 0.77 (95% CI, 0.57–1.04); p = 0.090). TTD was similar between the two treatment arms for the remaining key subscales of the QLQ-C30 and QLQ-MY20. Analyses based on a larger responder definition (Supplemental Table 1) demonstrated similar TTD between treatment arms for all key subscales.

Fig. 2
figure 2

Kaplan–Meier plots for time to first PRO deterioration. a EORTC QLQ-C30 Fatigue (Minimal important difference (MID) = 6). b EORTC MY-20 disease symptoms (MID = 9). c EQ-5D-5L utility index score (MID = 0.037). d EQ-5D-5L VAS score (MID = 7) EORTC QLQ-C30 European Organization for Research and Treatment of Cancer Quality of Life Core Module, EORTC QLQ-MY20 EORTC Quality of Life Multiple Myeloma Module 20, EQ-5D-DL European Quality of Life–5 Dimensions–5 Levels, Kd27 mg/m2 carfilzomib at 20/27 mg/m2 plus dexamethasone, Kd70 mg/m2 carfilzomib at 20/70 mg/m2 plus dexamethasone, PRO patient-reported outcome, VAS visual analog scale

Full size image

Discussion

A.R.R.O.W. is the first study evaluating PROs with once-weekly, IV proteasome inhibitor dosing, and as such provides valuable information regarding HRQOL associated with this regimen. In the primary A.R.R.O.W. analysis, once-weekly Kd70 mg/m2 significantly prolonged PFS vs. twice-weekly Kd27 mg/m2 (median PFS, 11.2 months vs. 7.6 months; HR, 0.69 (95% CI, 0.54–83); p = 0.0029). Overall safety was comparable between groups, although the rate of grade ≥ 3 treatment-emergent adverse events (TEAEs) was higher in the once-weekly Kd70 mg/m2 (68%) vs twice-weekly Kd27 mg/m2 (62%) arm [23]. Patients in the once-weekly Kd70 mg/m2 arm had longer treatment exposure with similar rates of exposure-adjusted discontinuation compared with the twice-weekly Kd27 mg/m2 arm [23]. Our study complements these results by demonstrating that the value of once-weekly Kd70 mg/m2 dosing derives from delayed worsening of symptoms and greater convenience compared with twice-weekly Kd27 mg/m2, which may allow for more reliable, accurate, and durable concordance with the prescribed dosing regimen.

In our analysis, the overall differences between the once-weekly Kd70 mg/m2 and twice-weekly Kd27 mg/m2 arms did not reach clinical significance for any of the HRQOL key subscales when assessing mean scores over time in a mixed model for repeated measures. However, TTD in PRO scales was longer in the once-weekly Kd70 mg/m2 arm compared with the twice-weekly Kd27 mg/m2 arm for the QLQ-C30 fatigue, QLQ-MY20 disease symptoms, and EQ-5D-5L index and VAS scores. The longer TTD associated with once-weekly Kd70 mg/m2 might be correlated with the longer PFS in this arm, indicating that the achievement of PFS benefit can translate into longer maintenance of baseline QoL. In both treatment groups, median TTD for GHS/QOL was approximately 4 months shorter than median PFS, suggesting that PFS may overestimate the time of maintained QOL. In addition, after adjusting for randomization stratification factors, patients in the once-weekly Kd70 mg/m2 arm reported greater convenience and higher satisfaction than patients in the twice-weekly Kd27 mg/m2 arm, likely due to the less frequent dosing schedule with the Kd70 mg/m2 dose. By delaying HRQOL deterioration and improving convenience, the once-weekly administration of Kd70 mg/m2 can provide patients with a greater chance to participate in meaningful activities and to maintain a social role, both crucial factors in the HRQOL of MM patients [5].

We note that the higher incidence of grade ≥ 3 TEAEs in the once-weekly Kd70 mg/m2 arm in the primary A.R.R.O.W. analysis [23] did not translate into worse HRQOL in our study. The AE profile and QOL may not align because AEs and PRO instruments measure different aspects of the disease experience [44,45,46], and AEs are clinician-reported whereas our analysis focuses on PROs. Furthermore, the recall period of the questionnaires (1 week) differs from the more extensive nature of AE data collection.

Previous studies have shown that patients with RRMM generally have a high symptom burden and low HRQOL [3,4,5, 9]. Functional impairment from disease- and treatment-related symptoms, as well as the overall burden of a terminal diagnosis, can greatly affect HRQOL [3]. In a qualitative study of patients with relapsed MM who received bortezomib or thalidomide treatment, physical and cognitive functioning were reduced due to disease symptoms and treatment-related toxicity, with persistent peripheral neuropathy being considered especially burdensome [5]. HRQOL was reduced in these patients due to concerns about underlying disease, disability, and relapse.

As patients with MM are experiencing longer overall survival with novel agents [6,7,8], with 5-year relative survival rates now over 60% in patients under 70 years of age [7], improving quality of life has become an increasingly important treatment goal of MM therapy. Despite the complementary value of HRQOL when assessing efficacy and safety outcomes of novel MM agents, inconsistencies and weaknesses in HRQOL data analysis and presentation complicate the interpretation of treatment impact on HRQOL [10]. In a systematic review of HRQOL in longitudinal studies of patients with MM, HRQOL improvements were more likely to occur during first-line than in relapsed treatment regimens [11]. Thus, it is critical that MM therapies not only improve efficacy and have tolerable side effects, but also that HRQOL is improved or maintained, particularly in relapsed forms of MM.

Despite the clear association between MM symptoms and HRQOL, there remains a paucity of evidence-based data regarding the impact of MM therapies on symptom burden and HRQOL of patients with RRMM [9]. In general, proteasome inhibitor-containing regimens have been found to maintain or improve HRQOL in patients with RRMM compared with control or standard of care treatments [3, 47,48,49]. An HRQOL analysis of the landmark phase 3 ASPIRE trial found that carfilzomib in combination with lenalidomide and dexamethasone (KRd) improved clinical outcomes and GHS/QOL compared with Rd alone [3, 50, 51]. In the phase III ENDEAVOR trial, Kd56 mg/m2 was associated with superior clinical outcomes and improved GHS/QOL scores compared with bortezomib and dexamethasone (Vd) [52,53,54]. Patients treated with Kd56 mg/m2 also experienced slower TTD in GHS/QOL, physical and cognitive function, side effects, and neurotoxicity symptoms compared with Vd-treated patients in ENDEAVOR [54]. Our study is an important addition to the current body of literature on HRQOL in MM patients treated with carfilzomib, and builds upon the primary A.R.R.O.W. analysis efficacy and safety findings [24] by demonstrating that the increased convenience of the once-weekly regimen compared with twice-weekly regimen allowed patients to stay on therapy longer and thus produced superior clinical outcomes and prolonged HRQOL. We note that although dose per cycle administered for the once-weekly Kd70 mg/m2 arm is approximately one-third higher than that of the twice-weekly Kd27 mg/m2 control arm, the once-weekly Kd70 mg/m2 dose represents a 37.5% reduction over the standard approved twice-weekly Kd56 mg/m2 dose.

This study has some limitations. Patient-reported convenience and satisfaction were measured using individual items. Given the lack of published MIDs for the EORTC QLQ-MY20, a single distribution-based estimate was used. This needs confirmation in further studies using multiple anchors to estimate MID. As this was an open-label trial, patients were aware of the treatment they were receiving. Despite this, the completion rates for the convenience and satisfaction questions were similar across arms. However, completion rates, out of the patients expected to complete a questionnaire at each time point, for the other PRO measures were higher at later cycles in the once-weekly Kd70 mg/m2 arm. The findings from the mixed model were confirmed using a sensitivity analysis (joint model) and appear to be robust despite the differential dropout between treatment arms and high dropout particularly from cycle 17. There was no clear trend prior to dropout for patients to be either improving or declining in either of the treatment arms across the EORTC QLQ-C30, QLQ-MY20, and EQ-5D-5L. In addition, at this analysis stage, data for TTD have high levels of censoring (>50% for all subscales), and calculation of median TTD was not possible for all subscales. The PRO analyses were preplanned but not adjusted for multiplicity; nominal p values were provided for descriptive purposes. The depth of response may have correlated with PFS and time of maintained QOL in this study; however, this was not within the scope of the statistical analysis plan. We plan to further analyze this phenomenon in a future study.

In conclusion, this study demonstrated delayed disease symptom worsening, as well as greater patient-reported convenience and satisfaction, for the once-weekly Kd70 mg/m2 arm compared with the twice-weekly Kd27 mg/m2 arm. TTD was longer for the once-weekly Kd70 mg/m2 arm in fatigue, physical and role functioning, and health status. In the primary A.R.R.O.W. analysis, once-weekly Kd70 mg/m2 improved PFS over twice-weekly Kd27 mg/m2, with the incidence of adverse events being similar between the treatment groups [24]. Our findings complement the primary A.R.R.O.W. analysis by showing that these improved clinical outcomes led to prolonged HRQOL for the once-weekly regimen. Our results also add to the current knowledge of the efficacy, safety, and HRQOL of carfilzomib-based therapies in RRMM. These results build upon the ENDEAVOR and ASPIRE studies, which showed that carfilzomib-based regimens produced superior clinical outcomes which translated to improved and prolonged HRQOL compared with control regimens in RRMM [3, 50,51,52,53,54]. Collectively, the primary A.R.R.O.W. safety and efficacy data [24] and the current PRO analysis reinforce that once-weekly Kd70 mg/m2 dose is superior and convenient while delivering more favorable HRQOL than the commonly used Kd27 mg/m2 dose. Thus, once-weekly Kd70 mg/m2 should be considered an important alternative to twice-weekly Kd27 mg/m2 in clinical practice.