Continuous or Fixed-Duration Maintenance Therapy in Multiple Myeloma.

INTRODUCTION

Treatment of multiple myeloma (MM) during the last two decades has improved survival, with potential cure in some patients. 1 , 2 During this time, the treatment of newly diagnosed myeloma (NDMM) changed from defined treatment duration with or without autologous stem cell transplant (ASCT) to addition of continuous use of either all drugs or a single drug. 3 – 8 Maintenance therapy is generally continued indefinitely until disease progression but the optimal duration of maintenance therapy is undefined. 3 , 7 – 9 Thus, no data address whether many years of continuous therapy with its attendant toxicities and cost improves outcomes compared with limited-duration maintenance. Most patients cannot continue treatment until progression because of side effects and expense, yet clinicians hesitate to stop maintenance given the uncertainty about optimal duration. We designed this trial to evaluate whether indefinite maintenance therapy improves outcomes in patients with NDMM compared to a predetermined interval of 2 years. This is an important question in the context of MM, as trials have suggested that a higher dosing or longer duration treatment strategy may not always be the optimal approach. 10 – 12 For example, better survival was observed with dexamethasone given once-weekly compared to a more intense dosing regimen, 11 and a planned dose reduction of lenalidomide and glucocorticoid discontinuation improved outcomes compared with continued intense dosing in another trial. 10 Therefore, we designed this phase 3 trial to address two important questions. In the first randomization (R1) we compared carfilzomib, lenalidomide, dexamethasone (KRd) and bortezomib, lenalidomide, dexamethasone (VRd) for initial therapy to compare the anti-myeloma efficacy of two proteasome inhibitors; reported previously. 13 Herein, we present the results of the second randomization (R2) comparing indefinite versus limited duration (2-years) lenalidomide maintenance following induction therapy.

METHODS

PATIENTS

This randomized, open-label phase-3 trial designed to demonstrate superiority of indefinite over limited duration maintenance was conducted by ECOG-ACRIN through the National Clinical Trials Network and enrolled patients between 2013 and 2019. Developed in collaboration with the National Cancer Institute’s (NCI) Cancer Therapy and Evaluation Program (CTEP), the study was approved by central and local ethics committees as required. The trial followed the Declaration of Helsinki and International Council for Harmonization (ICH) Good Clinical Practice (GCP) guidelines, with all patients providing written informed consent. The authors vouch for the data and adherence to the protocol, available at NEJM.org .

Eligibility criteria ( Appendix ) for enrollment and treatment initiation on Step 1 (S1, induction) have been previously described. 13 Briefly, adult patients with NDMM, considered ineligible for or not planning an upfront ASCT, were enrolled. Patients with high-risk MM (defined as one or more of: t(14;16), t(14;20), or Del17p on fluorescence in situ hybridization (FISH), serum lactate dehydrogenase (LDH)>2x the upper limit of normal, ≥20% circulating plasma cells on peripheral blood smear or 2000 plasma cells/microliter in peripheral blood, or high-risk gene expression signature (GEP70 14 , if tested) were excluded given another ongoing high-risk trial. 12 Patients receiving at least 6 VRd-cycles or 4 KRd-cycles for induction without disease progression or starting non-protocol therapy (alternate therapy before documented progressive disease) including ASCT, were eligible to enroll to Step 2 (S2) maintenance, provided any adverse event related to Step 1 therapy was resolved to grade 2 or less.

RANDOMIZATION AND TREATMENT

Patients enrolling to S1 were randomly assigned 1:1 to receive induction therapy with VRd (twelve 3-week cycles) or KRd (nine 4-week cycles); stratified by intent to ASCT at the time of disease progression. 13 Patients meeting eligibility for S2 were randomized (1:1) to limited (2-years) or indefinite-duration lenalidomide maintenance; stratified by induction therapy assigned ( Figure S1 ). Patients received 15mg lenalidomide orally, once-daily on days 1–21 of a 4-week cycle, continued for 2-years or indefinitely until progressive disease. Patients came off treatment for progressive disease, unacceptable toxicity, initiation of non-protocol therapy, or withdrawal of consent.

ENDPOINTS AND ASSESSMENTS

The primary endpoint for the maintenance phase was overall survival, defined as the time from R2 until death due to any cause or follow-up censored at the date last known alive. An additional planned analysis of overall survival censored patients at initiation of non-protocol therapy. A key secondary end point was progression-free survival, primarily defined as the time from R2 until the earlier of progressive disease or death due to any cause, occurring within 3 months of the last disease evaluation. Patients alive without progressive disease were censored at date of last disease evaluation. Additional post-hoc progression-free survival analyses evaluating the impact of non-protocol therapy (as an event at date of non-protocol therapy initiation or censored at date of last disease evaluation progression-free prior to non-protocol therapy) and timing of death (all deaths counted) were conducted. Other secondary end points included response, treatment exposure, quality of life (QoL), adverse events, and second primary cancers. ( Appendix )

Disease response and progression were assessed according to the International Myeloma Working Group (IMWG) uniform response criteria measured from S1 baseline. 15 Disease assessments were performed every three cycles; bone marrow examination was only performed for documentation of a complete response (CR) and after 24-cycles. All patients, including those who discontinued assigned treatment early, were followed until progressive disease and for survival through year-15.

Safety was monitored every 3 cycles and adverse events were graded according to National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE), version 4 or 5. Adverse events reported here occurred through 30 days post treatment. Treatment-emergent adverse events were defined by occurrence on maintenance whether new onset or existing since the last cycle of induction. Treatment-related adverse events were defined by an attribution of possible, probable or definite and serious adverse events were as defined by the protocol. Solicited adverse events (4 hematologic and 14 non-hematologic) required reporting of all grades and any attribution. Otherwise, grades 4–5 hematologic and grades 3–5 non-hematologic events were part of the required reporting.

Functional Assessment of Cancer Therapy (FACT) questionnaires were used assess patient reported health-related QoL. Assessments were performed at end of cycles 6, 12, 24 in both arms and after cycles 30, 36, 48, and 60 in the indefinite-arm and at corresponding time points in the limited-arm after 24 cycles. The primary QoL instrument was the FACT-Neurotoxicity Trial Outcome Index (FACT-Ntx TOI) which is the sum of the physical, functional and neurotoxicity scales (25 items; score 0–100) with a minimally important difference established as 5–7 points. There were three time periods of interest: changes from end of induction to end of cycle-6; from the end of 2-years to 3-years; and from the end of 2-years to 5-years of maintenance.

STATISTICAL ANALYSIS

This study was amended in June 2018 to increase S1 accrual to 1080 patients and to revise downwards the S2 enrollment yield from 85% to 65%. We estimated that a S2 sample size of 395 patients and 204 events at 9 years of follow-up would provide 80% power to detect a 50% improvement of median overall survival from 5 years on the limited-arm to 7.5 years on the indefinite-arm assuming exponential distribution of failures and 2-sided alpha of 5%. Inflated for potential treatment non-compliance (25% Lachin-Foulkes adjustment) 16 , the target sample size was 704 patients with 364 events. Time-to-event comparisons were performed in the intent-to-treat (ITT) population, which included all patients as randomized (R2). Additionally, landmark analyses at 2 years including only patients completing 24 cycles of treatment were conducted for overall survival and overall survival censored at non-protocol therapy. QoL was assessed in patients with evaluable data. Treatment response, treatment exposure, second primary cancers and adverse events were evaluated in the treated population. Survival distributions were estimated with the Kaplan-Meier method. 17 Superiority of the indefinite arm was assessed using the stratified log-rank test. 18 For the landmark analyses, treatment hazard ratios (HR; indefinite/limited) were estimated with the use of stratified Cox proportional hazards regression model. 19 Sequential monitoring of the maintenance overall survival comparison was incorporated. 20 In post-hoc analyses, restricted mean survival time was calculated for overall survival and progression-free survival along with the respective subgroup analyses with follow-up time evaluated primarily at 7-years based on the minimum overall survival or progression-free survival observed time within each arm among all subgroups. 21 , 22 To quantify the treatment effect, the restricted mean survival time ratio (indefinite/limited) and the restricted mean survival difference (indefinite-limited) were estimated. Exposure-adjusted incidence rates of adverse events were based on the sum of treatment duration for all patients by arm expressed as total person-years. Cumulative incidence of second primary cancers and treatment discontinuation were estimated in presence of competing risks. 23 The protocol did not include correction for multiplicity with respect to secondary endpoints and all additional time-to-event analyses; results are reported as point estimates with 95% confidence intervals. The widths of the confidence intervals have not been adjusted for multiplicity and should not be used to infer definitive treatment effects. Analyses were performed using SAS (v9.4) and R (v3.4.3).

RESULTS

PATIENTS

From among the 1087 patients randomized to induction therapy with VRd (n=542) or KRd (n=545) ( Figure-S1 ), 516 (47%) patients proceeded to R2; 260 patients to the indefinite-arm and 256 patients to the limited-arm ( Figure-S2 ), of whom 58% had received KRd and 42% had VRd as induction therapy ( Table-S1 ). The reasons for going off study before S2 are shown in Table-S2 . Demographic and clinical characteristics at R2 were well balanced between the arms ( Table-1 ), and the baseline characteristics of patients at R2 based on R1 assignment is provided in Table S3 . The response status at R2 is provided in Table-2 . While patients with certain high-risk cytogenetic abnormalities were excluded, nearly 35% of patients in each arm had t(4;14), 1q gain, or 1p deletion ( Table-S4 ). Cutoff date for the analysis was September 30, 2025, at which time only 35 (14%) patients on the indefinite-arm were continuing maintenance treatment. Mean (Standard Deviation (SD)) number of maintenance cycles administered was 18.2 (8.0) for limited-arm vs. 38.5 (31.9) for indefinite-arm ( Table-S5 ). All patients assigned to the limited-arm have either completed 2-years of maintenance (57%) or stopped earlier, while 49% in the indefinite-arm have gone off for progression ( Table-S6 , Figure-S3 ). Among patients without progressive disease in the indefinite-arm, 57% received lenalidomide for ≥3 years and 46% for ≥4 years, and the 35 patients still on treatment had a mean (SD) of 88 (18) cycles. At 2-years from R2, similar proportion of patients in the limited and indefinite-arms, respectively, had gone off for progression (22%, 25%), adverse events (10%, 6%), or other reasons. Beyond 2-years, patients in the indefinite-arm went off for progressive disease (17%), adverse events (9%), or patient decision (4%). Treatment exposure and lenalidomide dose over time are shown in Table-S7 ; 45% and 62% on the limited and indefinite-arms, respectively, required dose modifications due to adverse events, including 11% after cycle 24 in the indefinite arm ( Table-S8 ). The median lenalidomide dose at end of 24 cycles was 15mg and 10mg in the limited and indefinite arms, respectively. Non-protocol therapy was initiated in 33 (13%) and 21 (8%) patients in the limited and indefinite-arms, respectively, at a median of 22.6 and 20.9 months from S2 entry ( Table-S9 ). Overall, 60 (23%) and 55 (21%) patients in limited and indefinite arms, respectively, ever underwent ASCT, most for progressive disease.

EFFICACY

At an estimated median follow-up of 86 months, 160 deaths had occurred (43% full information). Of the 80 deaths in each arm, 34 (42.5%) and 33 (41.3%) were attributable to disease in the limited-arm and indefinite-arm, respectively. 7-year overall survival probability (95% CI) was 69.0% (62.3, 74.7) in the limited-arm and 68.6% (62.1, 74.3) in the indefinite-arm, a difference (95% CI) between arms (indefinite-limited) of −0.37 percentage points (−9.03, 8.29) p=0.933 ( Figure-1 ). Overall survival estimates from year-2 through year-7 are shown in Table-S10 . Overall survival censoring patients at non-protocol therapy had 48 fewer events (n=112) and yielded similar results ( Figure-S4 ). Additionally, results of the landmark analyses at 2 years from randomization were similar for both overall survival primary HR=1.06 (95%CI; 0.61, 1.85) and overall survival with non-protocol therapy censored HR=1.12 (95%CI; 0.60, 2.09) ( Figure-S5 ). At the 2 nd interim analysis, when the DSMC recommended release of data in part based on futility monitoring ( Table-S11 ), the conditional probability of a positive outcome was <20%, and the treatment HR approached 1.0, the futility boundary set for 48% information. On average when patients were followed for 7-years, patients in the limited-arm survived 72.8 months and patients in the indefinite-arm survived 72.0 months, with a restricted mean survival time difference (95% CI) of −0.77 (−4.43, 2.89) and a restricted mean survival time ratio (95%CI) of 0.99 (0.94, 1.04) ( Table-S12 ). Overall survival among subgroups is provided in Figure-S6 , Table-S13 , Table-S16 .

At an estimated median progression-free survival follow-up of 80 months, progressive disease or death has been observed in 317 (61%) of 516 randomized patients: 164 (64%) for the limited-arm and 153 (59%) for the indefinite-arm. In the primary progression-free survival analysis, median progression-free survival (95% CI) was similar between the two arms, 38.9 months (31.6, 46.2) for limited versus 42.5 months (33.1, 56.1) for indefinite. 7-year progression-free survival probability (95% CI) was 29.7% (23.5, 36.1) in the limited-arm and 36.1% (29.7, 42.4) in the indefinite-arm, a difference (95% CI) between arms (indefinite-limited) of 6.36 percentage points (−2.63, 15.40). ( Figure-2 , Table-S14 ). On average when patients were followed for 7-years, patients in the limited-arm survived without progression 45.6 months and patients in the indefinite- arm survived without progression 47.0 months, with a restricted mean survival time difference (95% CI) of 1.44 (−4.13, 7.01) and a restricted mean survival time ratio (95%CI) of 1.03 (0.92, 1.16), Table-S12 . In additional post-hoc analyses, based on restricted mean survival time ratio at 7 years of follow-up, the progression-free survival difference appeared greater when accounting for non-protocol therapy, more so when considered an event. ( Figure-S7 , Table-S14 ). Progression-free survival among subgroups is provided in Figure-S8 , Table-S15 , Table-S16 .

Change in response from start of maintenance is shown in Figure-S9 . Best overall response of partial response or better during maintenance was achieved by 239 (94%) (95%CI; 90, 97) of patients in the limited-arm and 236 (93%) (95%CI; 89, 96) in the indefinite-arm; corresponding rates of very good partial response or better were 88% (95%CI; 83, 92) and 84% (95%CI 78, 88), respectively ( Table-2 ). A complete response or better was observed in 125 (49%) (95%CI; 43, 56) of patients in the limited-arm and 119 (47%) (95%CI; 40, 53) of patients in the indefinite-arm.

SAFETY

Grade ≥3 hematologic and non-hematological adverse event rates were 46.9% for treatment emergent events and 33.5% treatment-related events for the limited-arm versus 62.4% and 44.7%, respectively, for the indefinite-arm p=0.009 for the difference in treatment-related events. Corresponding numbers for only non-hematological toxicities were 31.5% treatment-emergent and 16.9% treatment-related on the limited-arm versus 48.2% and 23.5%, respectively, on the indefinite-arm, p=0.064 for the difference in treatment-related events. Treatment-emergent solicited adverse events are shown in Table-3 , Figure-S10 . Treatment-related solicited adverse events are shown in Table-S17 . Exposure-adjusted Grade ≥3 treatment-related adverse event incidence rates are shown in Table-S18 , and exposure-adjusted incidence rates of Grades 1–4 solicited adverse events are provided in Table-S19
(TRAE) , Figure-S10
(TEAE) . Overall, 25 (10%) patients in the limited-arm and 39 (18%) patients in the indefinite-arm discontinued therapy for adverse events, primarily fatigue, anemia, neutropenia and diarrhea. Post-24 cycles exposure-adjusted Grade ≥3 treatment-related adverse events on the indefinite arm was 21/100 person years of treatment. Excluding non-melanoma skin cancers, 26 and 31 cases of second primary cancers were noted in the limited-arm and indefinite-arm, respectively. The related 5-year cumulative incidence in competing risk analysis is 8.3% for the limited-arm and 11.2% for the indefinite-arm ( Figure-3 ). The types of cancers are as shown in Table-S20 and the cumulative incidence with death as a competing risk is shown in Table-S21 .

The QoL compliance rate at the assessment timepoints on treatment were approximately 85% and higher, Table-S22 ), and the scores over time for each of the scales are shown in Tables-S23 – S27 , Figures-S11 – 15 . Definite conclusions regarding the changes in scores in either arm is limited by the small number of observations.

DISCUSSION

Treatment of NDMM has evolved leading to improved survival, with the current treatment approach built on two themes – more intense treatments leading to deeper responses and longer duration of treatment. 24 – 28 As initial therapy moved from two- to three- to four-drug combinations, the response duration has increased, with continuous therapy playing a role; but its optimal duration remains unclear despite toxicities and costs rising with prolonged treatment. In fact, with increasing duration of initial response, patients receive maintenance for longer, making it important to determine if benefits of prolonged therapy justify the added side-effects and incremental cost. Results of the current study suggest that continuous maintenance therapy with lenalidomide demonstrates no clear benefit on overall survival.

The result of the current study reflects the importance of conducting prospective trials to determine optimal treatment duration. When trials demonstrated indefinite maintenance being superior to no maintenance (an “all or none” design), maintenance therapy was quickly adopted but did little to inform us on the optimal duration. The lack of overall survival improvement from indefinite maintenance therapy reflects to some extent the improved efficacy of newer induction therapies, and the newer treatments available at relapse. The results suggest that the benefit from maintenance can be achieved with a 2-year treatment duration, with further treatment only adding toxicity. Our trial shows continued grade 3 and higher toxicity related to indefinite maintenance. We did find a slightly higher incidence of second malignancies at 5-years with indefinite therapy, but lower risk relative to ASCT or alkylating-agent exposure.

Some key considerations may affect the interpretation of our results. First, this study primarily included standard-risk patients and the optimal duration of maintenance in high-risk MM will be informed by ongoing randomized trials. Second, our trial did not incorporate 4-drug induction or ASCT as part of initial therapy, a scenario where the DETERMINATION and IFM2009 trials show different progression-free survival thought to be at least partly explained by the duration of maintenance. However, with deeper response at the initiation of maintenance among transplanted patients and those receiving 4-drug induction, indefinite maintenance seems unlikely to be superior. These results should help physicians evaluate maintenance duration, especially if patients are experiencing toxicity or have had long deep responses. Third, our results show that even if indefinite maintenance is pursued as a strategy, many patients cannot tolerate treatment indefinitely due to low grade, but quality of life disrupting side effects like fatigue, leg cramps, neuropathy, and diarrhea. Moreover, abbreviating the maintenance duration may allow reuse of lenalidomide at relapse, especially in limited-resource settings. Only about half of the initially enrolled patients proceeded to the maintenance step, primarily due to adverse events during induction or a decision to go to alternate therapy, most commonly SCT – these were similar in the 2 induction arms and given the stratification based on induction, this should not affect the results of the maintenance randomization.

Our trial was designed to test fixed versus indefinite-duration without regard to the response depth, but many ongoing trials are using measurable residual disease (MRD) negativity to guide the optimal time to stop therapy. 29
24 While these trials may provide additional clarity on the maintenance duration, the current results remain important as they address the duration of therapy in general in a prospective randomized design and add to the current body of evidence. 29

In summary, this prospective randomized trial directly examined the question of duration of maintenance therapy in NDMM and demonstrated that indefinite-duration maintenance with lenalidomide compared to limited duration (2-years) does not appear to have an overall survival benefit. Our results were in standard-risk MM in patients not receiving SCT, and not receiving a 4-drug induction, making generalization to all MM patients difficult. However, the concept of limited duration therapy, especially as the treatments are becoming more efficacious, the concepts resonate to all MM patients. 30 With longer duration of therapy adding toxicity and cost, studies to define optimal treatment duration must be done in parallel or shortly after new treatments are approved. It should not be assumed that longer duration of therapy is likely to be better. Our results have important implications for the cost of myeloma therapy; the cost of lenalidomide maintenance in the United States is over $1 billion annually. Finally, stopping lenalidomide at 2-years may also allow broader reuse of IMiDs in those patients who relapse after stopping maintenance.

Supplementary Material