Retrospective Study to Compare Outcomes in 159 Patients Undergoing First Autologous Stem Cell Transplantation for Myeloma Treated with Melphalan 140 mg/m or 200 mg/m

Introduction

Multiple myeloma (MM) refers to the malignant proliferation of plasma cells and presents with unique clinical features, including renal failure, lytic bone lesions, bone marrow failure, hypercalcemia, and pathologic fractures [1,2]. As an incurable disease, treatment-induced remissions are almost always followed by relapses after a variable period [2]. The main goal of therapy is to achieve disease control for as long as possible while minimizing toxicity, appreciating that with each relapse, the next remission is more difficult to attain and maintain [3]. Accumulation and selection of genetic aberrations that confer drug resistance and more aggressive behavior is a popular explanation for worsening response rates and remission durations with each line of therapy (LoT) [4]. An increasingly adopted notion is that the depth and duration of the first remission probably correlates with both the overall survival (OS) and quality of life [5].

MM is known to be highly responsive to melphalan, which is used in high doses followed by autologous stem cell transplantation (ASCT) [6]. This method has been the standard of care in newly diagnosed MM for the past 3 decades and is usually offered early, during the first remission, with the aim of prolonging its duration [2,7]. Considering the incurable nature of MM and the importance of the duration of first remission as mentioned above, careful modification to the intensity of anti-myeloma therapies is called for in frail patients. Reducing the dose of melphalan from 200 mg/m2 (Mel200) to 140 mg/m2 (Mel140) has been a widely recognized example for such an adjustment [8]. Mel140 is usually offered to patients with MM with significant comorbidities (renal failure, poor cardiac function) or older age, who are considered to carry a high risk for severe complications with Mel200 [1,9].

Although there is ongoing research to find the optimal conditioning regimen for ASCT in MM, including melphalan combinations with busulfan or bortezomib [10,11], single-agent melphalan remains the standard, with a few papers comparing the outcomes of its 2 common doses: Mel200 and Mel140 [2]. Two studies with very large sample sizes using registry data have reported conflicting results, especially concerning those patients with very good partial remission (VGPR) or complete remission (CR) at the time of ASCT [12,13]. The EBMT registry (CALM study, Auner et al) reported an unexpected association of increased mortality with Mel200 (standard of care [1,2]) among those patients with VGPR or better [12]. These two studies had inherent limitations of database analyses, such as moderate rates of patient exclusion due to missing data and heterogenous local practices of patient management. Single-center studies on this subject in the literature present comparisons between more homogenous groups; however, therapies other than ASCT were mostly outdated limiting interpretation, especially for longer-term data [14–17]. Given the perplexing findings from the literature, we planned to go through our institution’s archives to compare Mel200 with Mel140 for the first, single ASCTs among patients with MM. Since the local practice at our institution is to administer a second or even third LoT if VGPR is not achieved before ASCT, we were confident that our data could offer a decent contribution to the literature for the controversial subgroup of patients with satisfactory responses at the time of ASCT. We decided to keep data limited to a single center, to eliminate confounding factors of differences in local practice and data heterogeneity and to reduce the rate of missing data. To address the conflicting reports from the literature, especially that of the reports of increased mortality with Mel200 among those with VGPR or better, we conducted this retrospective study from a single center in Turkey aimed to compare outcomes of patients with a first ASCT for myeloma treated with melphalan 140 mg/m2 or 200 mg/m2 (Mel200 and Mel140).

Material and Methods

ETHICS STATEMENT AND PATIENT CONSENT:

All patients signed informed consent for the anonymous use of their medical data for research purposes, and the study received ethical board approval of the Clinical Research Ethical Board of Istanbul University-Cerrahpaşa on May 3, 2024, with an approval number of 978389 and confirmation code of BSASFEDYJZ.

STUDY DESIGN AND SAMPLE SIZE ESTIMATE:

This study was designed as a retrospective real-world analysis from a single hematopoietic stem cell transplant center with over 25 years of experience that has established its own local practice principles. The primary endpoint was OS, to address the issues raised in the earlier publication of the EBMT registry [12]. We calculated the sample size using a superiority design with the assumptions of a hazard ratio (HR) for death on the Mel200 arm of 0.5 and a 3: 1 ratio for number of patients between Mel200 and Mel140. Two-sided alpha of 0.05 and a type 2 error rate of 30% were used for calculation. Using preliminary data of the transplant center for 5-year mortality, we calculated the sample size at 172. The yearly number of transplantations for MM in the unit was documented between 15 and 20 (almost all directed from the same institution’s myeloma group) over the last 15 years. Based on the calculations and preliminary information, we planned to collect the data of all patients who received high-dose melphalan with ASCT support between 2012 and 2021, to achieve the target sample size while avoiding earlier transplants to reduce heterogeneity in terms of induction and post-transplant therapies and the availability of supportive measures. We decided on an inclusion cut-off to the end of 2021, to allow at least a 2-year post-transplant follow-up. Second or tandem ASCTs and patients with amyloidosis or POEMS syndrome were excluded.

DATA COLLECTION:

Data on demographic parameters, co-morbidities, renal-cardiac-hepatic organ function, performance status, induction therapies, International Staging System (ISS), Durie Salmon stage, bone marrow plasma cell percentage, genetic aberrations, response status at the time of transplantation, presence of high-risk genetic anomalies, treatments before ASCT, hematopoietic cell transplantation-comorbidity index (HCT-CI), melphalan dose for ASCT, and maintenance after ASCT were planned for collection as independent variables. Time to engraftment, time to discharge, re-hospitalization or intensive care admission rates, time to progression, time to next treatment, time to death, and cause of death were considered outcome measures. Maintenance therapy was categorized as lenalidomide, or others. Induction therapies were categorized as either VCD (bortezomib, cyclophosphamide, dexamethasone), VAD (vincristine, doxorubicin, dexamethasone), VRD (bortezomib, lenalidomide, dexamethasone), or others. We also categorized patients according to response state at ASCT, number of LoTs, and number of total chemotherapy cycles received before ASCT. Progression was defined in accordance with the International Myeloma working Group criteria. Progression-free survival (PFS) was defined as the time initiated at the day of the transplantation till any event of progression or death, and patients without an event at their last follow-up were censored. OS was defined similarly to PFS, except disregarding progression as an event, and censoring all living patients at the time of data accrual. Non-relapse mortality was defined as events of death before a documented disease progression.

STATISTICAL ANALYSIS:

Background data were presented with descriptive statistics, using frequency, mean, and median. The logistic regression model was used to calculate odds ratios (OR) and confidence intervals (CI) for progression of disease at 2 years, and OS at 5 years using melphalan dose, age, sex, high-risk genetic anomaly, response status, and number of LoT as covariates. Time to event analyses were calculated with the Kaplan-Meier method and Cox regression analysis. Patients without progression or death at the time of last evaluation before data cut-off were censored from PFS estimates. Patients alive at the time of last evaluation before data cut-off were censored for OS estimates. The Fisher exact test was used for correlations between background features. HRs with 95% confidence intervals and log-rank were used to determine statistical significance. Cox regression modelling using age, genetic anomalies, number of LoT, disease status before ASCT, ISS stage, and melphalan dose as covariates for multivariate analysis to calculate HR for progression and death. CI of 95% were used for all statistical analysis, considering one-sided P values <0.05 significant. IBM-SPSS version 20 was used for statistical analysis, and plots were re-generated on Microsoft Excel and PowerPoint.

Results

BACKGROUND DATA:

We evaluated the records of 194 consecutive ASCTs for MM in our transplant center within the predetermined study period. Thirty-five ASCTs were excluded, as these were either a second ASCT or performed on patients who also had primary amyloidosis. Three patients (1 from Mel200 and 2 from Mel140) were excluded from time-to-event analyses, as they were lost to follow-up. The data of 156 patients were evaluated for time-to-event outcomes, of whom 130 received Mel200 and 26 received Mel140. The ratio for number of patients between groups was more toward Mel200 than previously anticipated. The target number of OS events was reached; however, the sample size was 10% less than previously calculated. The median age at ASCT was 57 years, median follow-up was 69 months, and most patients (77.6%) had achieved VGPR or CR status before ASCT. As a demonstration of homogeneity in the pre-ASCT care for the cohort, the first LoT was VAD for 87.3% of patients transplanted before 2016 (when bortezomib received reimbursement for frontline use in Turkey) and VCD for 92.1% after 2016. Two or more LoT were administered to 48.1% of patients. For the second LoT, 85% received VCD in the first half of the study period, and 79% received VRD in the latter half, again highlighting the homogeneity. Most patients (86.6%) had received at least 2 courses of VCD or VRD before ASCT. Only about a quarter of patients who were from the second half of the study period received lenalidomide maintenance. The distribution of these background data between the Mel140 and Mel200 groups are presented in Table 1, which demonstrates the 2 key background differences between the 2 groups of either the presence of advanced kidney disease or older age in the Mel140 group. Median HCT-CI was higher for the Mel140 group (3 vs 2), but when renal failure was removed from scoring, the HCT-CI mean and median were similar. For other variables, such as the number and content of prior treatments, lenalidomide maintenance rates, response rates before ASCT, presence of high-risk genetic anomalies, paraprotein type, and sex, the 2 groups were similar, setting the ground for comparing melphalan dosing at ASCT.

PROGRESSION-FREE SURVIVAL:

There was no difference in the median PFS between the 2 groups, with a median time to progression of 42.4 months (95% CI=39.1–45.6) for the Mel140 and 33.1 months (95% CI=26.9–39.4) for the Mel200 groups (HR=0.8 for Mel140, log rank= 0.49). The PFS curves are presented in Figure 1A, demonstrating close estimates up to 7 years of follow-up. A particular subgroup of interest with 121 patients who had achieved VGPR or better before ASCT was also analyzed with regard to PFS, comparing Mel140 and Mel200, and demonstrated remarkably similar estimates (Figure 2A). Logistic regression analysis on 2-year PFS rates was significant only for the presence of high-risk genetic anomaly (OR=4.3, CI=1.9–7.6) and there was a statistically borderline trend for not achieving CR (OR=1.8, CI=0.93–3.1) (Table 2). Cox regression modelling identified a significant HR (HR=2.73, CI=1.8–4.6) for only the presence of high-risk genetic anomaly in multivariate analysis, along with a statistically insignificant trend for higher risk of relapse for those who received 3 or more LoT before ASCT (Figure 1B).

OVERALL SURVIVAL:

The most significant finding and the pre-defined primary endpoint of the study was longer OS with Mel200 over Mel140 (Figure 3A, P=0.002). The median OS estimate was 125.3 months (95% CI=80–170.6) for the Mel200 group vs 62.9 months (95% CI=44.9–77.8) for the Mel140 group (HR= 2.49, 95% CI=1.37–4.53, P= 0.003). There was 1 case of mortality within 100 days of ASCT in both groups. The difference in survival was particularly evident after the event of progression and will be discussed below (Figure 4). The OS superiority of the Mel200 group was firmly maintained among 121 patients with VGPR or CR, the controversial subgroup in the literature (Figure 2B, estimated median OS NR vs 53.7 months 95% CI=NR-NR, 47.3–60.1 P<0.001). Cox regression modeling on OS (Figure 3B) using the same covariates as used in the multivariate analysis on PFS demonstrated significantly higher HR ratios for high-risk genetic-anomalies (HR=5.1, 95% CI=2.6–10.2) and Mel140 (HR=2.6, 95% CI=1.3–5.1). The outcome comparisons including PFS and OS at certain timepoints with logistic regression analysis are outlined in Table 2 and are in line with the results from the multivariate analysis for PFS and OS. Deaths unrelated to MM were rare, with 3.1% and 3.8% in Mel200 and Mel140 groups, respectively (P=0.84), not providing an explanation for the higher rate of mortality in the Mel140 group.

SUBSEQUENT TREATMENTS:

To explain the substantial difference in OS despite similar PFS, we investigated the subsequent treatments between the 2 groups following progression. The second ASCT rate in the Mel200 group following relapse was 26%, clearly higher than the 4.5% in the Mel140 group (P=0.03). Among patients who did not undergo a second ASCT, there was also a statistically borderline trend for the MEL200 group to receive higher rates of carfilzomib or daratumumab-based therapy over patients in the Mel140 group (36.5% vs 14.3% P=0.056). All but 1 patient in the Mel140 group had significant treatment interruptions on carfilzomib or daratumumab due to toxicity, whereas 77.7% of those in the Mel200 group received at least 3 courses of these treatments without interruption (P=0.027). Median OS after progression was 52 months for the Mel200 group vs 25 months for the Mel140 group (HR=0.4 for Mel200, CI=0.2–0.72, P=0.004; Figure 4), further supporting the explanation that ability to provide effective subsequent treatments after progression may have played a role in the OS difference between the 2 groups.

EXPLORATORY ANALYSES:

We performed Kaplan-Meier survival analysis for PFS and OS, comparing melphalan dose using subgroups from each categorical covariate used in the multivariate and logistic regression analysis, to identify interactions between the melphalan dose and other features. None of these analyses identified significant associations. The PFS appeared to be independent of the melphalan dose, and OS was similarly better with Mel 200 in all studied subgroups except against those who received Mel140 due to renal failure (a much younger subgroup of 11 patients). These interaction analyses were also limited by the low number of patients in the Mel140 group with high-risk genetic lesions (n=4), and PR status at ASCT (n=5).

Interestingly, the use of lenalidomide maintenance was not found to improve PFS or OS in this cohort. HR for progression was 1.7 (P=0.018, 95% CI=1.1–2.6) among those receiving maintenance. However, maintenance therapy was administered to 42.9% of those without VGPR/CR and to 18% with VGPR/CR (P=0.002), suggesting a potential bias in patient selection for maintenance therapy, which may have also blunted the PFS differences between patients undergoing ASCT with different response states. We found no interaction between the melphalan dose and maintenance therapy on PFS or OS.

Discussion

This study provides the outcome data of 159 patients with MM receiving first, single ASCT over a 10-year period. Patients receiving Mel200 were shown to have better OS than those receiving Mel140, whereas both groups had similar PFS. Only 3 patients were excluded from evaluation due to missing data (1.6% of the cohort), most patients received induction therapies comparable to contemporary standards, and the 2 groups were homogenous in terms of demographic factors and management principles, including first and second LoT, and post-ASCT maintenance, highlighting the strength of this study, compared with other papers in the literature that will be discussed below. The non-randomized distribution based on frailty or renal function to melphalan dose, lack of data for formal frailty assessments, and the small sample size for the Mel140 group were the key limitations.

Early progression was closely associated with higher mortality across the cohort; however, despite the similar PFS, the OS superiority for Mel200 was maintained in all subgroup analyses, including those based on response states before ASCT. Non-relapse mortality was not different between the 2 groups. A significantly higher rate of patients in the Mel200 group had received more intensive subsequent therapies following progression after ASCT. These observations suggested that patients in the Mel140 group who were already found to be moderately frail at the time of ASCT were not considered for or could not tolerate intensive therapies when they progressed, possibly contributing to their shortened OS.

Mel200 and Mel140 were never directly compared in a prospective study. A randomized study comparing double transplants with Mel200 or Mel100 had demonstrated the superiority of Mel200 to the lower dose based on better efficacy and similar safety profile in a relatively fit and young cohort without renal failure [18]. Concerns of increased toxicity with Mel200 among elderly frail patients and those with renal failure led to the frequent use of Mel140 for this group of patients [8,9]. A pharmacokinetic study corelating the serum levels of melphalan demonstrated better disease control among those with higher serum melphalan concentrations during ASCT [19]. This suggested that some patients considered for Mel140 may be similarly benefiting with the lower dose as fit patients receiving Mel200, due to potentially slower drug elimination in the former group. The findings from this pharmacokinetic study may explain why patients in the Mel140 group who had renal failure had significantly better OS than the remaining of the Mel140 group and comparable OS to the entire Mel200 group. However, this interpretation was limited by the age difference between groups, as patients with kidney failure who received Mel140 were significantly younger than the remaining cohort.

The most comprehensive comparison of Mel140 to Mel200 was published in 2018 by Auner et al, using the multicenter EBMT registry data (CALM study) of 2253 patients to compare Mel140 with Mel200 [12]. Although the PFS was similar between the 2 groups, there was an unexpected interaction between the melphalan dose and response status concerning OS, significantly favoring Mel140 for patients with VGPR or CR at the time of ASCT. The authors attempted to explain this controversial finding with factors such as heterogenous local practices for selecting induction treatment, melphalan dosing, and post-transplant management, along with possible delayed physical recovery interfering with post-transplant treatments and late organ-specific toxicities, such as arrythmias with Mel200, and argued that a randomized trial was needed. There were also 289 (13%) excluded patients due to inadequate data in this study, which could have contributed to a form of unintentional bias.

The results from other studies addressing the same subject were not as unsettling [13–17]. Over 2000 elderly patients who received ASCT between 2013 and 2017 were compared for outcomes based on melphalan dose as a subset analysis [13]. This study demonstrated a small (4%) but statistically significant advantage in PFS and OS at 24 months from ASCT for the Mel200 group, whereas the relapse rates were found to be similar, and non-relapse mortality was higher for the Mel140 group, with an HR of 2.2. The comparison of patients in the Mel140 and Mel200 groups were not specified in terms of demographic data, genetic anomalies, stage, response before ASCT, induction and post-transplant treatments in this study, limiting the interpretation. Single-center studies with more homogenous data reported similar PFS and OS between Mel140 (or dose <150 mg/m2) vs Mel200 [14,15]. Particularly interesting data of 83 patients who had achieved CR at the time of ASCT demonstrated similar PFS and OS between Mel140 and Mel200 after a relatively short period of follow-up [16]. The relatively recent propensity score-matched comparison of Mel140 and Mel200 from the MD Anderson Cancer Center highlighted the similar OS between the 3 melphalan doses among those with VGPR or better response state at the time of ASCT, and a trend for better survival with Mel200 among those with PR or worse, similar to the EBMT registry [17].

The present study in the context of the other major studies mentioned above is relevant mainly from 2 perspectives. First, it is a counter-argument to the proposed association of increased risk of mortality with Mel200 among patients with VGPR or better response states at the time of ASCT by Auner et al [12]. Although it did not end up causing a significant shift in practice, the data from this EBMT registry is unsettling and needs to be taken seriously, since it is the study with the largest sample size and is credible by EBMT, a major authority on transplantation. By addressing some of the key limitations of the EBMT registry, including a high rate of missing data and heterogenous local practices, this study provides a different perspective in addressing the same question and has demonstrated reassuring survival outcomes favoring the standard practice as its primary endpoint. The ESMO guideline has maintained its recommendation that Mel200 is the standard of care [2] despite the EBMT data from the CALM study, which were published in 2018 [12]. The findings from this study are in favor of the standard of care, alleviating some of the concerns raised by the EBMT registry in terms of using Mel200.

The second important perspective is the similar PFS and non-relapse mortality rates between the 2 groups. Progressed patients in the MEL200 group were more likely to receive daratumumab and carfilzomib at recommended intensities, as well as second transplants, than those in the MEL140 group, which translated into improved survival. These findings are in line with previous literature and suggest that the remarkable and continuous improvement in the life expectancy of patients with MM over the last 2 decades [20,21] is related more to the increasing availability and efficacy of modern treatments including proteosome inhibitors, immunomodulators, monoclonal antibodies, bispecific antibodies, and chimeric antigen receptor T cells, than to the optimization in conditioning regimens.

This study had significant limitations, as mentioned throughout the text. The 2 key limitations were the small sample size, especially for the Mel140 group, and the non-randomized and retrospective patient allocation to each dose group. Nevertheless, the study was designed in awareness of these limitations, and the primary objective was purposefully restricted to assess the association of Mel200 with increased risk of death as proposed by the EBMT data. Consecutive patient inclusion, homogenous management principles, availability of data for subsequent therapies, and a very low rate of missing data were strengths of the study, whereas the lack of data for formal geriatric frailty assessments before ASCT was another key limitation, necessitating caution in interpreting results.

Conclusions

We found no evidence to suggest that Mel200 had an adverse impact on the survival of patients having achieved VGPR or CR before ASCT, unlike the data recovered from the EBMT registry. Superior OS with Mel200 as the primary endpoint was documented, supporting the standard of care [1,2]. The differences in findings between studies reflect the impact of local practices in comparing the outcomes of non-randomly selected patient populations and highlight the need for randomized comparison of Mel140 and Mel200, as suggested by the CALM study.