Early Relapse Risk in Patients with Newly Diagnosed Multiple Myeloma Characterized by Next-generation Sequencing

The expected survival of patients with newly diagnosed multiple myeloma (NDMM) is currently improving and approaching 8 years, thanks to the use of novel agents and better supportive care (1). Nevertheless, multiple myeloma still remains largely incurable and about 12,000 patients with multiple myeloma in the United States and 30,000 patients with multiple myeloma in Europe die each year, with the main cause of death being the development of refractory disease to the currently available drugs (2–4).

Relapse is caused by multiple myeloma cell clones with an increasing degree of drug refractoriness and genetic complexity eventually leading to shorter remissions (5). Because the longest remission period is usually induced by upfront treatment, the duration of first remission is one of the most important factors impacting patient prognosis (6).

This can become particularly important as a dynamic prognostic marker, if we consider the complexity associated with the evaluation of baseline prognostic features. The most widely used staging system is the Revised International Staging System (R-ISS), which is based on clinical and biological standard features [ISS, chromosomal abnormalities and lactate dehydrogenase (LDH) levels; ref. 7]. Many efforts aimed at improving the baseline stratification, including the use of gene expression profiles (GEP) and next-generation sequencing (NGS; refs. 8–10). Of note, according to R-ISS, only 10% of patients are at high risk of progression and/or death and, according to the NGS-based “double-hit” classification, only 6.1% of patients are at high risk of progression and/or death, but the overall rate of patients who relapse or die within 2 years from diagnosis is about 20% (11, 12). This highlights the importance of dynamic prognostic evaluation and the need for an improved baseline risk stratification. The identification and treatment of patients with high-risk multiple myeloma currently represent unmet medical needs. Our aims were (1) to characterize patients with early progressive disease [early PD; time-to-progression (TTP) ≤18 months] after first-line treatment including immunomodulatory drugs (IMiD) and/or first–second generation proteasome inhibitors (PI) incorporating baseline clinical and NGS molecular features; (2) to address the role of different upfront therapies in reducing the risk of early PD.

Data from patients enrolled in the prospective observational Multiple Myeloma Research Foundation (MMRF) CoMMpass study (NCT01454297) were included in this analysis. Ethics committees or institutional review boards at the study sites approved the study, which was conducted in accordance with the Declaration of Helsinki. All patients provided written informed consent.

Main inclusion criteria were: symptomatic NDMM, measurable disease, and upfront systemic therapy including an IMiD and/or a PI. CoMMpass data were generated as part of the MMRF Personalized Medicine Initiatives (https://research.themmrf.org and www.themmrf.org).

Data from patients receiving treatment in the context of clinical trials as well as with real word regimens were included. Therapy (source file “mmrf_commpass_IA14_stand_alone_treatment_regimen” available upon request on https://research.themmrf.org) was reviewed and classified according to: type of induction treatment (bortezomib-dexamethasone/bortezomib + chemotherapy triplets/lenalidomide-dexamethasone/bortezomib-lenalidomide-dexamethasone/carfilzomib-based/other), autologous stem-cell transplantation (ASCT; yes/no), and type of continuous therapy [(CT); IMiDs CT/PIs CT/IMiDs + PIs CT/fixed-duration therapy (FDT)]. FDT was defined as ≤1 year of upfront treatment (13). The definition of variables is detailed in Supplementary Tables S1 and S2. Patients were considered evaluable for the ASCT versus no ASCT analysis if they were alive and relapse-free after induction treatment and if the date of ASCT was available. Patients receiving ASCT before PD but after 18 months from the start of treatment (cut-off for the early relapse evaluation) were considered nonevaluable. Patients were considered evaluable for the CT analysis if they were alive and relapse-free after 1 year from the start of treatment, if the follow-up was >1 year, and if details of treatment administered after the 1-year time point were available.

The Interim Analysis (IA)14 release of CoMMpass was analyzed. Updated time-to-event endpoints for CoMMpass patients coenrolled in the NCT02203643 trial were used (data cut-off date: May 30, 2018; the treatment schedule is reported in the Supplementary Appendix).

Baseline bone marrow CD138⁺ cells were obtained before the initiation of systemic therapy (within 30 days before first-line treatment). Available data on samples at relapse, a preplanned objective within the CoMMpass study, were also evaluated. Long-insert whole-genome sequencing (WGS) and whole-exome sequencing (WES) were performed by the Translational Genomics Institute (TGen, Phoenix, AZ). Somatic tumor alterations were defined comparing tumor cells with patient-specific paired normal cells. Details on the definition of the risk factors explored in this work are provided in previous CoMMpass publications (14–16). Cytogenetic data reported by single study centers were heterogeneous in terms of fluorescence in situ hybridization (FISH) probes utilized, number of cells counted and cell sorting techniques. To uniformly define cytogenetic abnormalities in all patients, copy-number abnormalities (CNA), immunoglobulin heavy chain (IgH) translocations and Ig lambda (IgL) translocations were defined using molecular data (Seq-FISH; refs. 17–19). The concordance of Seq-FISH and conventional FISH in a subgroup of patients evaluated in the context of a clinical trial by a centralized laboratory showed a high degree of concordance. The presence or absence of recurrent CNAs [hyperdiploidy, deletion 13q, deletion 17p, gain 1q (three CSK1B copies), and amplification 1q (>3 CSK1B copies)], IgH translocations [t(11;14), t(4;14), t(14;16), t(14;20)] and IgL translocations were evaluated using calls on WGS long-insert data (19). The threshold for a positive detection of a CNA by Seq-FISH was 20%. Nonsynonymous alterations with an allele ratio of at least 5% in the tumor sample and less than 2% in the constitutional sample occurring in a customized panel of 21 genes known to be significantly mutated in multiple myeloma were also analyzed (Supplementary Table S1; refs. 20, 21). The cancer cell fraction of mutations of interest corrected by tumor purity and multiple myeloma cell ploidy was estimated using the ABSOLUTE algorithm (22). Moreover, we evaluated the aberrant activity of APOBEC cytidine deaminases (known to be associated with high mutational burden and poor prognosis in multiple myeloma; ref. 23), using the recently developed fitting algorithm mmsig (Supplementary Table S1; https://github.com/evenrus/mmsig; ref. 24). APOBEC activity was defined as high or low based on its quartile distribution (4th quartile vs. others; ref. 23).

Early PD was defined as occurring in the first 18 months from the start of treatment. Patients not experiencing PD within 18 months from the start of treatment were included in the reference population. The reference population was further classified in late PD (occurring after the first 18 months from the start of treatment) and no PD at the last follow-up. TTP was defined as the duration from start of treatment to PD; deaths from causes other than progression were censored (25).

Epanechnikov kernel-smoothed estimated hazard rates were used to study the risk of PD over time.

Best response to first-line treatment and drug refractoriness after first-line treatment were evaluated according to the International Myeloma Working Group guidelines (25, 26). The comparison of best response and drug refractoriness in the early versus late PD groups was performed according to two-sided Fisher exact test.

Overall survival (OS) was analyzed as time-to-event data using the Kaplan–Meier method. The Cox proportional hazards model was used to estimate the hazard ratio (HR) values and the 95% confidence intervals (CI). To account for potential confounders, the comparison of early PD versus reference population was adjusted for age, ISS, high-risk cytogenetics (27), induction treatment, ASCT, CT, and clinical trial enrollment. ASCT and CT were considered as time-dependent variables.

An 18-month landmark analysis for OS was also performed, comparing OS in the early PD versus late PD versus no PD groups.

To identify risk factors associated with early relapse, patients that were not at risk for progression for the entire 18-month period after the start of treatment were excluded from the reference population (n = 101; Fig. 1).

Univariate analysis of factors associated with early PD versus late/no PD was performed using Fisher exact test, Kruskal–Wallis test or χ² test as appropriate. Starting from the variables with a P < 0.15 in univariate analysis, the final logistic model was identified through a backward selection based on the minimization of the Akaike Information Criterion, keeping in the model the therapy-related variables. The final logistic regression model was used to estimate OR for early relapse risk, 95% CIs and P-values. A confirmatory analysis on the same patient population using death within 24 months as an endpoint was conducted (11).

All the analyses were conducted using R version 3.5.1 and bespoke code, which is available upon request.

The access to the IA 14 release of CoMMpass was approved by the Data Access Use Committee and downloaded from https://research.themmrf.org/rp/download. CoMMpass data are deposited in the database of Genotypes and Phenotypes (dbGaP; Study Accession phs000748.v7.p4—https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000748.v7.p4).

Data from 1,151 patients were available in the CoMMpass IA14. Patients without WES data (n = 213) and PD information (n = 12) were excluded from the analysis. The remaining 926 patients represented the population analyzed in the current work. Patient characteristics are shown in Table 1.

Median age was 63 years and most of the patients had an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1 (39% and 44%, respectively). Baseline prognostic factors were typical of a NDMM population. A total of 27% of patients presented with ISS stage III and 8% with high LDH levels; 13% of patients presented with del(17p), 14% with t(4;14), 5% with t(14;16), 1% with t(14;20), 27% with gain(1q), and 7% with amp(1q), while IgL translocations, a recently described marker of high-risk multiple myeloma (19), were present in 10% of evaluable patients.

Genes affected by somatic nonsynonymous alterations in at least 25 (3%) patients were analyzed (Supplementary Table S3). Mutational frequency was dominated by alterations in KRAS (25%), NRAS (21.5%), and IGLL5 (16%) gene.

The most frequent induction regimen administered was bortezomib-lenalidomide-dexamethasone (VRd; 34%), followed by bortezomib+chemotherapy triplets (23%) and carfilzomib-based treatment (23%).

Patients evaluable for the ASCT versus no ASCT comparison were 833. Nonevaluable patients experienced PD during induction (n = 40), died for reasons other than PD (n = 18), were lost to follow-up (n = 14), withdrew consent (n = 5), or discontinued the study for other reasons (n = 6). Ten patients received ASCT after the 18-month endpoint and were considered nonevaluable as well. High-dose chemotherapy followed by ASCT was received by 53% of the evaluable patients; the median time to ASCT was 169 days (range 78–508).

Patients evaluable for CT versus FDT comparison were 609. During the first year of treatment, nonevaluable patients had PD (n = 112), died for reasons other than PD (n = 32), were lost to follow-up (n = 21), withdrew consent (n = 16) or discontinued the study for other reasons (n = 15). In 121 patients, information of drugs used during CT was lacking at the current follow-up. A total of 74% of evaluable patients received CT (IMiDs 42%, PIs 14%, and IMiDs + PIs 18%); 26% of patients received FDT. The distributions of induction treatment and ASCT in each CT subgroup are shown in Supplementary Table S4.

The median follow-up of the entire population was 39 months. A total of 191/926 (20.6%) patients experienced early PD, while the remaining 735/926 (79.4%) patients were included in the reference population (Fig. 1).

In the early PD group, 126/191 (66%) patients discontinued the study at the last follow-up: 75 (39%) for death due to PD, 26 (14%) for death due to other reasons, four (2%) due to withdrawal of consent, three (2%) for being lost to follow-up, and 18 (9%) for other reasons.

In the reference population, 229/735 (31%) patients discontinued the study: 39 (5%) for death due to PD, 66 (9%) for death due to other reasons, 31 (4%) due to withdrawal of consent, 39 (5%) for being lost to follow-up, and 54 (7%) for other reasons. In the same reference population, 228/926 (24.6%) patients experienced a late PD (TTP >18 months), while 507/926 (54.8%) did not experience PD at the last follow-up.

Overall response rate (ORR) was significantly lower in patients with early PD compared with patients with late PD (80% vs. 96%, respectively, P < 0.001). Deep responses were also different, with very good partial response (VGPR) rates of 40% versus 57%, complete remission (CR) rates of 2% versus 18%, and stringent CR rates of 5% versus 8% in early versus late PD groups, respectively. This translated into a significantly different rate of ≥VGPR in the two groups (47% vs. 82%, P < 0.001; Table 2).

A significantly higher proportion of patients in the early versus the late PD group developed a refractoriness to PIs (50% vs. 18%, P < 0.001) and IMiDs + PIs (21% vs. 8%, P < 0.001), while no differences were found in terms of IMiD refractoriness (42% vs. 38%, P = 0.541; Table 2).

OS of patients with early PD versus the reference population is shown in Fig. 2.

Patients with early PD had a significantly higher risk of death compared with the reference population (HR, 4.89; 95% CI, 3.72–6.43; P < 0.001), with 53% of patient deaths at 3 years in the early PD cohort compared with only 12% in the reference cohort. This effect was maintained after adjusting the analysis for age, baseline prognostic factors (ISS, high-risk cytogenetics; ref. 27), treatment, and clinical trial enrollment (HR, 3.65; 95% CI, 2.70–4.93; P < 0.001). Of note, 61% of early relapsing patients presented with ISS stage I or II and 74% had conventionally defined standard-risk cytogenetics (27). The median OS of the early relapsing patients was 32.8 months, lower than that of the high-risk population defined using baseline ISS III (median OS, 54 months) or baseline high-risk cytogenetics (median OS, 65 months; ref. 27).

Patients with early PD were defined using a time-dependent endpoint (18 months); consequently, a landmark analysis of OS with a landmark point at 18 months was performed to validate our findings (Fig. 3). At the landmark time point, 121 patients with early PD and 640 patients in the reference population were evaluable. The main reasons for not being evaluable were death due to PD during the first 18 months in the early PD population (58/191, 30%) and death due to reasons other than PD during the first 18 months in the reference population (42/735, 6%). The difference in early death rates between the two groups led to a possible underestimation of OS differences after the landmark time point. Moreover, in this OS comparison, we split the reference population in patients with late PD and no PD. The 18-month landmark analysis showed a significantly worse OS in patients with early PD compared with both patients with late PD (HR, 2.05; 95% CI, 1.25–3.35; P = 0.004) and patients with no PD (HR, 8.05; 95% CI, 4.11–15.74; P < 0.001).

We investigated the clinical and prognostic variables impacting the risk of early relapse. In this analysis, we excluded from the reference population the patients who were not at risk for the entire 18-month period (101/926, 11%). Excluded patients were those that in the first 18 months died without a PD (n = 42), withdrew the consent (n = 14), were lost to follow-up (n = 25) or interrupted the protocol for other reasons (n = 20).

A significantly higher proportion of patients in the early PD group versus the reference population presented with ISS stage III (39% vs. 20%), gain(1q) (26% vs. 20%), IgL translocations (14% vs. 6%), high APOBEC signature (30% vs. 24%), high LDH (9% vs. 5%), ECOG ≥2 (23% vs. 11%), KRAS mutation (31% vs. 24%), IGLL5 mutation (20% vs. 14%), and TP53 mutation (9% vs. 3%; Supplementary Table S5). These variables were therefore included in multivariate analysis, together with age and treatment administered.

In multivariate analysis (Fig. 4), TP53 mutation [odds ratio (OR), 3.78, P < 0.01], high LDH levels (OR, 3.15, P < 0.01), IgL translocation (OR, 2.25, P = 0.03), and IGLL5 mutation (OR, 2.15, P < 0.01) were significantly correlated with a higher risk of early PD. Only a trend was found for gain(1q) and amp(1q) (Fig. 4).

Receiving ASCT (OR, 0.27, P < 0.01) and CT with IMiDs+PIs (OR, 0.34, P = 0.02) were significantly correlated with a lower risk of early PD. The effect of ASCT was confirmed in age-specific patient subgroups, showing similar ORs in patients aged ≤65 years (n = 531; OR, 0.27; 95% CI, 0.13–0.54) and aged 66–75 years (n = 222; OR, 0.30; 95% CI, 0.11–0.74).

A protective effect of carfilzomib-based induction was also observed (OR, 0.15, P = 0.01). Nevertheless, most of carfilzomib-treated patients were enrolled in a clinical trial and the enrollment effect itself was a protective factor as well (OR, 0.09, P < 0.01).

To confirm our results, we performed an additional analysis using death within 24 months as an endpoint (ref. 11; Supplementary Fig. S1). The adverse effects of TP53 mutation (OR, 3.35, P = 0.02) and IgL translocation (OR, 2.34, P = 0.046) were confirmed. Moreover, also 1q abnormalities were significantly correlated with an increased risk of death within 24 months. ASCT (OR, 0.44, P = 0.02) retained its protective effect.

In our analysis, TP53 mutation was the factor with the greatest effect size for early PD. Its association with patients with multiple myeloma carrying concurrent del(17p) is well known. In this cohort, 865 patients were evaluable for TP53 mutation and del(17p) (Supplementary Fig. S2A). 121 of 865 patients had del(17p) or TP53 mutation. Among them, 82/121 (68%) had del(17p) only, 10/121 (8%) had TP53 mutation only, and 29/121 (24%) had del(17p) and TP53 mutation. Rates of early PD in each patient subgroup are shown in Supplementary Fig. S2B. Patients with del(17p) but not TP53 mutation had an early PD rate of 17.1% (comparable with the general population), while the bi-allelic group [del(17p) + TP53 mutation] and the TP53-mutation-only group showed high early PD rates (41.4% and 50%, respectively). Of note, the TP53-mutation-only group was composed by only 10 patients and the majority of TP53-mutated patients experiencing early relapse were in the del(17p) + TP53 mutation group.

The use of a higher cut-off level to define del(17p) positivity (50% instead of 20%; Supplementary Fig. S2C and S2D) led to a slightly higher early PD rate in del(17p)-only patients (25%). However, the biallelic [del(17p) + TP53 mutation] and the TP53-mutation-only groups still showed the highest rates of early PD (40.7% and 50%, respectively).

Considering that TP53 mutation is important to confer early relapse risk, we hypothesized that TP53-mutated clones needed to be conserved at relapse. Only 6 patients with TP53 mutation at diagnosis had available molecular data at relapse, although in 6/6 cases TP53 mutation was conserved in relapse samples (Supplementary Fig. S3A). Moreover, despite the small numbers, if TP53 mutation was subclonal at diagnosis, a higher cancer cell fraction was found in paired samples at relapse. This effect was different from the IGLL5 mutations, in which subclonal cases tended to disappear at relapse (Supplementary Fig. S3B).

Multiple myeloma prognosis is improving and early relapse after upfront treatment is beginning to be recognized as a high-risk feature (28). The same observation had been done for other hematologic malignancies with an expected indolent course, such as follicular lymphoma and chronic lymphocytic leukemia (29, 30).

Here we proposed progression ≤18 months after the start of first-line treatment as a marker of high risk and demonstrated its detrimental effect on the OS of patients with NDMM.

The 18-month cut-off was chosen because our time to ASCT was approximately 6 months and the majority of published studies on patients with multiple myeloma with early PD defined early PD as a relapse within 12 months from ASCT. Indeed, the hazards of progression in our patient population increased over time with no identified peak of risk (Supplementary Fig. S4).

We incorporated in our analysis baseline clinical and biological features to identify risk factors of early PD. The characterization by NGS of this patient cohort allowed us to simultaneously study CNAs, translocations and mutations in genes of interest by using the same platform. This is an advantage of NGS versus conventional FISH, which cannot detect mutations and needs specific probes to detect prespecified translocations and CNAs. Moreover, NGS and conventional FISH showed high concordance in detecting the same CNAs and translocations, as shown in Supplementary Fig. S5 and by others (17, 18).

TP53 mutation, which is currently not included in the standard baseline evaluation of patients with multiple myeloma, was the most important factor increasing the risk of early PD emerging from our analysis. Its adverse effect was confirmed in the risk of death within 24 months from diagnosis. TP53 mutation is rare in patients at diagnosis (3.5%), but about 25% of patients with del(17p) has also TP53 mutation. As similarly observed by other groups (9), our data further supported the routine testing of TP53 mutation at least in del(17p)-positive patients. Indeed, the presence of del(17p) without TP53 mutation conferred an early PD risk that was similar to that of the overall population.

In our analysis, IgL translocation and IGLL5 mutation also emerged as risk factors of early PD. Both of them have already been associated with poor prognosis (19, 31). White and colleagues showed that mutations in IGLL5 can be associated with translocations juxtaposing IGLL5 (31). In our analysis, IGLL5 mutations and IgL translocations showed a trend toward cooccurrence, though not statistically significant (P = 0.06). The higher risk of early relapse observed in IgL-translocated patients, the loss of subclonal IGLL5 mutations at first relapse and the significant effect of IgL translocations but not of IGLL5 mutations in the risk of death within 24 months could suggest that IgL translocations impacted patients' prognosis more than IGLL5 mutations.

Only a trend toward a higher risk of early PD was found for gain(1q) and amp(1q). However, using death within 24 months as an endpoint, the effect of 1q abnormalities was more evident. This was possibly due to the use of a later time point allowing more patients to experience an event and to a possible more specific effect of 1q abnormalities on the risk of death.

In our analysis, the only clinical factor that increased the risk of early PD in multivariate analysis was baseline LDH, a well-known marker of disease aggressiveness in several hematologic diseases.

Other factors not included in the current analysis—such as circulating plasma cells (32), high-risk GEP (8, 33), and multiple myeloma cell-extrinsic factors (34)—could also play a role in determining the risk of early PD and should be investigated in future works. Moreover, our analysis focused on multiple myeloma cells derived from a random bone marrow aspirate, and spatial heterogeneity of high-risk features could also explain some of the early PD cases (35).

ASCT and CT with IMiDs + PIs showed a protective effect against early PD in this patient population. However, the majority of patients in the analyzed cohort were real-world patients and the analysis was consequently performed as per protocol, thus leading to a risk of overestimation of effects of ASCT and CT. With these limitations, our data support the intensification of therapy in patients at risk of early relapse and underline the importance of continuous treatment with combination regimens to optimize long-term disease control (36).

Carfilzomib-based induction also showed to reduce the risk of early relapse, although it is difficult to distinguish between treatment and trial effects because the majority of carfilzomib-treated patients were included in a clinical trial, whereas this was not the case for other induction regimens.

Besides clinical trial enrollment, this patient population was heterogeneously treated and our findings on early PD risk need to be confirmed in homogeneously treated patients. For instance, among the CT subgroups, heterogeneous upfront treatments before CT were received (Supplementary Table S4). Nevertheless, the multivariate analysis on the risk of early PD was adjusted for induction treatment, ASCT, CT, and trial enrollment effect, taking into account these differences.

The median age of the analyzed cohort was 63 years, younger than the usual median age of unselected patients with multiple myeloma. Elderly patients were underrepresented and the confirmation of our results in this patient population is warranted. However, other variables that are patient-related but not disease-related (e.g., frailty status) may have a major prognostic role in elderly patients (37).

Patients with early PD showed suboptimal responses and, at relapse, were more frequently refractory to PIs and double refractory to IMiDs + PIs, as compared with patients with late PD. IMiD refractoriness was not different between early PD and late PD groups. This was mainly due to the widespread use of PI-containing regimens during the first 18 months of therapy. On the other hand, after the 18-month time point, treatment with an IMiD as single agent was widely used in our patient population. Therefore, a high percentage of PI-refractory and IMiD + PI-refractory cases were observed in the early PD group, while IMiD-refractory cases were well represented in both the early PD and late PD groups.

In conclusion, early PD identifies a high-risk multiple myeloma population that still represents an unmet clinical need. As compared with FISH, extended genotyping through the routine use of NGS at diagnosis is feasible and may improve the patient stratification and identify patients at risk of early PD (38). Further research is needed to better identify baseline features predicting early relapse and the optimal treatment approach. Recently, clinical trials on patients experiencing PD within 18 months from the start of treatment are beginning to emerge (e.g., NCT03601078, cohorts 2a and 2b), thus suggesting that risk-adapted treatment in this patient population could soon become a feature of multiple myeloma clinical management.

M. D'Agostino reports personal fees from GSK (advisory board) outside the submitted work. S. Oliva reports personal fees from Amgen (honoraria; advisory board), Celgene (honoraria), Janssen (honoraria; advisory board), Adaptive Biotechnologies (advisory board), and Takeda (advisory board) outside the submitted work. S. Bringhen reports personal fees from Celgene (honoraria; advisory board), Amgen (honoraria; advisory board), Janssen (honoraria; advisory board; consultancy), Bristol-Myers Squibb (honoraria), Karyopharm (advisory board), and Takeda (consultancy) outside the submitted work. A. Larocca reports personal fees from Amgen (honoraria), Bristol-Myers Squibb (honoraria; advisory board), Celgene (honoraria; advisory board), Janssen (honoraria; advisory board), GSK (honoraria), and Takeda (advisory board) outside the submitted work. M. Boccadoro reports personal fees and other from Sanofi (honoraria; research funding), Celgene (honoraria; research funding), Amgen (honoraria; research funding), Janssen (honoraria; research funding), Novartis (honoraria; research funding), Bristol-Myers Squibb (honoraria; research funding), and AbbVie (honoraria); and other from Mundipharma (research funding) outside the submitted work. N. Bolli reports grants from European Research Council (grant agreement No. 817997) and personal fees from Celgene (honoraria), Janssen (advisory board), and Amgen (honoraria) outside the submitted work. F. Gay reports personal fees from Amgen (honoraria; advisory board), Celgene (honoraria; advisory board), Janssen (honoraria; advisory board), Takeda (honoraria; advisory board), Bristol-Myers Squibb (honoraria; advisory board), Roche (advisory board), AbbVie (advisory board), Adaptive (advisory board), and Seattle Genetics (advisory board) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

M. D'Agostino: Conceptualization, resources, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. G.M. Zaccaria: Conceptualization, resources, data curation, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. B. Ziccheddu: Resources, data curation, validation, investigation, visualization, writing-original draft, writing-review and editing. E.H. Rustad: Resources, data curation, validation, investigation, writing-review and editing. E. Genuardi: Resources, data curation, validation, investigation, writing-review and editing. A. Capra: Resources, data curation, software, formal analysis, validation, investigation, visualization, writing-original draft, writing-review and editing. S. Oliva: Conceptualization, resources, data curation, validation, investigation, methodology, writing-review and editing. D. Auclair: Conceptualization, resources, data curation, validation, investigation, methodology, writing-review and editing. J. Yesil: Conceptualization, resources, data curation, validation, investigation, methodology, writing-review and editing. P. Colucci: Resources, data curation, validation, investigation, writing-review and editing. J.J. Keats: Conceptualization, resources, data curation, validation, investigation, methodology, writing-review and editing. M. Gambella: Conceptualization, resources, data curation, validation, investigation, methodology, writing-review and editing. S. Bringhen: Resources, data curation, validation, investigation, writing-review and editing. A. Larocca: Resources, data curation, validation, investigation, writing-review and editing. M. Boccadoro: Conceptualization, resources, data curation, supervision, validation, investigation, methodology, writing-review and editing. N. Bolli: Conceptualization, resources, data curation, supervision, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. F. Maura: Conceptualization, resources, data curation, supervision, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. F. Gay: Conceptualization, resources, data curation, supervision, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing.

F. Maura is supported by the Memorial Sloan Kettering Cancer Center NCI Core Grant (P30 CA 008748).

The CoMMpass study is sponsored by the Multiple Myeloma Research Foundation (MMRF), which had no role in the data interpretation, writing of the report, or publication of this contribution. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit this manuscript for publication, together with the other authors.

We thank all the patients who participated in the study, the nurses Luisella D'Ambrosio and Tiziana De Lazzer, the data managers Debora Caldarazzo and Elena Tigano, and Ugo Panzani from the Torino site.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.