Toward personalized treatment in multiple myeloma based on molecular characteristics

Myeloma develops as the result of an evolutionary process during which a normal plasma cell moves through the premalignant state monoclonal gammopathy of uncertain significance to smoldering myeloma and myeloma that requires treatment.¹  Advances in therapy over the last 2 decades have improved patient outcomes, while the use of new technology has increased our understanding of the molecular drivers that underlie disease initiation and progression. Due to underlying molecular variation, the clinical disease course is very heterogeneous.²  While some patients experience long remission periods or functional cures, others relapse early or are refractory to therapy. In order to continue to improve outcomes, information regarding the molecular abnormalities driving these differences in outcomes needs to be incorporated into clinical care. These features may relate to messenger RNA (mRNA), DNA, or protein changes, but the aim is to identify aberrations that help inform the diagnosis, outcome, or treatment relevant to a specific patient or subgroup of patients. Such molecular features or biomarkers are defined by the NIH Biomarkers Definitions Working Group as “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention.”³ 

The purpose of using biomarker-driven, personalized, treatment approaches is to maximize benefit and reduce toxicity. In order to achieve this goal, the biomarker must be measurable in a robust and reproducible manner. Advances in technology have helped the identification and validation of myeloma biomarkers relevant to treatment, such as those that can predict outcome for patients based on differences in survival (prognostic biomarkers) or target treatment to subsets of patients based on specific molecular pathology (predictive biomarkers). Some biomarkers can clearly be both prognostic and predictive, and approaches to target these are likely to have the greatest impact on outcomes.

In this review, we describe the current use of prognostic and predictive biomarkers in myeloma and speculate on advances that may enable further improvement in patient outcomes by employing these biomarkers to define personalized treatment strategies.

The technologies enabling molecular profile analysis have evolved significantly over the last few decades, contributing to an increased understanding of myeloma pathogenesis (Figure 1). Initial studies were performed using G-banding cytogenetics that identified translocations involving the immunoglobulin heavy chain (IgH) gene locus and hyperdiploidy as initiating events.^4,5  Translocations including t(4;14), t(6;14), t(11;14), t(14;16), and t(14;20) place oncogenes, MMSET/FGFR3, CCND3, CCND1, MAF, and MAFB, respectively, under the control of the IgH gene enhancer.^6,7  The downstream effect of upregulation of these genes converges on the increased expression of cyclin D protein family members, ultimately driving G1/S checkpoint dysregulation.⁸  Hyperdiploidy, characterized by trisomies of odd-numbered chromosomes, also affects this checkpoint, although the mechanism of its acquisition and downstream effect is less well understood. Subsequent studies have shown that secondary acquired lesions compound the cell cycle dysregulation, driving further proliferation and disease progression.⁹ 

The use of fluorescence in situ hybridization (FISH), gene expression profiling (GEP), and single-nucleotide polymorphism (SNP) array technologies has expanded our knowledge of myeloma biology and enabled its further classification into subgroups.^8,10,11  Cases cluster mainly based on the underlying structural genetic event (translocations and hyperdiploidy), with 2 classification systems surviving the test of time: the TC (translocation/cyclin D) classification⁸  and University of Arkansas for Medical Sciences¹⁰  subgroups.

In more recent years, the introduction of next-generation sequencing (NGS) technologies has allowed the identification of single-nucleotide variants as well as larger structural changes, including translocations and copy-number abnormalities, more quickly and cheaply.^12-15  Dozens of myeloma driver genes have been identified. The most common occur in the RAS and NF-κB families,^12,14,16  with many mutations associated with the primary myeloma molecular subgroups suggesting the underlying background initiating event drives the acquisition of subsequent molecular abberations.¹³  A number of these new technologies are now Clinical Laboratory Improvement Amendments certified and available for diagnostic use, while additional techniques such as DNA methylation analysis remain confined to research laboratories.

Prognostic biomarkers are used to identify the likelihood of disease relapse and/or predict overall survival. Classically, they predict outcome irrespective of what therapy is given and enable more personalized outcome advice in the context of current treatment regimens. Several lesions have been identified as carrying an adverse outcome in myeloma. Some of these are clonal initiating lesions such as t(4;14), t(14;16), and t(14;20), while others are structural chromosomal changes or mutation events that tend to happen later in the evolutionary process.

t(4;14) results in the histone methyltransferase (MMSET) and tyrosine kinase (FGFR3) genes being placed downstream of IgH gene enhancers.¹⁷  The spiked expression of MMSET is likely responsible for the adverse outcome, as the outcome is equally as poor in the subset of patients with concomitant loss of FGFR3 expression.^17,18  MMSET results in epigenetic reprogramming, leading to a cascade of downstream effects, including altered adhesion, enhanced growth, and increased cell survival.¹⁹  This reprogramming also leads to genetic instability, including gain of chromosome 1q (gain(1q)), deletion of the short arm of chromosome 12 (del(12p)), del(13q), del(22q), and BIRC2/3 homozygous deletion, that may contribute to mediate the adverse outcomes.²⁰  In comparison with other risk groups, t(4;14) is particularly heterogeneous in terms of outcome, potentially influenced by these additional lesions and/or co-occurrence of del(17p).²¹ 

Upregulation of MAF and MAFB is seen in t(14;16) and t(14;20), respectively, and is associated with similar genetic instability including the lesions gain(1q) and del(17p). These subgroups are also associated with a mutational signature (a characteristic combination of mutation types) associated with the activity of the mRNA editing enzyme APOBEC and have an increased number of mutations.¹⁶  MAF and MAFb protein have been demonstrated to mediate resistance to proteasome inhibitors,^22,23  perhaps contributing to this subgroup’s adverse outcome seen in most,^2,24,25  though not all,²⁶  studies.

Del(17p) has been associated with adverse outcome, which is thought to be related to the loss of expression of the tumor suppressor gene TP53.²⁷  While occasional studies suggest deletions in <20% of cells detected by FISH may have some clinical impact,²⁸  most studies use a cutoff of >20% to demonstrate a significant effect.^29,30  As the clone size increases, the effect on outcome becomes more marked, with some studies suggesting that clonal deletion in at least 60% of cells is required.³¹  More recent data suggest that biallelic disruption by biallelic deletion, deletion in one allele and TP53 mutation in the other allele, or biallelic mutation is what mediates the adverse outcome.^29,32 

Gain(1q) has been associated with an adverse outcome, although given the large number of genes situated on this chromosome, it is less clear which gene is responsible. Implicated genes at the most commonly gained locus (1q21) include BCL9, MCL1, CKS1B, and ANP32E.^33-36  This locus is thought to be susceptible to gain due to instability of the pericentromeric chromatin.³⁷  Other genes may also be important (eg, CD45 at 1q32) when larger regions of the chromosome are gained.³⁸  There is an important distinction between gain, defined as one additional copy, and amplification, defined as >1 additional copy of 1q (amp[1q]), with amplified cases appearing to be associated with a more adverse outcome.^15,20,32 

Del(1p) (incidence 10% at diagnosis) frequently co-occurs with gain(1q) and has been shown to be associated with an adverse outcome in patients undergoing autologous stem cell transplant. This effect is potentially mediated by loss of CDKN2C and FAF1 at 1p32 and/or FAM46C at 1p12 and/or RPL5 and EVI5 at 1p22.^24,39,40 

Myc aberrations (incidence 15% to 20% at diagnosis) are common and may be mediated by secondary translocations to the MYC locus at 8q22 or copy-number change and are associated with adverse outcomes.^41,42 

t(11;14) and hyperdiploidy are usually considered standard risk. Some studies suggest that individual trisomies may be able to overcome some of the adverse impact of other lesions such as t(4;14) and del(17p), with trisomy 3 appearing to have the greatest impact.^43,44  Another study looking at the impact of hyperdiploidy in this setting had conflicting findings.⁴⁵ 

Mutations associated with adverse outcome that may function as prognostic biomarkers have also been identified and include those in CCND1 and DNA repair pathway genes (TP53, ATM, ATR and ZFHX4).^15,42  Some mutations associated with a favorable outcome have also been identified eg, IRF4 and EGR1. Mutational analysis of genomic instability can also predict for adverse outcomes with increases in genome-wide loss of heterozygosity associated with adverse outcomes.⁴⁶ 

While DNA-based assays are able to identify individual lesions and markers of global genomic instability, RNA and GEP can be used to detect markers of increased proliferation and specific pathway expression changes.⁴⁷  The GEP scores of 70 genes (GEP70; MyPRS)⁴⁸  or 92 genes (SKY92)⁴⁹  have prognostic capabilities better than using any single lesion discussed above. They identify high-risk outcomes in ∼15% of patients at diagnosis. Their perceived limitations lie in the lack of widespread availability and computational analysis required to interpret the results.

Novel fusion genes have also been identified in myeloma using RNA sequencing data, and some have been associated with adverse outcome, such as CSNK1G2 and CCND1 (with shortened progression-free survival [PFS]) and MMSET and BCL2L11 (with shortened overall survival [OS]).⁵⁰ 

Other features of disease may also indicate high-risk outcomes for patients and hence act as prognostic biomarkers. The presence of plasma cells with blastic morphology, renal failure, extramedullary disease,⁵¹  and plasma cell leukemia at diagnosis all predict for worse outcomes. Circulating plasma cells are also associated with adverse outcomes, even if levels do not meet the criteria for plasma cell leukemia.^52,53  Recent studies have shown that the number and size of focal lesions on positron emission tomography computed tomography and magnetic resonance imaging also predict for a poor outcome independent of molecular features.⁵⁴ 

With advances in technology and the increase in size of the data sets examined, the information concerning the clinical impact of the presence of these molecular lesions has changed. This has resulted in a shift from using a single lesion to define high-risk disease to the use of 2 or 3 collaborating lesions. In addition, as so-called high-risk lesions can occur in up to 30% to 50% of patients, the need to identify a smaller group (eg, <15%) of patients who truly perform poorly regardless of therapy has become apparent. Such patients can be considered “ultrahigh-risk” and have been identified by the presence of more than 1 adverse lesion, by the R-ISS, or as double-hit myeloma (Figure 2).

Translocations and copy-number change associated with adverse outcome as described above have been demonstrated to be cumulative such that the presence of >1 lesion predicts for a worse outcome than 1 lesion alone.^20,24  In the MRC Myeloma IX study (Figure 2A), patients with >1 adverse lesion were termed ultrahigh-risk and comprised 15% of patients.

The Revised International Staging System (R-ISS) score built on this concept and incorporates β2-microglobulin, albumin (from the previously used ISS), and lactate dehydrogenase (LDH) with structural lesions to more accurately predict risk. Risk is categorized into 3 groups, from low-risk R-ISS group I (ISS stage I, no high-risk cytogenetic abnormality [CA] [del(17p) and/or t(4;14 and/or 14;16)] and normal LDH level) to high-risk R-ISS group III (ISS stage III and high-risk CA or high LDH level) (Figure 2B).⁵⁵  Between 10% and 18% of patients are classified as R-ISS III.^55-58  In this system, each feature contributes equally to the risk group determination. As more data become available and the understanding of how these features reflect myeloma biology increases, it can be envisaged that staging systems will be refined and features will be weighted.

Building on the scores defined above the Myeloma Genome Project incorporated NGS and structural abnormalities to better define risk (Figure 2C).³²  The study defined the highest-risk patients as “double hit” myeloma; that is, patients with 2 hits to the same gene (either loss of both alleles of TP53 [by mutation, deletion, or both] or with 2 extra copies of 1q resulting in amplification rather than a single gain). This group comprises 6% to 10% of patients and has a greater prognostic power than the R-ISS. A number of other groups have confirmed the importance of knowing both the copy-number and TP53 mutation status, and this lesion along with amp(1q) now represents the most recent refinement in myeloma risk prediction.

Interestingly, there remains a subset of patients carrying none of the molecular lesions discussed above who still relapse early. Such patients can be considered high risk phenotypically, as these patients in addition to having a short first PFS have a poor OS. Hence, having become apparent, such patients may require therapy different to standard treatment at first relapse.⁵⁹  Ongoing molecular studies may help to identify the currently unrecognized drivers in this early relapse group, as altering upfront therapy remains likely to have the greatest benefit on long-term survival.

The biomarkers described above were largely described in newly diagnosed patients. Many adverse risk biomarkers become more frequent at later stages of the disease but still retain prognostic significance. Other biomarkers that can be incorporated later include response to therapy, especially when assessed by quantification of minimal residual disease (MRD) in the bone marrow (by NGS or next-generation flow cytometry)^60-62  or by imaging techniques.⁶³ 

Aside from providing important prognostic information, the true benefit of identifying patients at high risk of early progression or death is to intervene and deliver therapy other than the standard treatment. Such approaches are being investigated in several clinical trials. One of the challenges for these studies is that response rates generally do not differ between patients with high-risk and standard-risk disease, as the natural history of high-risk patients is to respond well but relapse early. As such, the depth of response may not have the same prognostic importance in high-risk groups. In patients defined as del(17p) or ≥2 cytogenetic abnormalities, stringent complete response and MRD negativity did not translate into a superior PFS or OS.⁶⁴  Early data suggest that at deeper levels of MRD analysis, this drawback may be overcome, and confirmatory studies are awaited.⁶⁵ 

Trials concentrating on high-risk disease can be performed in 2 ways. In all-comer trials, the high-risk subgroup can be analyzed and reported separately and compared with non–high-risk patients, or trials can be specifically designed to optimize therapy for a prospectively recruited high-risk group (Figure 3A-B). The first approach provides reassurance on subgroup analysis that a given treatment shows as much benefit in the high-risk as standard-risk population, but claims of efficacy in the smaller high-risk population will be limited by the statistical power, and so large trials are required. The latter approach ensures studies are correctly powered to assess impact specifically in high-risk patients; however, the definition of high risk needs to be uniform and use reproducible biomarkers.

Subgroup analysis of previously reported clinical trials has led to several approaches being suggested for high-risk patients. This includes the observation that proteasome inhibitors overcome some of the adverse outcomes associated with t(4;14)±del(17p) lesions.⁶⁶  The initial data were based on a small subgroup analysis of the VISTA study. Subsequent studies, and a meta-analysis, confirmed that bortezomib-based induction results in improved outcomes vs non–bortezomib-based induction but does not fully overcome the adverse prognostic impact of these lesions.^67,68  A similar pattern is seen with lenalidomide maintenance after autologous transplant,⁶⁹  suggesting that novel agents can ameliorate, but not abrogate, adverse outcomes associated with high-risk disease. Studies in relapsed patients of the novel proteasome inhibitors carfilzomib and ixazomib also support this concept with a benefit over the control arm of the relevant studies in high-risk patients but continued suboptimal outcomes compared with standard-risk patients.^70,71  More recently, tandem autologous transplant, posttransplant consolidation, and maintenance have all proved effective for newly diagnosed high-risk patients compared with standard of care⁷²  and may to some extent attenuate unfavorable outcomes, but no strategy to date is able to completely overcome the adverse effect of high-risk lesions. Prospective recruitment of high-risk patients to dedicated protocols is needed.

High-risk patients are currently being recruited into a number of ongoing trials (Table 1). One example is the Total Therapy (TT) series of studies that initially started as risk-agnostic (TT1–TT3a/b) and later moved to high-risk studies (TT5, TT5b, and TT7). The phase 2 TT5 trial⁷³  recruited patients with GEP70-defined high-risk disease and delivered dose-dense chemotherapy, minimizing breaks between treatment phases by administering less intense therapy blocks aiming to prevent relapses that have been observed to occur during treatment breaks, such as during recovery from autologous transplant. Patients were compared with risk-matched patients in TT3, and no significant differences in survival were identified. However, the number of patients relapsing in the early treatment courses decreased, with patients tending to relapse during the later maintenance phase. The latter iterations of the protocol (TT5b and TT7) have concentrated on this phase of treatment and are incorporating newer proteasome inhibitors and immune-based approaches.

Other ongoing studies are examining intensification of induction, the use of autologous and allogeneic transplantation approaches, and immunotherapy approaches such as CAR T cells. The first approach is exemplified by the UK MUK9b trial (NCT03188172), the US 2015-12 TT7 trial (NCT03004287), and the German GMMG-CONCEPT trial (NCT03104842), which all combine a CD38 antibody, proteasome inhibitor, and lenalidomide as intensified upfront therapy along with prolonged courses of consolidation and maintenance.

The major difference between each of these studies is the definition of high risk (eg, GEP70 or single or combinations of genetic lesions), which will make subsequent direct comparisons of PFS and OS challenging. However, there is little doubt that concentrating on this subgroup of myeloma will be a rewarding area for both patients and investigators. Given the long PFS and OS for standard-risk patients, trials designed for standard risk require large numbers of patients and long follow-up to demonstrate a statistical and meaningful clinical improvement of the intervention. The high-risk patient is an area of unmet clinical need and is also the ideal situation to demonstrate the clinical activity of a novel agent or novel approach, and as such, it is anticipated that other novel immunotherapy approaches such as CAR T cells, CAR natural killer cells, bispecific antibody therapy, and antibody drug conjugates will move into first-line studies for high-risk myeloma over the coming years.

An alternative approach to altering upfront therapy is to use the prognostic biomarker of MRD postinduction to alter treatment at this time point. Studies addressing questions around intensification of therapy for MRD-positive patients or de-escalating therapy for MRD-negative patients are in development.

In contrast to prognostic biomarkers, predictive biomarkers forecast the likelihood of a favorable or unfavorable outcome with a specific agent. From mutations alone, it can be estimated that two thirds of patients have potentially actionable lesions with agents currently available or in development.¹⁵  Other targetable lesions include primary translocation events and/or protein expression patterns. To date, only a limited number have been studied in clinical trials, and these are discussed below.

Venetoclax is an inhibitor of the antiapoptotic protein BCL2. In vitro data show a higher sensitivity to venetoclax for cell lines and patient samples with t(11;14). This is likely due to the higher BCL2 to MCL1 expression ratio that correlates with the presence of the translocation.^74,75  Two early-phase clinical studies have been published (Table 2). The first studied single-agent venetoclax in multiply relapsed/refractory patients and demonstrated an overall response rate of 40% among t(11;14) patients.⁷⁶  The study also correlated BCL2:MCL1 and BCL2:BCL2L1 mRNA expression levels with responses and with t(11;14) status. The second study examined venetoclax in combination with bortezomib and dexamethasone. It demonstrated an ORR of 67% in all patients and 78% in t(11;14) patients. Similar to the single-agent study, patients with higher BCL2 expression had deeper responses and longer PFS.⁷⁷  The high efficacy of the combination in patients without the t(11;14) or high BCL2 expression was speculated to be due to bortezomib upregulating NOXA, a proapoptotic factor that neutralizes MCL1, resulting in an increased BCL2:MCL1 ratio and sensitivity to venetoclax.⁷⁵ 

These findings suggest that moving forward venetoclax may not be limited to the t(11;14) subgroup when used in combination with a proteasome inhibitor and an assay measuring BCL2:MCL1 or BCL2:BCL2L1 mRNA expression ratios may be beneficial as a predictive biomarker. This biomarker-driven strategy can be clearly seen with the trial combinations being examined in ongoing studies (Table 2), where those with proteasome inhibitor combinations are open to all comers, whereas single-agent studies are restricted to t(11;14) patients.

Sequencing studies have identified mutations in RAS pathway genes in ∼50% of patients (25% NRAS, 25% KRAS, and 4% BRAF), leading to the evaluation of RAS pathway inhibitors. These include BRAF inhibitors (eg, vemurafenib and dabrafenib), and MEK inhibitors (eg, trametinib and cobimetinib). The published experience to date is mostly limited to case reports and case series (Table 3) and provides encouraging evidence of activity in relapsed/refractory patients (eg, with responses seen in 16/40 patients with measurable disease,⁷⁸  although therapy in this study was often in combination). Several prospective studies are now underway and should provide a more comprehensive analysis of efficacy (Table 3).

Further targets have been identified that may predict response to therapeutic agents, although these are at earlier stages of development. Examples include IDH1/IDH2 mutations and IDH inhibitors,^79,80  loss of heterozygosity or ATM/ATR mutations and PARP inhibitors,⁴⁶  and FGFR3 mutations and FGFR3 inhibitors.⁸¹  Several are being studied in large umbrella studies such as the Multiple Myeloma Research Foundation MyDRUG study and the Canadian “CAPTUR” study. These incorporate agents targeted to a large number of molecular drivers with a solid preclinical rationale, often repurposed from other diseases. Ongoing analysis is identifying further targets that can be incorporated into such studies. For example, fusion genes have been identified in myeloma, and although rare (1%),^50,82  the majority contain a kinase domain, suggesting kinase inhibitors may have a potential role.

Recent solid tumor studies have shown that patients with a high mutation burden respond exceptionally well to PD1/PDL1 inhibitors. Generally, the mutation burden is lower in myeloma than in solid tumors, but a percentage of cases with a t(14;16) MAF translocation have an APOBEC signature and high mutational burden, providing a biological rationale to explore checkpoint inhibitors in this small group.

An important caveat of targeted agents used on the basis of predictive biomarkers is the presence of clonal heterogeneity, as when this is present not all cells contain the target lesion. Biopsy specimens from distinct sites of disease within the same patient at one time point, or from the same site at different time points, have been shown to be molecularly diverse.⁸³  These findings suggest that initiating events, present in 100% of clonal cells, may make good predictive biomarkers, whereas secondary, usually subclonal events would need to be in a high percentage of cells to be clinically useful. However, it could be hypothesized that where the targeted lesion is also of some prognostic importance, targeted therapy may still be clinically beneficial; for example, eliminating a high-risk clone and allowing a standard-risk clone to predominate may prove an effective therapeutic strategy, especially if the standard-risk clone can be eliminated with a standard therapy given in combination. Using this logic, the best predictive lesions to target (but for which targeted agents unfortunately do not yet exist) are those that are present in a high proportion of clonal cells and are associated with adverse outcomes. For instance targeting MMSET, the oncogenic driver in t(4;14), fulfills both these criteria but has proved difficult for drug design to date. Other options include MAF targeted approaches and the identification and targeting of lesions associated with gain(1q), amp(1q), or del(17p).

Another limitation of mutation targeted therapy is the lack of integration of RNA- and protein-level data into the decision-making process. Since drugs mostly act on proteins, the effect of mutations at the protein level is important. For example, mutations in recurrent sites known to cause pathway activation may be acted upon clinically, but variants of as-yet-unknown significance should be treated with caution. As more integrative molecular models become available, this potential limitation may be overcome.

Finally, targeting individual lesions that occur in low frequencies in patient populations poses a logistical problem for clinical trial design and requires novel trial approaches. For instance, for t(11;14) and RAS pathway mutations (present in 15% and 50% of patients, respectively), it is still possible to run lesion-specific trials. In contrast, attempting to target lesions such as IDH mutations (present in <2% of patients) will require multicenter and potentially multidisease collaborations. Such concepts are employed in umbrella studies (Figure 3C) or Basket studies (Figure 3D) where patients with different tumors are enrolled in the same protocol based on a molecular lesion identified. In addition, given the scarcity of some lesions in specific disease, there is an argument that more single-patient experiences also warrant publication.

The improvement in survival for myeloma patients over the last decade has mainly benefited low-risk patients, and now that median PFS for this group of patients is in excess of 8 years, a new approach to improving outcomes is required. Current approaches to personalize myeloma therapy take into account age and comorbidities but rarely consider molecular information. However, as the information concerning genetic analysis has become stronger, it can be postulated that one way to quickly improve outcomes further would be to incorporate such information into clinical algorithms. We have described 2 possible approaches by which this might be achieved. The first targets a cohort of patients with high-risk markers using intensified therapeutic approaches agnostic to molecular lesions. Such approaches include combinations of quadruplet or even quintuplet regimens and/or novel immunotherapy approaches such as bispecific antibodies, antibody–drug conjugates, and CAR T cells. This approach has the benefit of targeting patients with the worst outcomes and highest unmet clinical need. The challenge, however, is the lack of understanding about whether this is best achieved by incorporating ever-increasing numbers of additional agents or novel immunotherapy agents or whether the focus should be on designing more optimal treatment delivery approaches, such as different schedules and sequencing approaches using currently available agents.

The alternative strategy is to aim to use molecularly targeted agents that target lesions specific to an individual patient’s disease and therefore have a higher likelihood of efficacy while avoiding unnecessary toxicity. The knowledge of the molecular basis of myeloma is ever expanding, and so we can use this to define rationale drug targets as well as to use drugs already available for known targets. With respect to such predictive biomarkers, utilizing therapies targeted to either initiating lesions or lesions with a high cancer clonal fraction seem most likely to be effective. In addition, it seems likely that molecularly targeted agents will not be used alone; instead, these agents will be combined in specific subsets of disease with other agents that target more general plasma cell biological functions, such as proteasome inhibitors, immunomodulatory drugs, and monoclonal antibodies.

In closing, it is important to note that work to date concerning prognostic and predictive biomarkers has concentrated on genetic lesions within the plasma cell. Advances in protein technologies are occurring rapidly with the advent of tabletop analyzers,¹¹  mass spectrometry,¹²  next-generation flow cytometry, mass cytometry, and whole-proteome analysis. These technologies will allow the study the myeloma proteome as well as components of the bone marrow microenvironment and immune environment. With the increasing use of immune therapies, it seems likely that biomarkers related to these areas will be identified and will need to be incorporated into current models and treatment decisions.

The authors acknowledge that many investigators have contributed to this field over the years and apologize if, due to a lack of space, any colleague’s work has not been mentioned.

C.P. is a National Institute for Health Research Clinical Lecturer.

Contribution: C.P. and F.E.D. wrote the manuscript.

Conflict-of-interest disclosure: C.P. received consultancy and travel support fees from Amgen and Takeda Oncology; honoraria and travel support from Janssen; and consultancy, honoraria, and travel support from Celgene Corporation. F.E.D. received consultancy and honoraria from Amgen, AbbVie, Takeda, and Janssen and consultancy, honoraria, and research support from Celgene Corporation.

Correspondence: Faith E. Davies, Myeloma Center, University of Arkansas for Medical Sciences, 4301 W Markham St, Slot 816, Little Rock, AR 72205; e-mail: fedavies@uams.edu.