Molecular studies have shown that multiple myeloma is a highly genetically heterogonous disease which may manifest itself as any number of diverse subtypes each with variable clinicopathological features and outcomes. Given this genetic heterogeneity, a universal approach to treatment of myeloma is unlikely to be successful for all patients and instead we should strive for the goal of personalised therapy using rationally informed targeted strategies. Current DNA sequencing technologies allow for whole genome and exome analysis of patient myeloma samples that yield vast amounts of genetic data and provide a mutational overview of the disease. However, the clinical utility of this information currently lags far behind the sequencing technology which is increasingly being incorporated into clinical practice. This paper attempts to address this shortcoming by proposing a novel genetically based “traffic-light” risk stratification system for myeloma, termed the RAG (Red, Amber, Green) model, which represents a simplified concept of how complex genetic data may be compressed into an aggregate risk score. The model aims to incorporate all known clinically important trisomies, translocations, and mutations in myeloma and utilise these to produce a score between 1.0 and 3.0 that can be incorporated into diagnostic, prognostic, and treatment algorithms for the patient.
Molecular studies have made it apparent that multiple myeloma is not a single disease entity but rather a collection of genetically diverse disease subtypes that manifest clinically as the clonal proliferation of plasma cells. With this, it is clear that a universal treatment approach is not sufficient and that patient management should be targeted towards the specific genetic disease subtype(s) a patient harbours. To fully achieve this, along with the somewhat established approaches of conventional cytogenetics and fluorescent
Despite a consensus within the field that the integration of genetic information into the diagnosis, treatment, and prognosis of myeloma would be of great benefit, there is currently no universally accepted system to achieve this. The main challenge in designing such a model for clinical use is that often an abundance of complex data must be simplified into an intuitive and useful form whilst remaining valid and applicable. This balance is difficult to obtain, as models which are too complex become clinically unintelligible whereas those which are too simple lose accuracy and informative power. As a variety of molecular techniques are available to analyse the myeloma genome, a proposed genetic based model must be able to either incorporate the findings from a range of methods or be specifically designed to unite one. For reasons discussed hereafter, the RAG model is designed to accommodate multiple analytical methods but crucially expands on other risk stratification systems by attempting to accommodate whole genome sequencing (WGS)/whole exome sequencing (WES) data, as although these techniques are currently still highly experimental and cannot currently be used to accurately inform treatment decisions or risk/prognosis, it appears likely that due to their power and increasing accessibility these techniques will play a key role in the workup of myeloma patients in the future. Furthermore, although the RAG model is presented here as a concept for risk stratification, the benefits of which are to optimize outcomes and stratify treatment regimes in order to minimize toxicity, it is possible that in the future such a model will inform many different areas of myeloma medicine such as identifying new disease biomarkers and therapeutic targets and helping better genetically categorise/diagnose myeloma disease subtypes. The advances that these areas would bring to myeloma treatment alongside risk stratification make it even more pressing for models to be developed which accurately interpret and utilise genetic information.
To establish a context for the requirement for a genetic risk stratification model, such as the RAG model, a literature review covering current myeloma risk stratification and prognostication follows.
For any malignancy to have a simple, accurate, and easily applicable universal staging system to inform prognosis is of obvious benefit. The first such staging system for myeloma, developed by Durie and Salmon in 1975 [
The international staging system for myeloma [
Stage | Criteria | Median survival (months) |
---|---|---|
I | Serum |
62 |
|
||
II | Neither stage I or III* | 44 |
|
||
III | Serum |
29 |
With the ISS proving that measurements of disease burden, renal function, and patient “fitness” can accurately guide myeloma prognosis, a number of studies postulated whether FISH could be used to similar effect. In 2009, the International Myeloma Working Group (IMWG) evaluated the data published on the role of FISH in myeloma prognostication to formulate a consensus [
Recently, an important study conducted by Boyd et al. used outcome data from the Medical Research Council (MRC) myeloma IX trial to conduct multivariate analysis on the interaction of genetic aberrations identified via a comprehensive FISH panel [
As the ISS considers different factors to FISH, several groups have attempted to combine these two methods to adjudge whether a more optimal prognostic model can be developed. These studies are important as any model utilising genetic data for prognosis will ultimately require some consideration of tumour burden and host “fitness” to optimise accuracy. Boyd et al., using their initial risk stratification groups from the MRC IX trial, integrated the ISS and identified that an ultrahigh-risk group could be defined at diagnosis by ISS stages II or III plus >1 of their defined adverse lesions [
Despite the contribution FISH analysis has made to improving myeloma understanding, a particular limitation to the technique is that it may only detect predefined genetic lesions determined by the specific probes employed. As the majority of key myeloma FISH lesions are already likely defined, this limitation is unlikely to prevent a well-constructed FISH library from becoming part of a universally accepted prognostic/risk stratification system. However, this limitation does prevent FISH from being utilised as a parallel screening tool to identify unknown genetic aberrations across the whole genome, an attribute which sequencing technology possesses and one which would improve therapeutic development and biological understanding. This point is well demonstrated by WGS recently identifying previously unobserved
As FISH proved useful in myeloma prognostication, studies began to assess whether gene expression profiling (GEP) could also be used effectively. Shaughnessy et al. first assessed 532 newly diagnosed myeloma patients with GEP and, using log-rank tests of expression quartiles, identified 70 genes which were linked to shorter durations of remission, event-free survival (EFS), and OS [
A second study conducted by Decaux et al. investigated the GEP of 182 myeloma patients and identified the 15 strongest genes associated with length of survival [
Interestingly, of the 17 and 15 genes identified by the two aforementioned studies, none were shared, a finding which demonstrates the genetic complexity of myeloma. Furthermore, in these GEP studies, as for FISH analysis, the gene signatures used to define high-risk disease were not always specific for a given clinical outcome, a problem which may again lead to potential over- or undertreatment with chemotherapy. Additionally, gene mapping arrays only analyse the genetic signature of the predominant clone, whereas WGS can provide semiquantitative analysis of the size of the clonal population carrying a given aberration allowing for characterisation of disease substructure [
The mayo stratification of myeloma and risk-adapted therapy model (mSMART) is a set of consensus guidelines developed by over 20 Mayo clinic myeloma physicians which aims to provide recommendations for the treatment of myeloma patients [
The mSMART risk stratification system in active myeloma [
High risk | Intermediate risk | Standard risk |
---|---|---|
FISH | FISH | FISH |
del(17p) | t(4;14) | t(4;14) |
t(14;16) | t(6;14) | |
t(14;20) | ||
GEP | Cytogenetic del(13) | All other patients |
High-risk signature | ||
Hypodiploidy | ||
Plasma cell labelling index ≥3% |
As is evident, the inclusion of genetic information into a myeloma risk stratification system is both required and achievable at a given level. The biggest obstacle to progression in this area is how best to interpret and use the large amounts of complex data that the aforementioned molecular techniques generate. Answers to this problem must soon be developed as the accessibility, speed, and cost of the techniques will soon mean they are available for utilisation in the workup of a newly diagnosed myeloma patient. The following subsection therefore outlines a proposed genetic risk stratification model, developed as a concept, to try and create a starting point for how genetic information, particularly sequencing data, may be utilised.
Many groups have produced different biological classification systems for myeloma developed through a range of molecular techniques [
RAG model categories.
RAG category | Initiation/primary event | Progression/secondary event |
---|---|---|
Hyperdiploidy |
|
|
t(4;14) |
|
|
t(6;14) |
|
|
t(11;14) |
|
|
t(14;16) |
|
|
t(14;20) |
|
|
+1q |
|
|
Del(1p) |
|
|
Del(11q) |
|
|
Del(12p) |
|
|
Del(13/13q) |
|
|
Del(14q) |
|
|
Del(16q) |
|
|
Del(17p) |
|
|
Secondary t(8;14) |
|
|
Bone disease |
|
|
Proliferation |
|
|
Apoptosis and NF- |
|
|
Differentiation |
|
|
DNA repair |
|
|
RNA editing |
|
|
Epigenetic |
|
Using conventional cytogenetic and FISH techniques, it has long been recognised that gross structural and numerical chromosomal changes are important in defining myeloma risk. With the recent employment of GEP and WGS/WES, however, the important genes in these instances are being identified whilst other salient genes, disrupted through more subtle structural changes/mutations, are becoming apparent. With this, the RAG model aims to expand on current systems of risk stratification by incorporating individually mutated genes deemed to be important in driving myeloma pathogenesis. For this, the genes relevant to the categories outlined in Table
RAG model categories and their candidate genes.
RAG category | Candidate genes | Reference |
---|---|---|
Hyperdiploidy |
|
[ |
t(4;14) |
|
[ |
t(6;14) |
|
[ |
t(11;14) |
|
[ |
t(14;16) |
|
[ |
t(14;20) |
|
[ |
+1q |
|
[ |
Del(1p) |
|
[ |
Del(11q) |
|
[ |
Del(12p) |
|
[ |
Del(13/13q) |
|
[ |
Del(14q) |
|
[ |
Del(16q) |
|
[ |
Del(17p) |
|
[ |
Secondary t(8;14) |
|
[ |
Bone disease |
|
[ |
Proliferation |
|
[ |
NF- |
|
[ |
Differentiation |
|
[ |
DNA repair |
|
[ |
RNA editing |
|
[ |
Epigenetic |
|
[ |
The model for calculating a RAG score is outlined in Figure
RAG scores and their risk stratification groups.
RAG score | RISK Stratification |
---|---|
2.5–3.0* | High-risk* |
2.0–2.5** | High-intermediate-risk** |
1.5–2.0** | Low-intermediate-risk** |
1.0–1.5*** | Low-risk*** |
The RAG model. The genes and categories selected for the RAG model are placed into their respective red, amber, and green groups. To generate a RAG score, the average score for lesions correlating between a patient sample and the model is calculated. The RAG score is then used for risk stratification.
To determine the selected genes and categories to compile the RAG model, the evidence from studies linking genetic aberrations to prognosis/risk was reviewed. The criteria used to evaluate the evidence was based on established methods used by other groups and is represented in Table
Classification system for levels of evidence and grades of recommendation [
Type of evidence | |
---|---|
Level I—meta-analysis of multiple, well-designed, controlled studies. Randomized studies with low type 1 and type 2 errors (high power) are also considered. | |
Level II—evidence obtained from at least one, well-designed experimental study. Randomised trials with high type 1 and/or type 2 errors (low power) are also considered. | |
Level III—well-designed, quasiexperimental studies such as nonrandomised, controlled single-group, prepost, cohort, time, or matched case-control series. | |
Level IV—well-designed, nonexperimental studies, such as comparative and correlational descriptive and case studies. | |
Level V—case reports and clinical examples. | |
|
|
Grade of recommendation | |
|
|
Grade A—evidence of level I or consistent findings from multiple levels II, III, and IV studies. | |
Grade B—evidence of levels II, III, or IV with generally consistent findings. | |
Grade C—evidence of levels II, III, or IV but findings are inconsistent. | |
Grade D—minimal or no systematic empirical evidence. |
It is envisaged that cost and availability permitting, a model such as the RAG model will be used both at diagnosis and at later disease stages/relapse to build a genetic pattern over time. As it is recognised that myeloma progresses in union with an ever-changing genetic landscape through the advancement and regression of clonal tides [
To help visualize the RAG score and facilitate a better understanding of its meaning in the clinical environment, it is proposed that the score will be represented as a RAG “pizza” plot, whereby the genetic categories are displayed as colour plots as outlined in Figure
RAG “pizza” plots. The RAG “pizza” plots are colour plots which represent how the RAG score will be presented. The size of the segment each aberration represents is proportional to the “weighting” that lesion is given in calculating the RAG score; that is, a red group lesion will be represented by a larger segment when compared to the segments of amber and green lesions. Representation of the RAG score as a “pizza” plot helps to visualize the score and improve understanding.
It is recognised that by simply stating candidate genes within the RAG model that the system fails to outline the specific mutations relevant within that gene. This is important, as the functional consequence of two different mutations within a gene may well be different. Furthermore, it is recognised that synonymous mutations may occur within genes and therefore impart no alteration to protein structure. However, in a WGS/WES study by Chapman et al., [
The RAG model is presented here as a concept for myeloma risk stratification system based on the genetic characteristics of the disease. The justification for the need of such a model is apparent from reviewing the literature which demonstrates that the underlying genetic makeup of an individual’s disease contributes significantly towards behaviour and outcome and can be used to inform risk. The RAG model has been designed to begin the process of bridging the gap between experimental research and clinical medicine whereby there is a necessity to simplify large amounts of complex data into a useful and formative structure. Although principally designed to inform disease risk, it is envisaged that in the future a model such as the RAG model would be able to also contribute towards the identification of new disease biomarkers and therapeutic targets and aid in the genetic categorisation and diagnosis of myeloma disease subtypes.
In conclusion, despite its lack of clinical validation and optimisation, we believe that the RAG model has enormous potential to simplify otherwise complex genetic data to guide clinicians and improve treatment outcomes for patients. For this reason, we wish to present the model as a prototypic concept of how state-of-the-art genomic data can complement other established technologies to usher in an era of personalized myeloma medicine.
The authors declare that there is no conflict of interests regarding the publication of this paper.