Modeling progression risk for smoldering multiple myeloma: results from a prospective clinical study

Abstract

The risk of progression to multiple myeloma (MM) from the precursor condition smoldering MM (SMM) varies considerably among individual patients. Reliable markers for progression to MM are vital to advance the understanding of myeloma precursor disease and for the development of intervention trials designed to delay/prevent MM. The Mayo Clinic and Spanish PETHEMA have proposed models to stratify patient risk based on clinical parameters. The aim of our study was to define the degree of concordance between these two models by comparing the distribution of patients with SMM classified as low, medium and high risk for progression. A total of 77 patients with SMM were enrolled in our prospective natural history study. Per study protocol, each patient was assigned risk scores based on both the Mayo and the Spanish models. The Mayo Clinic model identified 38, 35 and four patients as low, medium and high risk, respectively. The Spanish PETHEMA model classified 17, 22 and 38 patients as low, medium and high risk, respectively. There was significant discordance in overall patient risk classification (28.6% concordance) and in classifying patients as low versus high (p<,0.0001), low versus non-low (p= 0.0007) and high versus non-high (p<0.0001) risk. There is a need for prospectively validated models to characterize individual patient risk of transformation to MM.

Introduction

The detection of multiple myeloma (MM) precursor disease, classically specified as monoclonal gammopathy of undetermined significance (MGUS) or smoldering MM (SMM) [1], signifies an increased risk of developing MM relative to the general population. For the individual patient, however, the magnitude of this risk is not entirely clear. In the aggregate, the annual risk of transformation to MM from MGUS and SMM is approximately 1% and 10% per year, respectively [2,3], but the rate of progression to malignant disease varies considerably among patients. The International Myeloma Working Group has called for individual follow-up and development of early treatment trials for patients with high-risk SMM [4], and reliable models for predicting progression to MM are crucial to achieve these clinical goals [5].

There are currently two models which classify SMM patients as low, intermediate or high risk for progression to clinically detectable MM. Both risk models were derived from single-center retrospective cohort studies that identified, through multivariate analyses of historical clinical data, patient characteristics at the time of SMM diagnosis that were later associated with progression to MM. The Mayo Clinic model incorporates bone marrow plasma cell percentage and serum protein levels to stratify patients to risk groups [6]. Patients with SMM defined as low, intermediate and high risk by this model had median times to progression of 10, 5.1 and 1.9 years, respectively. The Spanish PETHEMA group (Programa de Estudio y Tratamiento de las Hemopat í as Malignas) model stratifies patient risk by assessing immuno paresis (reduction of one or more uninvolved immunoglobulin isotypes below the lower limit of normal) and proportion of plasma cells that are abnormal by flow cytometry to stratify patients, and found that low, intermediate and high risk patients had median times to progression of not reached (NR), 73 months (6.1 years) and 23 months (1.9 years) [7]. Here, we present the first head-to-head comparison of these models based on a well-characterized cohort of patients with SMM enrolled in a prospective natural history study. The aim of our study was to define the degree of concordance between these two models by comparing the distribution of patients with SMM classified as low, medium and high risk for progression.

Patients and methods

Between April 2010 and July 2012, 77 patients with SMM were enrolled in our ongoing prospective natural history study, NCT01109407. Th is study was approved by the Institutional Review Board of the National Cancer Institute, and written informed consent was obtained for all patients. Diagnosis of SMM and determination of risk score were performed using methods highly similar to those described by the models under comparison. Laboratory analyses included serum free light chain (FLC) assay and protein and immunofixation electrophoresis to identify and quantify circulating proteins, microscopy and immunohistochemical staining of bone marrow biopsy and aspirate samples to quantify plasma cell populations, and flow cytometric analysis of bone marrow samples for immunophenotyping of marrow plasma cells. Risk scores were assigned according to criteria for each model. Possible scores using the Mayo Clinic model were 1 (low risk), 2 (intermediate risk) or 3 (high risk) points; scores were assigned based on bone marrow plasma cell percentage greater than or equal to 10% (one point), serum M-protein greater than or equal to 3 g/dL (one point), and serum immunoglobulin FLC ratio either less than 0.125 or more than 8 (one point). One or both of the plasma cell percentage and M-protein criteria are required for diagnosis of SMM. Possible scores using the Spanish PETHEMA model were 0 (low risk), 1 (intermediate risk) or 2 (high risk) points; scores were assigned based on the proportion of bone marrow plasma cells with abnormal cell surface markers (less than 95%, zero points; greater than or equal to 95%, one point) and the presence (one point) or absence (zero points) of immunoparesis. The distribution of patients to low, intermediate and high risk groups was compared between the two models. Concordance proportions, significance levels and binomial 95% confidence intervals (CIs) for these comparisons were calculated using Microsoft Excel software and IBM SPSS Statistics 20.

Results

Overall, the Mayo Clinic and Spanish PETHEMA models for predicting risk of progression from SMM to MM concurred in the classification of 22 out of 77 patients, a 28.6% (95% CI: 18.8–40.0%) rate of agreement. The distribution of patients between the two models is summarized in Table I. Among patients with SMM classified as high risk by the Mayo model (n= 4), four patients (100%; 95% CI: 39.8–100.0%) were also classified as high risk by the Spanish model. Among patients classified as high risk by the Spanish model (n= 38), four patients (10.5%; 95% CI: 2.9–24.8%) were also classified as high risk by the Mayo model; the remaining 22 (57.9%; 95% CI: 40.8–73.7%) and 12 (31.6%; 95% CI: 17.5–48.7%) were classified as intermediate and low risk under the Mayo criteria, respectively. There was significant (p< 0.0001) discordance between the two models in classifying patients as high risk versus non-high risk (low risk plus intermediate risk groups). Similarly, among patients with SMM classified as low risk by the Mayo model (n= 38), 11 (28.9%; 95% CI: 15.4–45.9%) were also classified as low risk by the Spanish model; 15 (39.5%; 95% CI: 24.0–56.6%) and 12 (31.6%; 95% CI: 17.5–48.7%) were classified as intermediate and high risk by the Spanish criteria, respectively. Among patients classified as low risk by the Spanish model (n= 17), 11 (64.7%; 95% CI: 38.3–85.8%) were also classified as low risk by the Mayo model; the remaining six (35.3%; 95% CI: 14.2–61.7%) were classified as intermediate risk under the Mayo criteria. There was significant (p= 0.0007) discordance between the two models in classifying patients as low risk versus non-low risk (intermediate risk plus high risk groups). Comparisons between the two models are summarized in Figure 1.

Table I.

Distribution of 77 patients with SMM between two clinical models to predict risk of progression to MM.

	Spanish PETHEMA low	Spanish PETHEMA intermediate	Spanish PETHEMA high
Mayo Clinic low	11	15	12
Mayo Clinic intermediate	6	7	22
Mayo Clinic high	0	0	4
		Overall agreement: 22/77 (28.6%)

SMM, smoldering multiple myeloma; MM, multiple myeloma; PETHEMA, Programa de Estudio y Tratamiento de las Hemopatías Malignas.

Figure 1.

Discrepancy between models persisted when comparisons were made: (A) across all risk groups (including models restricted to low and high risk groups only); (B) between low and non-low risk (i.e. intermediate and high risk) groups; and (C) between non-high risk (low and intermediate risk) and high risk groups.

Discussion

In this first prospective head-to-head comparison between currently available clinical risk models for progression from SMM to MM (the Mayo Clinic [6] and Spanish PETHEMA [7] models), we assessed the distribution of patients with SMM classified as low, intermediate and high risk by the two models in a well-characterized prospective natural history study of patients with SMM. Per our study protocol, each patient was assigned risk scores based on both the Mayo Clinic and the Spanish PETHEMA models. Importantly, we found highly significant discordances between the two models. Indeed, the overall rate of agreement between the two risk models was less than 30%, reflective of the fact that many patients with SMM were both “ high risk” and “ l ow risk” at the same time when using the two models in parallel [6,7]. Perhaps the clinically most challenging observation was that among patients with SMM classified as high-risk by the Spanish PETHEMA model (n= 38), only ~10% were classified as high-risk by the Mayo Clinic model, while the remaining ~90% were classified by the Mayo model as intermediate and low risk. Furthermore, among patients with SMM classified as low risk by the Mayo Clinic model (n = 38), ~30% were classified as low-risk by the Spanish PETHEMA model, while the remaining ~70% were classified by the Spanish model as intermediate and high risk. Given the emerging trend to develop early intervention studies for patients with high-risk SMM [1], the observed discordances between the two models are important in that they highlight the dilemma of comparing results obtained across studies based on different models. In addition to the discordance between the two models, it seems reasonable to believe that there is a high degree of biological and clinical heterogeneity within the risk groups defined by the two models.

As stated above, the present study is the first prospective comparison of the Mayo Clinic [6] and Spanish PETHEMA [7] models for predicting risk of progression to malignancy among patients with SMM. Using a predefined study protocol, this investigation was designed to assess the degree of concordance between the two models by evaluating the distribution of patients with SMM classified as low, intermediate and high risk. Both the Mayo Clinic [6] and Spanish PETHEMA [7] models are retrospective, single-center cohort studies. As such, they used clinical information collected at SMM diagnosis and modeled the risk of developing MM in relation to included variables. In the present study, we used prospectively collected information on risk factors obtained at SMM diagnosis (as per our predefined study protocol), to allow for comparison with the published Mayo Clinic [6] and Spanish PETHEMA [7] models. One may speculate that cohort studies are not the optimal study design to capture risk profiles for progression from SMM to MM. For example, it is possible that the biological risk profile is similar for many patients with SMM at diagnosis, while there are evolving characteristics over time (e.g. acquired molecular lesions in the tumor cells and an altered microenvironment) [1]. If this is true, retrospective cohort studies (such as the Mayo Clinic [6] and Spanish PETHEMA [7] models)–designed to capture information at SMM diagnosis and then assess progression versus non-progression–may not be able to capture the answers we need. Alternative approaches include prospective cohort studies which monitor patients and collect data and samples in a longitudinal fashion. Such studies are expensive and time-consuming. The present investigation designed to assess the distribution of patients with SMM classified by the Mayo Clinic [6] and Spanish PETHEMA [7] models is based on a prospective natural history study of patients with SMM at the National Cancer Institute, National Institutes of Health, Bethesda, MD, USA (www.ClinicalTrials.gov; trial identifier NCT01109407). In a few more years, when the prospective natural history study has collected additional follow-up information for individual patients, it will allow assessments of various risk factors (including the Mayo Clinic [6] and Spanish PETHEMA [7] models) in parallel with molecular profiling, functional imaging and longitudinal measures of disease and host biology. Until such data are available, we feel it is important for patients and clinicians to be aware of the observed discordances between the two available clinical models. Clearly, there is a need for future studies designed to defi ne reliable risk models for individual patients.

In 2012, the standard of care for patients with SMM is expectant observation of laboratory and clinical characteristics [4]. No therapy is currently approved for the treatment of patients with these premalignant and variably progressive conditions. In general, predictive markers are important for making choices between treatment options [8]. In the absence of approved therapies for SMM, however, the development of markers predicting progression to MM is crucial for determining the intensity of clinical monitoring to detect transformation to malignant disease and the identification of patients who would benefit from enrollment in early treatment trials [5].

In summary, our results reveal significant discrepancies between the two clinical models (Mayo Clinic and Spanish PETHEMA models) [6,7] currently used in clinical practice to determine individual patient projected risk of progression from SMM to MM. These findings highlight the need for the development and prospective validation of molecularly derived risk factors that can distinguish between clinically similar phenotypes to more accurately characterize the risk of transformation to MM [9,10]. Future models, based on platforms such as reliable biological markers and functional imaging, will improve counseling and follow-up for patients with SMM and will help to identify patients who would most benefit from enrollment in clinical trials for the treatment of high-risk myeloma precursor states (“early myeloma” ) [11]. Such insights will also play an important role in our understanding of the pathogenesis of the disease, allowing us to define novel treatment targets with the ultimate goals to delay and prevent progression from SMM to MM.