Validation of multiple myeloma risk stratification indices in routine clinical practice: Analysis of data from the Czech Myeloma Group Registry of Monoclonal Gammopathies

Abstract This study used data from the Czech Myeloma Group Registry of Monoclonal Gammopathies to validate the International Myeloma Working Group (IMWG) and revised International Staging System (R‐ISS) indices for risk stratification in patients with multiple myeloma (MM) in clinical practice. Patients were included if they had symptomatic MM, complete data allowing R‐ISS and IMWG staging (including cytogenetic information regarding t(4;14), t(14;16), and del(17p)), and key parameters for treatment evaluation. Median overall survival (OS) in included patients (n = 550) was 47.7 (95% CI: 39.5‐55.9) and 46.2 (95% CI: 38.9‐53.5) months from diagnosis and initiation of first‐line therapy, respectively. Patients categorized as higher vs lower risk had reduced survival; median OS from diagnosis was 35.4 (95% CI: 30.5‐40.3) vs 58.3 (95% CI: 53.8‐62.9) months in high‐risk vs other patients (IMWG; P = .001) and 34.1 (95% CI: 30.2‐38.0) vs 47.2 (95% CI: 43.4‐51.0) months in Stage III vs Stage II patients (R‐ISS; P < .001). In conclusion, IMWG and R‐ISS risk stratification indices are applicable to patients with MM in a real‐world setting.


| INTRODUCTION
Multiple myeloma (MM) is a clinically heterogeneous disease, as evidenced by considerable variation in rates of response to treatment and overall survival (OS); indeed, OS in patients with MM has been shown to range from a few months to more than a decade. 1 Much of the clinical heterogeneity of MM is thought to arise from multiple genomic events that result in tumor development and progression. 2 These include genetic and epigenetic alterations, including point mutations, translocations (eg, t(4;14) and t(14;16)), deletions (eg, del(17p)), 3 aberrant DNA and histone methylation, 4 and/or abnormal microRNA expression. 5 Several groups have worked to develop systems that use prognostic markers to stratify patients with MM into homogeneous survival subgroups. 1,[6][7][8][9] Risk stratification facilitates prognostication, allowing patients to be categorized as having lower risk (ie, longer OS) or higher risk (ie, shorter OS) disease.
In 2014, the International Myeloma Working Group (IMWG) recommended a new staging system based on (1) the criteria of the International Staging System (ISS) (which was first presented in 2005 and was based on high serum β 2microglobulin and low serum albumin levels) 10 and (2) cytogenetic information. 11 With the IMWG system, patients are categorized as low risk (ISS Stage I/II and the absence of t(4; 14), del(17p), and 1q21 and age <55 years), intermediate risk (neither low risk nor high risk), or high risk (ISS Stage II/III and t (4;14) or del(17p)). In 2015, a revised version of the ISS (R-ISS) was presented, 1 which incorporated chromosomal abnormalities detected by interphase fluorescent in situ hybridization (iFISH) and serum lactate dehydrogenase (LDH). With the R-ISS, patients are categorized as Stage I (ISS Stage I and standard-risk chromosomal abnormalities by iFISH and normal LDH), Stage II (neither Stage I nor Stage III), or Stage III (ISS Stage III and either high-risk chromosomal abnormalities [del(17p) and/or t (4;14) and/or t(14;16)] by iFISH or high LDH). Despite the availability of these and other risk stratification indices, however, no single tool is uniformly used in patients with MM.
This study was undertaken to validate the IMWG and R-ISS indices for risk stratification in a cohort of patients who were being treated in routine clinical practice, including in subgroups of patients who received new drugs and who underwent ASCT. Data for the study were derived from the Czech Myeloma Group Registry of Monoclonal Gammopathies (RMG), a large-scale project that aims to monitor the diagnosis and treatment of monoclonal gammopathies in the Czech Republic and Slovakia.

| SUBJECTS AND METHODS
This was a retrospective registry-based analysis that used data derived from the RMG (https://rmg.healthregistry.org/). Details regarding the registry have been presented previously 12,13 ; in brief, the registry, which has become one of the flagship projects of the Czech Myeloma Group, was established in 2007 and retrospectively and prospectively collects data from patients with MM, monoclonal gammopathies of undetermined significance, amyloid light-chain amyloidosis, and Waldenström's macroglobulinemia. In the Czech Republic, all patients with myeloma who are diagnosed undergo examination in one of the registry participating centers and the patient is registered at that time. Every year there is a recruitment of 200-300 newly diagnosed patients which roughly corresponds to expected incidence of myeloma in Europe and patients who die early (after diagnosis) are captured as well. Parameters of interest captured by the registry include demographic data, disease characteristics, treatment choice(s), and response to treatment (including OS, TTP [time to progression], PFS [progression-free survival], and time to next treatment) for each line of treatment and each treatment interval. Information regarding diagnosis, treatment response, and time-to-event endpoints is assessed according to current IMWG criteria. The registry is regularly monitored, and data are validated by an external monitor. Upon entering the RMG, all patients are required to sign an informed consent form for data collection; the consent forms are approved by the ethics committees of participating hospitals. Patients were included in the current analysis if they had symptomatic MM, a complete dataset of parameters allowing R-ISS and IMWG staging, and key parameters for treatment evaluation; parameters required for calculation of IMWG and R-ISS scores included levels of albumin, β 2 -microglobulin, and LDH and information regarding chromosomal abnormalities. As we were not able to obtain +1q data (these were not uniformly reported since the beginning of the registry), we simplified the analysis of IMWG score as high risk vs others. Cytogenetic data were obtained from multiple laboratories with various cutoffs for positivity reporting; depending on the threshold defined by each local laboratory, patients were considered positive for a translocation when it was present in a percentage ranging from 5% to 20% (with the most frequently reported cutoff value being 20%, as used in previous research). 1 High LDH was defined as a serum level greater than the upper limit of normal. β 2 -microglobulin testing and albumin testing were as per nationwide standardized methods.
Data were described by absolute and relative frequencies for categorical variables and by median (5th to 95th percentile) values for quantitative variables. For comparisons, the maximum likelihood Chi-square test was used for categorical variables, while the Kruskal-Wallis test was used for continuous variables. Using the Kaplan-Meier method, TTP and OS were plotted. Kaplan-Meier estimates were completed using the Greenwood confidence interval (CI), with a log-rank test used to estimate the statistical significance of any differences between curves. A Cox proportional hazards model was performed to explore the univariate significance of risk factors. P values less than 0.05 were considered statistically significant (all tests were two-sided). Analyses were performed using SPSS software (IBM Corp., released 2013, IBM SPSS Statistics for Windows, Version 23.0, Armonk, NY, USA) and software R version 3.2.3 (www.r-project.org).

| RESULTS
Data from the RMG showed that the mean number of newly diagnosed cases of MM per year was 478; based on a population of approximately 10 million people, the annual incidence was estimated to be 4.8 per 100 000 individuals.
To be included in the current analysis, it was necessary that patients had valid data for chromosomal abnormalities. Patients were considered high risk if they were positive for at least one of the following genetic markers: t(4;14), t(14;16), or del(17p) and standard risk if they were negative in all the markers; 3460 patients in the RMG were excluded due to failure to meet this criterion. An Table 2 summarizes IMWG risk stratification and demographic and disease characteristics of all patients at diagnosis. Table 3 summarizes R-ISS risk stratification and demographic and disease characteristics of all patients at diagnosis. Regarding IMWG risk classification, statistically significantly higher proportion of Durie-Salmon Stage III patients is present in the high-risk group as well as patients with renal failure. Logically albumin, creatinine, and β 2 -microglobulin levels differ significantly too ( Table 2). Regarding R-ISS risk stratification, there were significant differences across the 3 risk groups for multiple patient characteristics (Durie-Salmon stage and substage, bone marrow aspiration cytology, and various clinical laboratory tests (eg, hemoglobin, thrombocyte count, albumin, creatinine, β 2microglobulin, LDH, and CRP) ( Table 3). For certain patient characteristics (eg, age, performance status, and M-protein type), there was a significant association with R-ISS risk stratification but not with IMWG risk stratification.
An analysis of the data showed that patients who were categorized as higher risk with both risk stratification indices had reduced survival. The median OS from diagnosis in patients who were categorized as high risk according to the IMWG criteria was 35.4 months (95% CI: 30.5-40.3), compared with 58.3 months (95% CI: 53.8-62.9) for all other patients (P = .001; Figure 1). Median OS from diagnosis in patients who were categorized as R-ISS Stage III was 34.1 months (95% CI: 30.2-38.0), compared with 47.2 months (95% CI: 43.4-51.0) for patients categorized as Stage II (P < .001). Median OS from diagnosis for patients categorized as R-ISS Stage I was not reached (Figure 1). An analysis of OS and PFS for both of the risk stratification systems is shown in Table 4. The data showed a statistically significant survival disadvantage for higher vs lower risk patients for all comparisons.    An analysis of OS and PFS according to ISS Stage, cytogenetic factors and LDH is shown in Table 5. The data showed a statistically significant survival disadvantage for the presence vs the absence of t(4;14) and the presence of high-risk vs standard-risk chromosomal abnormalities. The same was true for the presence of Stage II/III vs Stage I disease according to ISS criteria and for a per-unit increase in LDH. There was a statistically significant survival disadvantage (PFS only) associated with the presence vs the absence of del(17p). In contrast, a statistically significant survival disadvantage (OS and PFS) was not found for the presence vs the absence of t(14;16) even though it has been described as a high-risk chromosomal abnormality.
Regarding the IMWG criteria, median OS from diagnosis in patients who were categorized as high risk and who underwent ASCT was 62.  Table 6. There was a survival disadvantage for higher vs lower risk patients for all comparisons. Results were uniformly statistically significant in the group of patients who did not undergo ASCT.

| DISCUSSION
It is important that any risk stratification scheme is suitably validated. As part of its development, the original ISS was tested and found to be effective in patients from different geographic regions (North America, Europe, and Asia), patients of different ages (<65 and ≥65 years), patients receiving different treatments (standard therapy or autologous stem cell transplantation [ASCT]), and patients at different study sites (single sites and cooperative groups). 10 In the same way, the R-ISS was tested in patients of different ages (≤65 and >65 years) and in patients receiving different treatments (ASCT, proteasome inhibitors [PIs], or immunomodulatory drugs [IMiDs]). 1 Although the ISS risk stratification tool has been evaluated outside of a clinical trial framework, 14 the majority of patients (69.1%) on which the system was based were participating in a clinical trial 10 ; similarly, the R-ISS was based wholly on patients enrolled in experimental trials. 1 This is relevant because patients in cancer trials may not be representative of the overall patient population and derivation of risk stratification criteria based on a highly selected set of patients may limit their applicability in the real world. As a result, testing of risk stratification criteria in populationbased studies is important. Validation in such studies is also important to determine the day-to-day practicality of conducting the tests that form the risk stratification criteria.
The current analysis was undertaken to validate the IMWG and R-ISS indices for risk stratification in patients with MM in a real-world setting. Using data from the Czech Myeloma Group RMG, it was shown that the prognostic value of the IMWG and R-ISS indices for risk stratification is applicable to patients treated in routine clinical practice. These results extend the findings of studies conducted in patients who were, for the most part, participating in clinical trials. 1,6 Together, this study, and others of a similar nature, 15,16 indicate that the IMWG and R-ISS indices for risk stratification are applicable to a broad spectrum of patients with MM.

| Impact of LDH and cytogenetic abnormalities
The impact of LDH and cytogenetic changes in patients with MM has been well documented. Within the current cohort of patients, 31.7% had elevated LDH (ie, levels above the upper limit of normal), which is higher than what has been N = 555 a,b n (%) Median (5th-95th percentile)  17 ; similarly, in a study of 203 patients with symptomatic MM, 7% had elevated LDH. 18 The higher rate of elevated LDH reported here may suggest shorter OS for our cohort of patients, compared with the original dataset from which the R-ISS system was developed (where only 13% of evaluable patients had elevated LDH). 1 In an interesting manner, in the current analysis, when high vs normal LDH was evaluated as a stand-alone marker in a univariate analysis undertaken to explore the significance of different risk factors, it did not show any significant impact on OS. Del(17p) is often encountered in patients with MM and is considered a predictor of adverse outcomes (ie, it has a negative impact on PFS and OS). 19,20 In the current cohort, we failed to show a significant impact of del (17p) 23 They demonstrated that 0% of MM patients without del(17p) presented with a mutation in TP53; conversely, not all (ie, only 37%) of patients with del(17p) exhibited a TP53 mutation. This fact may explain why only a cytogenetic test without further molecular analysis might fail to show a significant impact on patient survival. Nonetheless, the results regarding LDH, del(17p), and t(14;16) and OS (and t(14;16) and PFS) in the current analysis are inconsistent with previous research that showed chromosomal abnormalities (del(17p) and/or t(4;14) and/or t(14;16)) and LDH each had significant prognostic value in terms of survival. 1,16

| Impact of treatment administered
Both indices of risk stratification appeared to have an effect (ie, have prognostic value) on OS in the overall study population. This finding held true for patients who received new drugs and, to some extent, patients who did not receive new drugs. Although statistical significance was not demonstrated for all comparisons in the latter group, hazard ratios suggested an increased mortality risk in higher vs lower risk patients.
Although the prognostic value of both risk stratification indices appeared to apply to patients who had not undergone ASCT, the picture was less clear in those who had undergone this form of treatment, particularly for the IMWG score; notably, we failed to show an impact of the IMWG score on OS in a transplant setting. The original work describing the value of IMWG risk stratification was based on a pooled analysis of patients treated with transplantation or conventional treatment; however, for the subgroup treated with high-dose melphalan and ASCT, the model worked just as well. 6 A possible explanation for the findings of the current study could be that the small number of ASCT patients, particularly within the high-risk groups, limited the validity of the indices; indeed, only 32.6% of patients overall underwent ASCT (compared with 59% of patients in the original IMWG work), with rates of 28.7%  in IMWG high-risk patients and 24.8% in R-ISS Stage III patients. Possible reasons for the low uptake of ASCT in the current study may include progression or death during induction chemotherapy or a suboptimal health status. In particular, poor performance status reported in our cohort of Stage II and Stage III R-ISS patients could potentially worsen the outcome of these patients (10.6% and 14.3% PS 4) due to the fact that proper treatment could not be delivered. Performance status is an important independent prognostic indicator of survival 24 not reflected in R-ISS. There is also the fact that R-ISS patient Stage II and Stage III are older than those of Stage I and poor performance status as well as more cytogenetic changes are expected in this population. The process of ASCT itself might also have influenced the prognostic value of risk stratification; that is, ASCT might partly overcome the predictive power of conventional predictors, changing a patient's prognosis vs the period prior to ASCT. No maintenance treatment is currently approved in our country; therefore, no data on possible impact of this issue are available.

| Overall results
In the current dataset, median OS in the overall study population was 47.7 months (95% CI: 39.5-55.9), which is somewhat shorter than what has been reported by other investigators. 1,16 Data from a study that evaluated the R-ISS algorithm in 3,060 patients with newly diagnosed MM reported a median OS of 83 months for R-ISS Stage II patients and 43 months for R-ISS Stage III patients. 1 In another study that aimed to validate the R-ISS in an independent cohort of 475 unselected, consecutive patients with symptomatic MM treated with contemporary regimens, the estimated median OS was 63 months. 16 Such between-study differences in OS may be explained by several factors, including the difference in study setting. The current analysis included unselected patients in a real-world setting, while one of the comparator studies included selected patients who were participating in experimental trials. 1 This is relevant as data show that clinical trial participants are typically younger and healthier than the overall cancer population, resulting in differences in OS between trial participants and real-world patients. 25,26 Indeed, the median age in the current analysis was 66 years, compared with 62 years in the comparator study undertaken in clinical trial participants. 1 Another contributing factor to the lower OS rates in the current analysis could be the markedly lower proportion of patients who had undergone ASCT in the current study (32.6%) vs the comparator study in patients in clinical trials (60%). 1 Numerous studies have shown that intensive therapy with ASCT is associated with improved survival, compared with conventional chemotherapy in patients with newly diagnosed MM; however, this type of treatment is typically reserved for patients who are aged 65 years or younger. 27 In the other comparator study, 16 which was also conducted in a real-world setting, only 36% of patients were reported to have undergone ASCT, which is similar to the rate in the current study. In an interesting manner, OS in R-ISS highrisk patients in both of the real-world studies was comparable (29 months in the earlier study vs 34.1 months in the current the median OS of high-risk (R-ISS Stage III) patients was 34.1 months in the current analysis, compared with an estimate of less than 2 years (despite the use of novel agents) in patients who are considered high risk according to IMWG criteria. 11 On the one hand, this increase in OS highlights the progress made in diagnosing and treating patients with MM, even those with a poor prognosis. On the other hand, it highlights the need for predictors of early relapse so that treatment can be adjusted in order to extend survival.

| LIMITATIONS OF THE STUDY
The current study has an obvious limitation in that the results are based on patients in central and eastern Europe and therefore may not be generalizable to the global MM population as a whole. In addition, owing to missing data relating to genetic markers of interest, it was not possible to include all of the patients in the RMG in the current analysis; although this may introduce bias, it should be noted that the 555 patients included represents a substantial sample size, giving confidence in the conclusions drawn from the data. However, it is acknowledged that there were small sample sizes for some of the subgroup analyses (eg, the low number of patients who did not receive new drugs and who received ASCT). In an interesting manner, information about the number of patients who were ineligible to participate in the current analysis because of missing data relating to genetic markers provides a useful insight into the day-to-day practicality of evaluating cytogenetic abnormalities in a real-world setting. Such challenges, however, are not restricted to clinical practice; in the previously described study that evaluated the R-ISS algorithm, more than 30% of patients in experimental clinical trials did not have chromosomal abnormality data available (in addition to simultaneous ISS and LDH data). One of the major limitations is the heterogeneity of cutoff levels for evaluation of chromosomal abnormalities. It is important to note that this is an evolving field and the cutoffs used during the past is not valid nowadays, but the retrospective nature of data does not allow us to precisely differentiate the cutoffs. Furthermore, as with all retrospective registry-based approaches, a central limitation is the reliance on accurate and complete patient records and data collection, and also the possibility of researcher/physician selection or information bias.

| STRENGTHS OF THE STUDY
The RMG includes information from a broad range of realworld MM patients across various categories (eg, use and nonuse of new drugs and ASCT). Owing to the noninterventional nature of the RMG, no specific drugs or treatment procedures are required for patients to be included, ensuring that patients are in a naturalistic setting where treatment choice is based on the current standard of practice and/or available treatments. Second, the registry is robust in that data are prospectively collected, regularly monitored, and validated by an external monitor. Third, the database allows an analysis of multiple baseline factors that may influence OS, including those that are not accounted for by the IMWG and R-ISS indices such as age, comorbidities, and performance status.

| CONCLUSION
Using data from the Czech Myeloma Group RMG, it was shown that the IMWG and R-ISS risk stratification indices are applicable to patients with MM in routine clinical practice. In addition to supporting previous validation studies conducted primarily in patients participating in experimental clinical trials, the current analysis provides important information about baseline factors that may influence OS in patients with MM, including those that do not form part of the IMWG and R-ISS indices. There is merit in conducting an analysis of outcomes in larger populations of patients who did not receive new drugs and who did receive ASCT. Nonetheless, the current analysis, along with additional analyses by other researchers, confirms the validity of risk stratification using the IMWG and R-ISS indices in a broad range of patients with MM. Beyond simple prognostication, defining MM subgroups, as per the IMWG and R-ISS, will prove useful in providing suitable patient counseling, delivering more effective personalized therapies, and facilitating better between-trial comparisons of patient groups. From the current data, R-ISS seems to show better stratification in a real-world setting especially among patients not treated with ASCT.