A simple additive staging system for newly diagnosed multiple myeloma

Risk stratification in multiple myeloma is important for prognostication, patient selection for clinical trials, and comparison of treatment approaches. We developed and validated a staging system that incorporates additional FISH abnormalities not included in the R-ISS and reflects the additive effects of co-occurring high-risk disease features. We first evaluated the prognostic value of predefined cytogenetic and laboratory abnormalities in 2556 Mayo Clinic patients diagnosed between February 2004 and June 2019. We then used data from 1327 patients to develop a risk stratification model and validated this in 502 patients enrolled in the MMRF CoMMpass study. On multivariate analysis, high-risk IgH translocations [risk ratio (RR): 1.7], 1q gain/amplification (RR: 1.4), chromosome17 abnormalities (RR: 1.6), ISS III (RR: 1.7), and elevated LDH (RR: 1.3) were independently associated with decreased overall survival (OS). Among 1327 evaluable patients, OS was 11.0 (95% CI: 9.2–12.6), 7.0 (95% CI: 6.3–9.2), and 4.5 (95% CI: 3.7–5.2) years in patients with 0 (stage I), 1 (stage II), and ≥2 (stage III) high-risk factors, respectively. In the MMRF cohort, median OS was 7.8 (95% CI: NR-NR), 6.0 (95% CI: 5.7-NR), and 4.3 (95% CI: 2.7-NR) years in the 3 groups, respectively (P < 0.001). This 5-factor, 3-tier system is easy to implement in practice and improves upon the current R-ISS.


INTRODUCTION
Multiple myeloma (MM) is the second most common hematologic malignancy and is responsible for approximately 2% of all cancer deaths in the United States [1]. With the advent of novel therapeutic agents and drug combinations, survival outcomes have improved considerably [2]. Despite these therapeutic advances, survival outcomes remain highly variable even in uniformly treated clinical trial populations [3,4]. This disparity highlights the importance of risk stratification at the time of diagnosis for prognostication and patient selection for clinical trials [5,6]. Several risk stratification systems have been proposed based on clinical characteristics, laboratory studies, bone marrow cytogenetics, and gene expression profiling. Among them, the International Staging System (ISS) and its successor, the revised ISS (R-ISS), have stood the test of time in clinical practice. The ISS was first introduced in 2005 and is based on β2-microglobulin and albumin, which are thought to reflect tumor burden and host status [7]. In clinical practice, the ISS remains a popular choice for risk stratification due to its simplicity, although its ability to discriminate among lower-risk patients is limited in the era of novel therapies [8]. Interphase fluorescence in-situ hybridization (FISH) enabled the identification of high-risk disease independent of ISS stage based on abnormalities such as t(4;14), t (14;16), and del(17p) [9,10]. The revised ISS (R-ISS) was introduced in 2015 and includes these cytogenetic abnormalities as well as elevated lactate dehydrogenase (LDH) as high-risk markers [11]. The R-ISS has subsequently been validated in several studies [12,13] and is currently used to risk-stratify patients with newly diagnosed MM [11]. It is thought to perform better than its predecessor in certain patient populations [8,12,13]. Recent studies have identified additional cytogenetic abnormalities not included in the R-ISS to be associated with adverse survival outcomes including monosomy 13 [14,15], gain/amplification of 1q21 [16,17], and rearrangements involving the MYC gene [18][19][20]. Even though more powerful high-risk classifiers have been developed based on gene expression profiles and cytogenetics, their acceptance in clinical practice remains poor due to the involved resources and need for complex modeling [21][22][23][24][25]. While the presence of high-risk laboratory and cytogenetic features has been incorporated into the R-ISS, the additive effects of multiple cooccurring high-risk disease features have not been accounted for [15,16]. Importantly, as RISS was developed, assessment of 1q abnormalities was not commonly employed and there was limited data to explore the value of adding this variable. Furthermore, unlike the ISS which had a similar proportion of patients in each of the three stages, R-ISS resulted in majority of patients being classified as intermediate. In light of the identification of additional cytogenetic risk factors and an improved understanding of the prognostic implications of multi-hit disease, we conducted this study to evaluate a new simple additive staging system in patients with newly diagnosed MM [14,17,19,20].

PATIENTS AND METHODS Patients and study design
We included all patients with MM seen in Mayo Clinic in Rochester, Minnesota between 2004 and 2019, who had cytogenetic analysis by FISH performed within 1 year before diagnosis, or within 6 months from the start of first-line treatment. The cohort included 2556 patients 18 years and older, diagnosed with MM between February 2004 and June 2019. Patients were identified from a preexisting database and additional data was obtained by review of electronic medical records. We extracted data on FISH results, standard serum and urine tests, bone marrow biopsy results, first-line treatment, transplant status, and date of death or last follow up. For model development, we included 1327 Mayo Clinic patients who had simultaneous data for all high-risk abnormalities found to be prognostic on multivariate analysis. The validation population included 502 patients enrolled in the MMRF CoMMpass study (MMRF); all data were obtained through the MMRF Researcher Gateway (https://research.themmrf.org). The study was approved by the Mayo Clinic Institutional Review Board. Informed consent was obtained from all patients included in the study.

Statistical analysis
Using the entire Mayo cohort (n = 2556), we first examined the impact of each of the cytogenetic abnormalities detected by FISH on overall survival (OS) using univariate analysis. These abnormalities included: (1) high-risk translocations involving the immunoglobulin heavy chain (IgH) locus, (2) t (11;14) translocation, (3) trisomies of at least 1 chromosome, (4) rearrangements involving the MYC gene locus, (5) gain of 1 or more copies of 1q (1q gain/amplification), (6) monosomy of chromosome 13 (not including patients with 13q deletion alone), and (7) chromosome 17 abnormalities [del(17p) or monosomy of chromosome 17]. Cytogenetic abnormalities significantly associated with survival in the univariate model were included in a multivariate model; those significantly associated with survival in the multivariate model were then included in a final multivariate model with ISS stage III (vs I/II) and elevated LDH. Values above the upper limit of normal of the reporting laboratory were considered elevated for lactate dehydrogenase (above 222 U/l for the Mayo Clinic Laboratories). ISS stages were defined as: β2-microglobulin <3.5 mg/dL and albumin ≥3.5 g/dL (stage I); β2microglobulin ≥5.5 regardless of albumin (stage III); all other cases were considered stage II [7]. For R-ISS, stage I included patients with ISS stage I, normal LDH, and absence of high-risk cytogenetics; R-ISS stage III included patients with ISS stage III, elevated LDH and/or high-risk cytogenetics; all other cases were considered stage II [11]. Overall survival (OS) was defined as the time from diagnosis of multiple myeloma to death from any cause. Progression-free survival (PFS) was defined as the time from the start of first-line treatment to first disease progression or death from any cause (whichever occurred first). Patients without an event at the end of follow-up were censored. Risk ratios (RR) were calculated based on the univariate and multivariate cox proportional hazards models.

Model development and validation-The Mayo Additive Staging System (MASS)
First, we updated the R-ISS such that the definition of HR cytogenetic abnormalities included all the FISH abnormalities found to be significantly associated with survival in our final multivariate model. We then compared the PFS and OS of patients based on the updated R-ISS. Next, given that the R-ISS does not give similar weight to all the prognostic factors, we explored a simpler approach that incorporates all the independent prognostic factors, and grouped patients by virtue of the number of risk factors. For this analysis, we included 1327 Mayo Clinic patients who had simultaneous data available for all the variables significantly associated with survival in the final multivariate model. We examined PFS and OS of patients based on the number of high-risk abnormalities; patients with 0, 1, or ≥2 HR abnormalities were considered as stage I, II, or III, respectively. We then performed a subgroup analysis for OS based on age group (<65 years vs. ≥65 years), transplant status, and across 2 time periods (before and after 2012). Next, we evaluated the performance of this model in an independent population of 502 patients enrolled in the MMRF CoMMpass study (MMRF). The performance of the model was measured by Harrell's Concordance Index (C) [29,30]. Overall and progression-free survival estimates were calculated using the Kaplan-Meier method, and survival was compared between groups using the Log-rank test [31]. For all tests, 2-sided P values <0.05 were considered statistically significant. Statistical analyses were performed using the JMP (SAS, Cary, NC) and Stata statistical softwares.

Baseline patient characteristics
The Mayo clinic cohort included 2556 patients diagnosed with MM between February 2004 and June 2019. The median age was 64 years and 62% were males. The clinical characteristics of patients and the first-line treatments are included in Table 1. Among all patients, 143 (6%) had no abnormality detected by FISH using the probes tested. A trisomy of at least one chromosome was found in 55% of patients tested. t (11;14) was the most common IgH translocation, found in 21%; t(4;14) and t(14;16) were found in 10% and 4%, respectively. A gain of 1 or more copies of chromosome 1q was detected in 31%, and a rearrangement involving the MYC gene locus was found in 9%. A monosomy of chromosome 13 was seen in 37%, and an abnormality in chromosome 17 [del(17p)/monosomy 17] was seen in 13%. These results are presented in Table 2.

Univariate and multivariate survival analysis
The median follow-up in the entire cohort was 6.2 (95% CI: 5.9-6.5) years. At the time of analysis, 58% of patients were alive; the median OS was 7. including all FISH abnormalities significantly associated with OS on univariate analysis, monosomy 13 (RR: 1.1, P = 0.20) and trisomies (RR: 0.9, P = 0.08) were no longer prognostic for OS. ISS III and elevated LDH were both associated with increased risk of death on univariate analysis, so these were included in a final multivariate model with high-risk IgH translocations, MYC rearrangements, 1q gain/amplification, and chromosome 17 abnormalities. HR IgH translocations (RR: 1.7, P < 0.001), 1q gain/amplification (RR: 1.4, P < 0.001), chromosome 17 abnormalities (RR: 1.6, P < 0.001), ISS III (RR: 1.7, P < 0.001), and elevated LDH (RR: 1.3, P = 0.01) were all independently associated with decreased OS in the final multivariate model. MYC rearrangements were associated with decreased survival, but this was not statistically significant (RR: 1.3, P = 0.06). These results are presented in Table 3.

Risk stratification based on the MASS over 2 time periods
Stage migration using the MASS Among all patients, 1269 patients had simultaneous data available for R-ISS stage and MASS, including 244 (18%) patients with R-ISS I, 791 (62%) with R-ISS II and 234 (18%) with R-ISS III. Overall, 469 (37%) had stage migration when the MASS was used for risk stratification (Fig. 5). Among R-ISS I patients, 21% were reclassified as stage II using the MASS system; 32% and 21% of patients with  R-ISS II were classified as stages I and III, respectively using the MASS system. All patients in R-ISS III were classified as stage III using the MASS system. Among 147 patients with double-hit myeloma, 71 (48%) were classified as stage II using the R-ISS and 76 (52%) were classified as stage III. Among 17 patients with triple hit myeloma, 4 (24%) were classified as R-ISS II and 13 (76%) as R-ISS III. By definition, all patients with double and triple hit MM were classified as stage III using the MASS system. Stage migration from ISS to MASS is also shown in Fig. 5. Among patients with R-ISS I, patients who were classified as MASS II had inferior OS (median OS: 6.8, 95% CI: 6.3-9.4 years) compared to patients classified as MASS I (median OS: 9.4, 95% CI: 8.9-12.8 years), P = 0.03 (supplemental Fig. 2a). Among patients with R-ISS II, median OS was 11.0 (95% CI: 8.9-NR), 7.0 (95% CI: 5.9-9.3), and 4.8 (95% CI: 3.5-5.8) years in patients with MASS I, II, and III, respectively (P < 0.001) (supplemental Fig. 2b). Among patients with ISS I, those who were classified as MASS II (median OS: 8.7, 95% CI: 6.3-9.8 years) had decreased OS compared to patients with MASS I (median OS: 9.4, 95% CI: 8.9-12.8 years) (P = 0.01), but no significant difference in OS compared to patients with MASS III (median OS: 7.7, 95% CI: 3.3-NR years) (P = 0.39) (Supplemental Fig. 2c). Among patients with ISS II, a higher MASS stage was associated with inferior survival; median OS was 11.0 (95% CI: 8.9-NR), 7.5 (95% CI: 5.7-9.5), and 3.9 (95% CI: 3.2-5.7) years in patients with MASS I, II, and III, respectively (P < 0.001) (supplemental Fig. 2d).
The MASS as a 4-tier staging system When the MASS was used as a 4-tier staging system in the Mayo cohort, OS was 11.0 (95% CI: 9.2-12.6), 7.0 (95% CI: 6.3-9.   but there was no difference in OS between all other pairs. Among patients with two HR cytogenetic abnormalities (235 patients), there was no difference in OS between patients with one primary and one secondary HR abnormality, (median OS: 4.5, 95% CI: 3.5-6.4 years) and those who had two secondary HR cytogenetic abnormalities (5.7, 95% CI: 2.9-8.1 years) (P = 0.77). There was also no difference in OS associated with different combinations of HR cytogenetic abnormalities (global P value: 0.66).

DISCUSSION
Risk stratification in MM remains an important concept both for patient counseling and the development of risk-adapted treatment strategies [5]. While minimal residual disease assessments are emerging as powerful predictors of survival outcomes for patients undergoing treatment, risk stratification in newly diagnosed patients has to rely on pre-treatment patient and disease characteristics [32]. The ISS and its successor, the R-ISS,  have prevailed in clinical practice and research alike due to their simplicity and reliance on readily available baseline characteristics [7,11]. The recognition of additional risk factors motivated the revision of the ISS and resulted in a more powerful predictive model [11]. Since the introduction of the R-ISS, data on additional risk factors including chromosome 13 abnormalities, 1q gains, and MYC rearrangements have come to light [15][16][17]20]. Furthermore, it has become evident that the co-occurrence of multiple high-risk disease features compounds the risk for adverse outcomes, leading to the concept of double-and triple-hit myeloma [15,16]. This insight led us to propose a refined additive risk stratification system.
We first evaluated the prognostic impact of known cytogenetic abnormalities in MM. In addition to the high-risk abnormalities included in the R-ISS, 1q gain demonstrated independent prognostic value in our final multivariate model, while the prognostic impact of MYC rearrangements did not reach statistical significance. Trisomies and monosomy 13 were not independently associated with OS when adjusting for other cytogenetic abnormalities, and thus the poor prognostic impact of nonhyperdiploidy [33] and chromosome 13 abnormalities [34][35][36] may be due to their association with other HR cytogenetic abnormalities. When we used an updated version of the R-ISS, where 1q gain/amplification was included in the definition of high-risk disease, we discriminated 3 groups of patients with significantly different outcomes. However, as with the original R-ISS, most patients had intermediate risk disease (stage II) using this classification (66%). In addition, the R-ISS as originally designed did not consider the compounding effects of co-occurring highrisk disease features. Thus, we proposed a system that stratifies patients based on the number of high-risk factors present at diagnosis: HR IgH translocations, 1q gain/amplification, chromosome 17 abnormalities, ISS III, and/or elevated LDH. Using this system, we discriminated 3 groups of patients with significantly different progression-free and overall survival with nearly a third of the patients distributed between the 3 stages.
The prognostic utility of this system was demonstrated in both age groups: ≥65 years and <65 years, in transplant-eligible and ineligible patients, and over 2 time periods. In addition, its prognostic value was shown in 2 independent populations. Compared to the R-ISS, the 3-tier MASS re-classified approximately one third of patients, underscoring how much our understanding of high-risk disease has been reshaped by the concept of multi-hit disease. Furthermore, the MASS retained its performance and discriminatory ability when used as a 4-tier risk stratification system which can be utilized if additional discrimination among high-risk patients is desired. This application has the potential to serve as an important tool in the design of clinical trials exploring intensification of treatments in high-risk patients. Importantly, this model will lend itself for incorporation of other risk factors as they are identified in the future.
Some limitations of this study include the long period of time over which data were collected during which treatments and transplant-eligibility criteria changed, and the exclusion of nonevaluable patients which may create selection bias. In addition, we did not evaluate some cytogenetic abnormalities for which data were not available like deletion 1p32 which has an estimated prevalence of 7% in newly diagnosed patients and has been associated with worse outcomes [37]. Furthermore, the lack of prognostic value of MYC abnormality in this study may be due to the limited number of patients with available cytogenetic data for MYC, and thus its added value in risk stratification requires further evaluation in larger studies. The impact of the clonal plasma cell percentage harboring secondary cytogenetic abnormalities and the genetic mutational profile on risk stratification were not assessed in this study and also warrant exploration in future studies.

CONCLUSION
In summary, we developed and validated a simple, additive 5-factor 3-tier risk model for newly diagnosed MM in two diverse patient populations that is easy to implement in clinical practice and can play an important role in patient selection for clinical trials.