The prognostic utility of dynamic risk stratification at disease progression in patients with multiple myeloma

ABSTRACT Objectives: There may be a shift in risk stratification at progression compared to that at diagnosis in patients with multiple myeloma (MM). We aimed to evaluate whether re-staging and stage migration is of prognostic impact. Methods: Real-world data from the National Longitudinal Cohort of Hematologic Diseases-multiple myeloma were collected; 263 consecutive patients demonstrating disease progression were finally included. Staging at diagnosis and re-staging at progression were performed using the International Staging System (ISS) and Revised International Staging System (RISS). Results: Based on ISS re-staging, the median post-progression survival (mPPS) of patients with stage I, II, and III was 44.2, 21.7, and 11.6 months, respectively (P < 0.0001). Based on RISS re-staging, the mPPS of patients with stage I, II, and III was 50.3, 22.2, and 11.4 months, respectively (P < 0.0001). The mPPS in patients with improved, maintained, and deteriorated ISS stage migration from diagnosis was 33.6, 20.9, and 16 months, respectively (P = 0.0051) and that with improved, maintained, and deteriorated RISS stage migration was 48.4, 23.1, and 13.9 months, respectively (P < 0.001). Compared to patients with maintained or improved disease stage, those with deteriorated ISS/RISS migration showed significantly higher incidence of Del(17P) at progression and worse PPS. Multivariate analyses indicated both re-staging and stage migration by ISS/RISS at progression were independent predictors for PPS. Conclusions: We demonstrated that ISS/RISS re-staging showed superior prognostic utility over ISS/RISS staging in predicting PPS. Patients with deteriorated stage migration or maintained advanced stage at progression may need more individualized treatment.


Introduction
Despite significant improvement in the prognosis of patients with newly diagnosed multiple myeloma (NDMM), the recurrence rate after remission remains high [1,2].Previous studies on relapse kinetics have shown that early relapse in patients with multiple myeloma(MM) often indicates poor prognosis [3][4][5][6][7].Moreover, MM progression is a risk-accumulating process [8][9][10][11][12][13][14][15][16][17], with possible differences in the risk stratification at progression compared to that at diagnosis.Thus, in addition to focusing on recurrence kinetics in patients with relapsed/progressive disease, it is also critical to perform dynamic risk re-stratification by re-staging the disease before second-line therapy initiation [18,19].However, a uniform prognostic staging system for assessing post-progression survival (PPS) in patients with relapsed MM has not been established.
Given that clonal evolution and changes in aggressiveness may occur with MM progression due to treatment-induced selection pressure, the risk stratification [20][21][22][23][24][25] is likely to alter from that at initial diagnosis, altering the prognostic trajectory.Therefore, a restaging strategy at progression may help to indicate the prognostic significance of dynamic risk stratification; it can also help to identify unified prognostic stratification systems for patients with MM in relapsed/ progression population.Previous studies [26][27][28] have used International Staging System (ISS) or Revised International Staging System (RISS) to stratify patients with MM at relapse in clinical trials [29].However, given that the patients in the above study were not part of the same cohort that experienced relapse after receiving first-line treatment, it is still unknown whether stage migration (change of the patient's clinical stage at progression compared with the initial stage) will affect post-progression survival (PPS).
In a consecutively treated cohort of NDMM showing disease progression, we performed this real-world retrospective study to evaluate whether re-staging or stage migration using ISS/RISS is of prognostic significance.

Patients and study design
Real-world data were collected through the National Longitudinal Cohort of Hematologic Diseases-multiple myeloma (NICHE-MM) (NCT04645199), which contains historical, anonymised electronic and paper medical records, and follow-up data.
After reviewing data of 912 patients with NDMM diagnosed between March 24, 2013, and December 2, 2019, at our hospital, 265 patients who were referred to a local hospital or discontinued induction treatment were excluded.Of the remaining 647 consecutive patients, 384 were further excluded owing to continuous remission.A total of 263 consecutive patients showing disease progression were finally included in the study.Briefly, patients received induction therapy mainly with proteasome inhibitor (PI)-based (such as bortezomib plus cyclophosphamide and dexamethasone) or immunomodulatory drug (IMiD)-based(such as lenalidomide, cyclophosphamide, and dexamethasone) triplet regimen.After at least four cycles of induction treatment for partial remission (PR), patients underwent either autologous stem cell transplantation (ASCT) or consolidation therapy.Subsequently, patients were treated with lenalidomide or bortezomib maintenance for 1 year, unless intolerant or continual progression was observed.Routine monitoring of treatment response was carried out every two courses during the induction treatment, as well as before and 3 months after transplantation, or post consolidation treatment, and approximately every 3 months during maintenance, and when patients started exhibiting signs of recurrence, follow-up monitoring was performed approximately every month.Renal insufficiency was defined as creatinine > 2 mg/dL.Response assessment was conducted as per the International Myeloma Working Group (IWMG) Criteria [30].
The NICHE-MM study was reviewed and approved by the Human Genetic Resource Administration of China and this study was approved by the institutional ethics committees.

Re-staging and stage migration
We employed the ISS [22] and RISS [25] for staging at diagnosis and re-staging at progression.Patients who had ISS stage II disease but lacked iFISH and LDH analysis at diagnosis or progression, were also classified to be at RISS stage II according to the RISS definition.Re-staging by ISS and RISS was uniformly done according to clinical information at first indication of disease progression to ensure that all patients were re-stratified at relatively consistent time points.
Patterns of stage migration from diagnosis were defined as follows: improved ISS/RISS stage migrationpatients who migrated to lower ISS/RISS stage at progression; maintained ISS/RISS stagepatients who maintained their original ISS/RISS stage at progression; deteriorated ISS/RISS stage migrationpatients who migrated to higher ISS/RISS stage at progression.

Statistical analyses
The chi-square or Fisher exact test was used to compare categorical variables.The t-test was used to compare continuous variables.Correlation was assessed using Spearman's test and two-tailed Pvalues were represented.Multivariable-adjusted Cox proportional hazards survival models were developed to assess the prognostic utility of staging at diagnosis, first-line treatment options, and re-staging at progression on PPS based on the univariable analysis including ISS staging, upfront ASCT, progression-free survival (PFS), and ISS re-staging in model 1 and RISS staging, upfront ASCT, PFS, and RISS re-staging in model 2. Additionally, multivariate analyses were also performed to evaluate the impact of stage migration on PPS: variables include ISS stage migration, PFS, and ASCT in model 1 and RISS stage migration with PFS and ASCT in model 2. Variables with a P-value < 0.1 in the univariable analysis were included in multivariable models and all results were reported as hazard ratios with 95% confidence intervals (CIs).PFS was calculated as the time from diagnosis to any form of disease progression or death; PPS was defined as the time from disease progression to death or last follow-up [31][32][33].Overall survival (OS) was calculated as the time from diagnosis to death or last follow-up.Survival analyses were performed using the Kaplan-Meier method, followed by intergroup comparisons using the log-rank test.The median follow-up duration was analysed using the reverse Kaplan-Meier method.Statistical significance was set at a two-sided P-value < 0.05.Statistical analyses were performed using Statistical Package for Social Sciences (SPSS version 25.0, IBM, Chicago, IL) and R language (version R 4.1.2Foundation for Statistical Computing, Tsinghua, China).

Characteristics at diagnosis and progression
Among the 263 included patients, the median age at diagnosis was 57 years, and 161 (61.2%) patients were male; 213 (80.9%) patients received PI-based induction treatment and 129 (49%) were alive.
Comparison of the characteristics of all included patients at diagnosis was done based on whether or not they had received first-line ASCT.There was no difference in sex distribution, types of M-protein, and ISS/RISS distribution, while the ASCT group patients were younger, had a longer duration of PFS, and a higher rate of PI-based induction regimen (Supporting Information, Table S1).Baseline characteristics at progression and second-line treatment options are shown in Table 1.The rates of renal insufficiency were lower at progression (6.4%) compared to those at diagnosis (11%), (P = 0.085).Regarding the HRCA at progression, the rates of t(4;14) and t(14;16) were comparable to those at diagnosis.However, the rate of Del(17P) at progression was 26.9%, which was significantly higher than 12.7% at diagnosis (P = 0.001).When we grouped the HRCA of relapsed patients according to different stage migration patterns, we further found that the incidence of t(4;14) and t (14;16) were also comparable among different migration patterns; however, compared to patients at maintained or improved original stage, the incidence of Del(17P) was significantly higher in the deteriorated stage migration for both ISS and RISS (P < 0.001, Table 1).Besides, there were no obvious difference in length of PFS and incidence of renal insufficiency between the three patterns of stage migration at progression for both ISS and RISS.Second-line treatment mainly consists of the following regimens: PI-based, IMiD-based, and VRD (bortezomib plus lenalidomide and dexamethasone)-or IRD (ixazomib plus lenalidomide and dexamethasone)-based; others include high-dose chemotherapy followed by second ASCT, CD38 monoclonal antibody, and chemotherapy.Under the three patterns of stage migration, the second-line treatment regimens among the different groups of relapsed patients were roughly comparable.

OS and PFS according to staging at diagnosis
At diagnosis, ISS data were available for all included patients.There were 51 (19.4%), 94 (35.7%), and 118 (44.9%) patients at ISS stage I, II, and III, respectively.
Regarding the RISS stage at disease progression, 52 patients migrated to a lower stage (by one stage, 50 patients; by two stages, 2 patients), 107 patients maintained the original stage, and 25 patients migrated to a higher stage (by one stage,  ), and 13.9 (95% CI: 6.2-25.9)months, respectively (P < 0.001; Figure 2B).Moreover, as the stage migration can be further separated into different circumstances (for example, both ISS/RISS stage 3 migration to stage 1 or stage 2 were classified as improved stage migration), we conducted a more thorough survival analysis and PPS according to each category of ISS/RISS stage migration, and the results are shown in Figure S3.

Prognostic re-staging univariate and multivariate models on PPS
In the univariate analysis for PPS, higher ISS/RISS stage at diagnosis and progression predicted worse outcomes (Table 2).Besides, patients without ASCT therapy or those who showed rapid progression of PFS <20 months also related to worse PPS.Subsequently, to assess the relationships among staging at diagnosis, prior ASCT, and re-staging at progression, we formed a separate multivariable-adjusted Cox proportional hazards model (Table 2) using each ISS and RISS data.Regarding PPS adjusted by PFS, re-staging  using ISS/RISS re-staging at progression showed superior prognostic utility over ISS/RISS staging at diagnosis.Moreover, re-staging by ISS and RISS at progression predicted PPS regardless of whether the patient had previously received first-line ASCT or had experienced early progression.

Prognostic stage migration univariate and multivariate models on PPS
Spearman's analysis between re-staging at progression and stage migration showed that ISS re-staging is moderately correlated with ISS stage migration (r = 0.588, P = 0.01), while RISS re-staging is moderately correlated with RISS stage migration (r = 0.560, P = 0.01).Thus, dependent Cox models were prepared including ISS stage migration and RISS stage migration separately to evaluate the significance of stage migration in PPS prediction (Table 3).Univariate analysis demonstrated that both non-first-line ASCT, early progression, and deteriorated ISS/RISS stage migration predicted worse PPS.In multivariate analysis on PPS adjusted by PFS, maintained and deteriorated ISS/ RISS migration showed worse PPS compared to improved ISS/RISS stage migration during progression, regardless of whether the patient was treated with first-line ASCT or had experienced early progression.

Discussion
In this retrospective real-world study, we performed staging at diagnosis and re-staging at progression using both ISS and RISS in our consecutive, treated MM cohort, to investigate whether re-staging and stage migration from diagnosis can predict PPS in relapsed MM patients.
Initially, we employed survival analysis and then demonstrated that re-staging using ISS/RISS had the ability to discriminate the three groups of patients having significantly different PPS.More specifically, our study showed the mPPS of patients with ISS/RISS staging I, II and III at progression is approximately 4, 2 years, and 1 year respectively; the prognostic discrimination ability according to RISS re-staging at relapse/progression is consistent with a previously published report [29].Furthermore, as expected, restaging at progression not only showed superior prognostic utility over stratification at diagnosis for both models, but was also an independent prognostic factor for PPS, regardless of whether patients had previously received ASCT or had early progression.
Migration to a higher ISS/RISS stage at progression was associated with poor PPS, which could be attributed to aggressive progression with higher disease burden and genetically high-risk abnormalities [3,8], as we found significantly higher incidence of Del  (17P) in deteriorated ISS/RISS migration group compared to stage improved or maintained group.In contrast, migration to a lower stage at progression was associated with significant PPS, which could be attributed to the indolent biological behaviour of recurrent plasma cells with monoclonal gammopathy of undetermined significance-like features [34,35].To further explore the impact of stage migration on PPS, we performed the correlation analysis between re-staging at progression and staging migration at first diagnosis.Given that a moderately moderate intensity of correlation was observed between them in the analysis (patients with advanced stage at progression are more likely to be classified as the deteriorated or maintained stratification group, while patients with lower stage are more likely to be classified as the improved or maintained stratification group), dependent Cox hazard ratio survival models were established with stage migration, first-line ASCT, and PFS.Not surprisingly, both stage migration by ISS and RISS at progression are independent prognostic impactors in PPS, regardless of whether patients had previously received ASCT or had early progression.To the best of our knowledge, this is the first study to demonstrate the prognostic utility of stage migration for MM.This study has several limitations.First, this was a retrospective single-centre study involving heterogeneous treatment regimens.Second, our findings are only applicable to patients with relatively complete staging system parameters at relapse/progression but not to those with sustained remission or missing staging system parameters.Moreover, stage deviation can occur during re-staging, as the stage criteria based on biochemical indicators can be affected by medical intervention.Third, 57.1% (105/184) of the patients were restaged to the RISS II at progression and such a large group may stand a good chance of having different cytogenetic architecture and anticipated PPS; thus, applying R2-ISS re-staging may help overcome this limitation.Fourth, compared to the previously established risk stratification algorithm [18] for patients with relapsed myeloma that takes both patient frailty and disease aggressiveness into consideration, the ISS and RISS does not consider patient frailty (age and Eastern Cooperative Oncology Group performance status).Fifth, owing to the study design, stage migration could only occur in two ways in two of the three cases (for patients with stage I, it could only migrate to a higher stage or be maintained at the original stage, and for patients with stage III, it could only migrate to a lower stage or be maintained at the same stage).However, when the analysis was restricted to patients with ISS II/ RISS II at diagnosis, stage migration can also discriminate relapsed patients into three groups with different PPS.
Although ISS/RISS re-staging is not perfect, our findings demonstrated the prognostic value of simple re-staging methods at progression using the ISS or RISS.This is particularly important for patients maintained at ISS/RISS III disease or those that migrated to advanced ISS/RISS stage at progression as it suggests that prior therapy was not sufficient to overcome the poor prognostic factors carried by patients; thus, post-progression treatment options with new drug regimens involving different mechanisms of action are needed.Previous studies showed that these patients may further benefit from consolidation therapy of high-dose chemotherapy followed by salvage ASCT, CD38-targeting monoclonal antibody therapy, or anti-BCMA CAR-T immunotherapy after progression [36][37][38][39][40].
Owing to the lack of a direct comparison between the prognostic assessment values of ISS/RISS/R2-ISS and previously established models [18,41] in unified research, and as miscellaneous parameters or calculation methodologies cannot be easily applied in clinical implementation, further explorations on more practical score-weighted re-staging models like the R2-ISS, or a model combining patient frailty with cytogenetics, progression kinetics [42,43], and minimal residual disease [44][45][46] may better predict PPS and guide clinical second-line treatment strategies in relapse setting.We expect that this model can be established using a large-scale, multicentre, prospective controlled trial in the future.

Conclusion
Our study shows that risk stratification of patients with MM at relapse/progression may change from that at initial diagnosis altering the patient's original prognostic trajectory.Thus, it is critical to perform dynamic risk stratification and develop an individualized treatment strategy at the time of relapse/progression in patients with MM.

Figure 1 .
Figure 1.PPS according to re-staging at progression.(A) PPS according to ISS re-staging at progression.(B) PPS according to RISS re-staging at progression.Abbreviations: ISS: International Staging System; PPS, post-progression survival; RISS: Revised International Staging System.

Figure 2 .
Figure 2. PPS according to stage migration at progression.(A) PPS according to ISS stage migration at progression, (B) PPS according to RISS stage migration at progression.ISS: International Staging System; PPS: post-progression survival; RISS: Revised International Staging System.

Table 1 .
Characteristics of patients at diagnosis and progression based on different patterns of stage migration.

Table 2 .
Prognostic re-staging univariate and multivariate models on post progression survival

Table 3 .
Stage migration univariate and multivariate models on post progression survival.