Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Apr 1.
Published in final edited form as: Am J Hematol. 2024 Jan 21;99(4):523–533. doi: 10.1002/ajh.27207

Development and validation of an individualized and weighted myeloma prognostic score system (MPSS) in patients with newly diagnosed multiple myeloma

Xuehan Mao 1,2,3,#, Wenqiang Yan 1,2,#, David Mery 4,#, Jiahui Liu 1,2, Huishou Fan 1,2, Jingyu Xu 1,2, Yan Xu 1,2, Weiwei Sui 1,2, Shuhui Deng 1,2, Dehui Zou 1,2, Chenxing Du 1,2, Shuhua Yi 1,2, Frits van Rhee 4, Bart Barlogie 4, John D Shaughnessy Jr 4, Kenneth C Anderson 5, Fenghuang Zhan 4,*, Lugui Qiu 1,2,*, Gang An 1,2,*
PMCID: PMC10947864  NIHMSID: NIHMS1958049  PMID: 38247315

Abstract

Current standard predictive models of disease risk do not adequately account for the heterogeneity of survival outcomes in patients with new-diagnosed multiple myeloma (NDMM). In this retrospective, multicohort study, we collected clinical and genetic data from 1792 NDMM patients and identified the prognostic impact of all features. Using the top ranked predictive features, a weighted Myeloma Prognostic Score System (MPSS) risk model was formulated and validated to predict overall survival (OS). In the training cohort, elevated LDH, ISS Stage III, thrombocytopenia, and cumulative high-risk cytogenetic aberration (HRA) number were found to have independent prognostic significance. Each risk factor was defined as its weighted value respectively according to their hazard ratio for OS (thrombocytopenia 2, elevated LDH 1, ISS III 2, one HRA 1, and ⩾2 HRA 2, points). Patients were furtherly stratified into four risk groups: MPSS I (22.5%, 0 points), II (17.6%, 1 points), III (38.6%, 2–3 points), and IV (21.3%, 4–7 points). MPSS risk stratification showed optimal discrimination, as well as calibration, of four risk groups with median OS of 91.0, 69.8, 45.0, and 28.0 months, for patients in MPSS I to IV groups (P<0.001), respectively. Importantly, the MPSS model retained its prognostic value in the internal validation cohort and an independent external validation cohort, and exhibited significant risk distribution compared with conventional prognostic models (R-ISS, R2-ISS, and MASS). Utilization of the MPSS model in clinical practice could improve risk estimation in NDMM patients, thus prompting individualized treatment strategies.

Keywords: multiple myeloma, cumulative cytogenetic aberrations, prognostic factors, risk stratification

Graphical Abstract

graphic file with name nihms-1958049-f0001.jpg

Introduction

Multiple myeloma (MM) is a malignancy of terminally differentiated antibody secreting plasma cells in the bone marrow with highly variable survival outcomes1. Precise risk stratification not only plays a vital role in predicting patient prognoses, but critically can be used to develop risk-adapted treatment regimens. Various prognostic factors and staging systems have been developed to predict patient outcomes2–5. The most common predictive models used in clinical practice are the International Staging System (ISS)2 and the Revised ISS (R-ISS)3. ISS was thought to be the first simple and robust risk stratification in MM. R-ISS integrated ISS with specific high-risk cytogenetic aberrations and serum lactate dehydrogenase level (LDH), which better reflected the biological features of MM.

However, the current stratification systems fail to account for all outcome variability. ISS did not initially incorporate cytogenetic abnormalities and showed restricted utility to guide risk-adapted treatment6,7. In addition, while several cytogenetic abnormalities were included in R-ISS, improvements and updates from recent studies had been made. Thus, in addition to del(17p), t(4:14), and t(14;16), the independent negative impact of 1q21 gain has now been confirmed in multiple studies8,9. Yet, 1q21 gain is not considered in R-ISS. Moreover, there is emerging evidence that compared with one specific high-risk aberration (HRA), the coexistence of two or more HRA confers a poorer prognosis than any one alone10–12.

Recently, two new risk models have been proposed including the Second Revision of ISS (R2-ISS)4 and the Mayo Additive Staging System (MASS)5, integrating 1q21 gain alongside the conventional R-ISS model. Remarkably, the t(14;16) was excluded in R2-ISS and MASS does not consider the inherent imbalance of hazard weight among different predictors. Weighted prognostic systems accounting for essential risk factors have not yet been developed completely and sufficiently in MM.

Host status or bone marrow microenvironment is another key factor impacting patient prognosis, especially in the context of immunotherapies13,14. Mass myeloma cells occupying the bone marrow reshape the marrow microenvironment, promoting immunosuppression15. Current prognostic models of MM do not incorporate such host-associated predictors like monocyte counts16 and thrombocytopenia2,17,18.

In the present study, ISS Stage III, LDH, thrombocytopenia, and number of HRA were integrated into one comprehensive and weighted Myeloma Prognostic Score System (MPSS), capable of stratifying NDMM patients into four risk groups. The MPSS model showed significantly better discrimination and calibration compared with R-ISS, R2-ISS, and MASS. Further analysis also confirmed its utility and efficacy in different subgroups, as well as in an independent validation cohort.

Patients and Methods

Construction of the MPSS model was carried out based on the National Longitudinal Cohort of Hematological Diseases in China (NCT04645199) cohort, collected from 2000 to 2020 at the Institute of Hematology & Blood Disease Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College (IH & BDH, CMAS & PUMC). As shown in Fig 1A, exclusion criteria include: other plasma diseases (pAL, POEMS, PCL), MGUS or SMM with no treatment indicated, and NDMM patients with incomplete clinical information. The study was conducted in accordance with the Declaration of Helsinki and approved by Institutional Review Boards. Informed consents were obtained from patients before recruitment. After exclusion, a total of 1016 NDMM patients were enrolled. It is worth noting that in real-world analysis, patients who have severe cardiopulmonary comorbidities were also included. The whole dataset was then randomly divided into training cohort (n=710) and validation cohort 1 (n=306) by a predefined selective ratio (7:3): training cohort was for the selection of risk factors and construction of the MPSS model; validation cohort 1 was for internal validation of the MPSS model. In the BDH cohort, HRA was defined using Fluorescence in situ hybridization (FISH) which was performed on enriched plasma cells. Abnormal cytogenetic detection included gain of chromosome 1q, deletion of chromosome 13q, deletion of chromosome 17p, and IgH translocations: t(11;14), t(4;14), and t(14;16). The positive threshold for translocation probes was set at 10%, while the copy number probe had a threshold of 20%. Given the comparable effects of different copy numbers of 1q21 on prognosis (Fig S1), we combined 1q21 gain (3 copies) and 1q21 amplification (≥4 copies) as gain(1q).

Figure 1: Study profile and prognosis predictions by MPSS model.

Figure 1:

Open in a new tab

(A) workflow of patient selection of the MPSS project; (B) PFS and OS of patients in training cohort; (C) PFS and OS of patients in validation cohort 1; (D) PFS and OS of patients in validation cohort 2. Training cohort: the random selected 70% of Blood Disease Hospital (BDH) cohort; Internal Validation cohort: the random selected 30% of BDH cohort; External Validation cohort: TT2, TT3a, and TT3b cohorts from a series of randomized prospective studies directed by University of Arkansas for Medical Sciences (UAMS). pAL: primary amyloidosis; PCL: plasma cell leukemia.

External validation cohort 2 was created using TT2, TT3a, and TT3b cohorts from a series of randomized prospective studies from the University of Arkansas for Medical Sciences (UAMS)19–21. Baseline characteristics for cases from the TT2 and TT3 cohorts showed a difference from the real-world cohort in that all cases were ≤ 75 years old; Performance Status score < 3; and only patients with enough cardiopulmonary function that can tolerate high dose anti-myeloma treatment were enrolled in the clinical trials. HRA in this cohort were defined using GEP. Deletion of TP53 (del17p) was defined when the gene expression signal from the Affymetrix U133Plus2.0 for the TP53 gene was less than or equal to 73322, the t(4;14) and t(14;16) translocations were defined by “spiked expression” for FGFR3 and/or NSD2 and MAF and if the case was classified as either a MS or MF subtype23. The gain of 1q was defined using an algorithm that accurately predicts chromosome gains and losses in MM using genes whose expression is copy number dependent24.

Development and validation of the MPSS model

Clinical and laboratory variables previously demonstrated to be associated with patient survival outcomes were assessed in samples before the start of first-line treatment. Factors for PFS and OS identified in univariate analysis are shown in supplementary material. All factors included were checked and satisfied the proportionality assumption.

Significant covariates with p value < 0.05 in univariate analysis were selected and incorporated into the Cox multivariate model. Development of the final model was performed using backward stepwise regression based on the Wald statistic. According to the hazard ratios (HR) derived from Cox analysis, a weighted risk score was then assigned for each factor, and a weighted MPSS score was formulated based on the sum of the points. The decision tree was then adopted for determining the MPSS risk stratification.

Performance of MPSS was then checked by assessing discrimination and calibration. Discrimination was assessed by Kaplan-Meier curves, estimating HR value and Harrell’s concordance index (C-statistic) by bootstrap resampling (1000 resamples). Calibration curves were illustrated to compare MPSS-predicted survival probability with the observed patient survival at 3 and 5 years of follow-up.

Survival and statistical analyses

The primary endpoint of the study was overall survival (OS), as calculated from the initiation of treatment to the date of death or last follow-up. Survival curves were plotted with Kaplan-Meier curves and compared using the log-rank test. P values < 0.05 (two-sided) are considered statistically significant. Statistical analyses were performed using IBM SPSS Statistics, Version 25.0, and R 4.1.2 software (http://www.r-project.org).

Results

1. Clinical characteristics and treatments

From January 2000 to December 2020, 1016 of 1978 patients with complete clinical characteristics and survival data were enrolled, with a median follow-up time of 52.9 (95%CI, 48.7–57.1) months. In the training cohort (n=706), the median age was 58 years old, and 62.3% of patients were male. In the first-line treatment, 56.2% of patients received with PIs+IMiDs-based therapy, 21.1% with PIs, and 15.8% with IMiDs. A total of 27.7% patients received ASCT after induction treatment. In the validation cohort 1 (n=306), the median age was 58 years, and 54.2% of patients were male. In the first-line treatment, 52.3% of patients received PIs+IMiDs-based therapy, 22.5% with PIs, and 21.4% with IMiDs. A total of 24.9% patients received ASCT after induction treatment. The external validation cohort 2 (n=776) consisted of patients from randomized prospective TT2 and TT3a/b cohorts directed by UAMS. The clinical features of training cohort and validation cohorts are listed in Table 1.

Table 1.

Baseline features and treatments of patients in the MPSS project.

Clinical features (%) BDH cohort UAMS cohort
Total (n=1016) Training Cohort (n=710) Validation Cohort 1 (n=306) Validation Cohort 2 (n=776)
Age, median (range) 58 (26–83) 58 (26–83) 58 (30–81) 58 (25–76)
Age ≥ 65 24.6% (250/1016) 23.9% (170/710) 26.1% (80/306) 23.5% (182/776)
Male 59.8% (608/1016) 62.3% (442/710) 54.2% (166/306) 60.8% (472/776)
Anemia 43.0% (437/1016) 42.3% (300/710) 44.8% (137/306) 11.1% (86/776)
Thrombocytopenia 15.6% (159/1016) 16.3% (116/710) 14.1% (43/306) 14.6% (113/776)
Renal failure 14.1% (143/1012) 14.4% (102/707) 13.4% (41/305) 8.2% (64/776)
Elevated LDH 16.4% (167/1016) 17.3% (123/710) 14.4% (44/306) 29.0% (225/776)
High β2MG 43.0% (429/1016) 41.2% (287/697) 47.2% (142/301) 21.6% (168/776)
BMPC ≥60% 45.6% (447/980) 45.0% (308/685) 47.1% (139/295) 43.9% (332/756)
M protein type
 IgA 24.5% (246/1005) 23.9% (168/701) 25.7% (78/304) 22.2% (172/776)
 IgG 50.6% (509/1005) 50.4% (353/701) 51.3% (156/304) 77.3% (600/776)
 Others 24.9% (250/1005) 25.6% (180/701) 23.0% (70/304) 0.5% (4/776)
Light chain type
 κ 47.4% (476/1005) 46.9% (329/701) 48.4% (147/304) 62.2% (483/776)
 λ 50.7% (510/1069) 51.5% (361/701) 49.0% (149/304) 37.8% (293/776)
DS stage
 I 4.9% (49/1007) 5.3% (37/704) 4.0% (12/303) NA
 II 8.6% (87/1007) 9.7% (68/704) 6.3% (19/303) NA
 III 86.5% (871/1007) 85.1% (599/704) 89.8% (272/303) NA
ISS stage
 I 19.3% (196/1016) 20.1% (143/710) 17.3% (53/306) 46.0% (357/776)
 II 36.5% (371/1016) 37.3% (265/710) 34.6% (106/306) 32.3% (251/776)
 III 44.2% (449/1016) 42.5% (302/710) 48.0% (147/306) 21.6% (168/776)
R-ISS stage
 I 14.3% (145/1016) 14.2% (101/710) 14.4% (44/306) 28.1% (218/776)
 II 65.5% (665/1016) 66.2% (470/710) 63.7% (195/306) 59.4% (461/776)
 III 20.3% (206/1016) 19.6% (139/710) 21.9% (67/306) 12.5% (97/776)
Treatment
 PIs based 22.6% (214/1016) 21.1% (150/710) 22.5% (64/306) NA
 IMiDs based 18.3% (173/1016) 15.8% (112/710) 21.4% (61/306) 22.4% (174/776)
 PIs+IMiDs based 59.1% (559/1016) 56.2% (399/710) 52.3% (160/306) 55.9% (434/776)
 Chemotherapy 6.9% (70/1016) 6.9% (49/710) 6.9% (21/306) 21.6% (168/776)
ASCT 26.9% (258/959) 27.7% (187/674) 24.9% (71/285) 94.8% (736/776)
Cytogenetic
 Del(13q) 45.1% (457/1013) 43.7% (309/707) 48.4% (148/306) 42.5% (330/776)
 Del(17p) 10.0% (102/1016) 11.0% (78/710) 7.8% (24/306) 9.9% (77/776)
 Gain(1q) 45.6% (463/1016) 46.1% (327/710) 44.4% (136/306) 44.2% (343/776)
 t(11;14) 16.6% (166/1000) 17.2% (120/698) 15.2% (46/306) 9.3% (72/776)
 t(4;14) 18.4% (187/1016) 18.2% (129/710) 19.0% (58/306) 13.4% (104/776)
 t(14;16) 3.3% (34/1016) 3.7% (26/710) 2.6% (8/306) 6.2% (48/776)
 HRT 21.8% (221/1016) 21.8% (155/710) 21.6% (66/306) 19.6% (152/776)
HRA
 0 HRA 43.2% (439/1016) 43.0% (305/710) 43.8% (134/306) 47.9% (372/776)
 1 HRA 38.6% (392/1016) 37.9% (269/710) 40.2% (123/306) 32.9% (255/776)
 ≥2 HRA 18.2% (185/1016) 19.2% (136/710) 16.0% (49/306) 19.2% (149/776)
Open in a new tab

Notes: BDH cohort: enrolled newly-diagnosed multiple myeloma patients diagnosed in blood disease hospital from 2000 to 2020; UAMS cohort: TT2, TT3a, and TT3b cohorts from a series of randomized prospective studies directed by University of Arkansas for Medical Sciences (UAMS); Training cohort: the random selected 70% of BDH cohort; Validation cohort 1: the random selected 30% of BDH cohort; Anemia: Hb <90g/L; Thrombocytopenia: platelet levels <100*109/L in BDH and <150*109/L in UAMS; Renal failure: serum creatinine >2.0 mg/dl; Elevated LDH: LDH ≥247 U/L in BDH and ≥190U/L in UAMS; High β2MG: β2-MG >5.5mmol/L; BMPC: Bone marrow plasma cell; ASCT: autologous stem cell transplantation (includes 11 cases with allo-stem cell transplantation); HRT: High risk IgH translocations, including t(4;14) and/or t(14;16); HRA: cumulative number of high risk cytogenetic aberrations, including del(17p), gain(1q) gain, and HRT (t(4;14)/t(14;16))

2. Frequencies of cytogenetic aberrations and related survival analyses

Frequencies of the common cytogenetic aberrations detected by FISH were shown in Table 1. Here, we defined del(17p), gain(1q), and t(4;14)/t(14;16) as HRA. In the training cohort, there were 405 HRA events, with 1–3 HRA accounting for 37.9% (269/710), 17.2% (122/710), and 2.0% (14/710), respectively. Similar distribution of the HRA were observed in the two validation cohorts (P>0.05; Table 1, Table S1).

To clarify the specific influence of different cytogenetic aberrations on patient survival, we compared OS of patients with different HRA separately in the training cohort. The median OS of patients with 0, 1, 2, and 3 HRA were 73.6 (95%CI, 61.6–85.5) months, 48.0 (95%CI, 39.3–56.8) months, 29.0 (95%CI, 23.9–34.1) months, and 23.0 (95%CI, 15.6–30.4) months, respectively (P<0.001; Fig S2 A–B). Considering the similar poor survival of patients with 2 or 3 HRA (P=0.236), we grouped these patients into the ≥2 HRA class. Moreover, compared with the impact of any specific HRA, acquisition of two or more HRA was associated with an even higher risk and poor prognosis regardless of the type or combination of co-existing HRA (Fig S2 C–D).

3. Multivariate Cox analysis for survival and selection of independent prognostic factors

The individual prognostic impact of each risk factor was estimated in the training cohort. Univariate and multivariate Cox analysis of predictors for OS are listed in Table S2. According to the results of multivariate analysis, thrombocytopenia (HR=1.63, 95% CI, 1.23–2.17, P<0.001), elevated LDH (HR=1.43, 95% CI, 1.10–1.88, P=0.01), ISS III (HR=1.62, 95% CI, 1.26–2.09, P< 0.001), with 1 HRA (1 HRA vs.0 HRA: HR=1.45, 95% CI, 1.11–1.88, P=0.006) and ≥ 2 HRA (≥ 2 HRA vs. 0 HRA: HR=2.44, 95% CI, 1.82–3.28, P< 0.001) were significant independent risk factors for shortened OS. The final HR values for each factor are provided in Table 2A. Likewise, these predictors also independently conferred an adverse impact on PFS, as shown in Table S3. Similar independent prognostic factors could be observed in the additional analysis of the entire BDH cohort (n=1016; Table S4–6).

Table 2A.

MPSS score definition on the basis of the training cohort

Risks β OS HR (95% CI) p Score assigned
Thrombocytopenia 0.471 1.60 (1.24–2.08) <0.001 2
Elevated LDH 0.347 1.41 (1.09–1.84) 0.01 1
ISS III 0.556 1.74 (1.41–2.16) <0.001 2
0 HRA — Reference — 0
1 HRA 0.399 1.49 (1.17–1.90) 0.001 1
≥2 HRA 0.846 2.33 (1.76–3.10) <0.001 2
Open in a new tab

4. Development of Myeloma Prognostic Score System (MPSS)

Based on the corresponding regression coefficient as well as the HR value associated with each risk factor in the Cox analysis, weighted risk score 1 was assigned to elevated LDH and 1 HRA given their similar HR values; while thrombocytopenia, ISS stage III, and ≥ 2 HRA received a risk score of 2 (Table 2A). By adding risk scores together, the final grading of MPSS risk stratification was then established. With an MPSS score ranging from 0 to 7, patients in the training cohort experienced progressively inferior outcomes as the score increased (Table S7).

The patients with different scores but similar OS were combined (Figure S3), and the whole training cohort was segregated into four risk categories: MPSS I (score 0, 22.5%), MPSS II (score 1, 17.6%), MPSS III (score 2–3, 38.6%), MPSS IV (score 4–7, 21.3%). The risk of death for patients within MPSS II, III, IV were 1.57, 2.51, and 4.51 times greater than MPSS I group patients, respectively (Table 2B). The median PFS was 54.0 (95%CI, 41.3–66.8) versus 39.4 (95%CI, 28.3–50.4) versus 30.0 (95%CI, 24.8–35.2) versus 18.0 (95%CI, 14.2–21.8) months in the MPSS I, II, III, and IV groups, respectively (P<0.001). The median OS was 91.0 (95%CI, 70.8–111.2) versus 69.8 (95%CI, 61.0–78.5) versus 45.0 (95%CI, 36.8–53.2) versus 28.0 (95%CI, 24.1–32.0) months in the MPSS I, versus II, III, and IV groups, respectively (P<0.001; Fig 1B). The differences among different MPSS groups were statistically significant.

Table 2B.

MPSS risk stratification in the training cohort

MPSS group Proportion HR (95% CI) p MPSS score
I 22.5% (160/710) Reference 0
II 17.6% (125/710) 1.57 (1.07–2.31) 0.021 1
III 38.6% (274/710) 2.51 (1.84–3.44) <0.001 2–3
IV 21.3% (151/710) 4.51 (3.24–6.27) <0.001 4–7
Open in a new tab

Notes: Thrombocytopenia: platelet levels <100*109/L; Elevated LDH: LDH ≥247 U/L; ISS: international staging system; HRA: cumulative number of high risk cytogenetic aberrations, including del(17p), 1q21 gain, and HRT (t(4;14)/t(14;16)); p was for HR of different MPSS group compared with MPSS I group.

Subgroup analyses of the utility of MPSS model was next evaluated. In subgroups with different ages and diverse therapeutic eras, MPSS stratification successfully differentiated patient prognosis, confirming both efficacy and reproducibility (P<0.001; Fig S4–5).

5. Internal and external Validation of the MPSS model

In the internal validation cohort 1, the MPSS I patients were 63 (20.6%), MPSS II 50 (16.3%), MPSS III 132 (43.1%), and MPSS IV 61 (19.9%). The median OS was not reached (NR) versus 70.0 (95%CI, 52.6–87.4) versus 51.3 (95%CI, 34.5–68.1) versus 33.8 (95%CI, 22.9–44.6) months in the MPSS I, II, III, and IV groups, respectively (P<0.001). The stratification performance of MPSS was similarly excellent in the median PFS (P<0.001). The differences in diverse MPSS groups were statistically significant (Fig 1C).

In the independent external validation cohort 2, patients were also classified by MPSS model into four groups accounting for 30.2%, 25.2%, 29.1%, and 15.5%, respectively. Median OS of patients from MPSS I to MPSS IV was 183.0 (95%CI, 150.0-NR), 144.0 (95%CI, 126.0–173.0), 101.0 (95%CI, 83.3–119.0), and 48.7 (95%CI, 38.2–64.4) months, respectively (P<0.001). Similarly, MPSS could effectively discriminate diverse median PFS for patients in external validation cohort (P<0.001). The differences in diverse MPSS groups were also statistically significant (Fig 1D).

Using bootstrap resampling, calibration curves for survival probability at 3-year and 5-year after treatment were then drafted, which showed an optimal correlation between the MPSS-prediction and actual outcome in the training cohort. The calibration curves at 3-year and 5-year in the validation cohort 1 and validation cohort 2 showed similar excellent correlation (Fig S6). These results confirmed both the efficacy and utility of the MPSS model in an independent validation dataset.

6. Comparison of MPSS with other predictive models

Discrimination and calibration for the MPSS model were then evaluated. In training cohort, C-statistics of OS prediction was 0.650 (95%CI, 0.623–0.677) for MPSS, which was significantly higher than R-ISS (0.605, 95%CI, 0.580–0.630, P<0.001). We also described the redistribution of patients from R-ISS to MPSS, and the redistribution from R2-ISS to MPSS. In terms of the restratification by MPSS, a total of 470 patients with R-ISS II with an unbalanced proportion (66.2%), could be furtherly divided as follows: 88 (18.7%) patients to MPSS I, 100 (21.3%) to MPSS II, 236 (50.2%) to MPSS III, and 46 (9.8%) to MPSS IV (Fig 2A). And these newly classified patients showed a statistically significant difference in the OS (P<0.001; Fig 2B) and PFS (P<0.001; Fig S7). Similar restaging performance of MPSS model could be observed in these heterogenous R2-ISS III patients (Fig 2 C–D; Fig S8). Importantly, the C-statistic of OS prediction for MPSS was also evidently higher than the R2-ISS (0.628, 95%CI, 0.601–0.655, P=0.014), and MASS (0.630, 95%CI, 0.603–0.657, P=0.043). Similar results could be found in the validation cohort (Fig 2E). Compared with R-ISS, R2-ISS, and MASS, the MPSS model displayed even better discrimination in predicting patient survival.

Figure 2: Comparison of MPSS with other predictive models.

Figure 2:

Open in a new tab

(A) redistribution from R-ISS to MPSS in training cohort; (B) OS of the heterogeneous R-ISS II patients redistributed by MPSS model; (C) redistribution from R2-ISS to MPSS in training cohort; (D) OS of the heterogeneous R2-ISS III patients redistributed by MPSS model; (E) comparisons of C-index among different risk models in training and validation cohorts.

Discussion

In this study, we constructed a weighted MPSS risk model integrating four significant independent variables with established implications in patient prognosis. By stratifying patients into four risk categories, the MPSS model stratified patients better than the current clinical staging systems R-ISS, R2-ISS, and MASS. According to the results of our Cox multivariate analysis which included most of the reported risk factors, four independent risk factors for PFS and OS were identified: thrombocytopenia, elevated LDH, ISS stage III, and HRA number (1 HRA and ≥2 HRA).

Thrombocytopenia was present in 10–20% of patients with MM, but was always associated with bone marrow suppression and tumor burden. In an analysis of 10750 MM patients2, thrombocytopenia was recognized as an independent adverse predictor of prognosis, with an increased risk of death (HR 1.63, P<0.001). In our longitudinal cohort, thrombocytopenia remains an independent predictor regardless of therapeutic regimens18. Serum LDH is another established important biomarker in MM associated with increased disease aggressiveness and high proliferation rate of MM cells25. Since elevated LDH has a major impact on patient survival, it has been incorporated into R-ISS26 as well as other clinical risk models4,5,27. Thus, we adopted thrombocytopenia representing aggressive host immune status18 and elevated LDH as symbols of high tumor burden in the MPSS model. It is important to note that thrombocytopenia and elevated LDH do not necessarily have fixed values. Different institutions set different thresholds to define normal upper and lower limits or to establish their own definitions, based on variations in race and detection methods.

The t(4;14), del(17p), and gain(1q) are now recognized as high-risk cytogenetic aberrations in MM 6,28–31, while the independent prognostic impact of t(14;16) continues to be a matter of debate due to its infrequent incidence and association with renal dysfunction32,33. Besides specific adverse cytogenetic lesions, there is increasing evidence that co-occurrence of more than one HRA is associated with poorer patient outcomes10,12,34. In fact, the concept of ‘multiple-hit’ HRA is exactly based on the observation that genetic aberrations in MM are not mutually exclusive, but rather frequently co-segregate during the clinical course of disease10–12. According to a recent report, the European Myeloma Network (EMN) built a weighted R2-ISS model consisting of 1q CNA, del(17p), t(4;14), high LDH, and ISS II, ISS III from over 7000 NDMM patients enrolling in European clinical trials35. In their study, patients were divided into four risk strata, with OS of not reached vs 109.2 vs 68.5 vs 37.9 months, respectively35. Similar to our study, 1q CNA was adopted and a weighting algorithm was constructed; however, the t(14;16) and host factors were not considered in the R2-ISS model.

In our analysis, gain(1q) was evaluated together with del(17p), t(4;14) and t(14;16). Patients with 0 to 3 HRA experienced progressively worse OS. Analysis of combination high-risk cytogenetics suggested that the number of high-risk genetic lesions can better reflect the multi-hit model of myeloma progression. Thus, the cumulative number of HRA was assessed as one variable in the Cox multivariate analysis, and then included in the final MPSS model.

In CLL and DLBCL, weighted prognostic systems like CLL-IPI and NCCN-IPI have been established, reflecting both tumor complexity and providing additional prognostic information36,37. Although current prognostic models in myeloma have attempted to incorporate the most important markers known, weighting various prognostic factors has not yet been done. The recently proposed MASS model could achieve effective risk stratification using a simple summation of five risk factors5; however, this pattern does not consider the inherent imbalance of hazard weight among different predictors, which may cause deviations in accurate risk distribution. In our analysis, the impact of each indicator was assessed by their HR value in Cox models before incorporating it into the MPSS model. This strategy allowed us to use a weighted combination of host, biological, and cytogenetic markers for accurate prognostication.

R-ISS is by far the most used predictive model in MM. One controversy associated with R-ISS was the inappropriate large proportion of R-ISS II (nearly 60%)38. Moreover, validation of R-ISS in subsequent studies showed less consistency due, at least in part, to the heterogeneity of patients between clinical trials and real-world practice39,40. Within R-ISS II groups in our study, four subgroups with diverse survival outcomes could be furtherly identified by MPSS, indicating the high degree of heterogeneity in patients classified as R-ISS II. Meanwhile, we must acknowledge the nearly 40% proportion of MPSS III in the BDH cohort, indicating slight imbalance and heterogeneity. This may arise from the higher frequency of chromosome 1q21 abnormality (⩾50%) in the Chinese MM cohort30,41. Nevertheless, upon MPSS model validation in the UAMS cohort, we found a more balanced and rational distribution with around 30% for both MPSS I and III. This leads us to believe that the application of MPSS can improve survival prediction among myeloma patients in different regions and identify those who may derive less advantage from current treatments.

In subgroups identified by different age and diverse therapeutic eras, the MPSS risk model still retained its utility and reproducibility, which emphasizes the universality of MPSS model across diverse treatment patterns in different regions of the world. Although treatment categories and transplantation have been incorporated into some prognostic models42, they were not incorporated into our model. In fact, with the rapid progress in the era of myeloma treatment, the exclusion of therapy-related factors from the predictive model will facilitate its wider usage; further validation of this predictive model in the era of evolving novel therapies may identify patients at higher risk before the initiation of any treatment.

Sample size of the independent external validation cohort was comparable to our training set. Of note, more high-risk patients with renal failure, high β2MG, thrombocytopenia, and ISS III were observed in the training cohort than validation cohort 2, probably due to the selection of patients for TT clinical trials19–21. As a result, less proportion of MPSS III/IV were observed in the UAMS dataset (60.9% vs. 44.6%, P<0.001; Table S8). In addition, almost of all patients in validation cohort 2 underwent ASCT but the proportion of those getting ASCT was only 26.9% in BDH cohort43. Both of these factors may account for the slightly worse performance of risk stratification by MPSS in external validation cases compared with the training set. But it is noteworthy that the MPSS model may be more generalizable and better appropriate for NDMM cohorts receiving standard treatment in the real-world.

Inevitably, there are several potential defects in our studies. First, prognostic factors employed in our analysis were selected based on standard guidelines, as well as our previous clinical experience to facilitate clinical adaptation in real-world settings. Factors such as BMI, CRP, and others were not included in our analysis for the interest of developing a practical model that can be more universally applied. Second, indicators reflecting patients’ physical condition such as WHO performance status or Eastern Cooperative Oncology Group (ECOG) were not included in our analysis, as they are subjectively evaluated, ie pain in previously-untreated patients with bone fractures at diagnosis. Third, considering the rapid advances in novel immunotherapies represented by anti-CD38 monoclonal antibodies and CAR-T, and refinement of our model in the context of large-scale use of immunotherapies will optimize its utility in the future.

In conclusion, we formulated and validated the MPSS risk model to predict the prognosis of patients with MM using readily available standard clinical and genetic test data. The established MPSS profile shows a better performance in risk discrimination than the current R-ISS, R2-ISS, and MASS. A score-based risk stratification is derived, and identifies patients with high and ultra-high risk of death after diagnosis and may therefore aid the development of more personalized treatment strategies, especially for patients for whom current therapies are likely to fail.

Supplementary Material

Supinfo

Figure S1: Prognostic impact of different copy numbers of chromosome1q21. (A, B) PFS and OS of different copy numbers of chromosome 1q21

Figure S2: Prognosis of different molecular risk groups. (A, B) PFS and OS of patients with different number HRA; (C, D) PFS and OS of patients with different co-existed HRA.

Figure S3: Prognosis for training cohort within the MPSS score segment. (A, B) PFS and OS of different MPSS scores

Figure S4: Predictive performances of MPSS model in different age subgroups. (A) OS prediction by MPSS in young patients (age < 65 years old); (B) OS prediction by MPSS in elder patients (age ≥ 65 years old).

Figure S5: Predictive performances of MPSS model in diverse therapeutic eras subgroups. (A) OS prediction by MPSS in patients diagnosed from 1999 to 2013; (B) OS prediction by MPSS in patients diagnosed from 2014 to 2017; (C) OS prediction by MPSS in patients diagnosed from 2018 to 2020.

Figure S6: MPSS model calibration in training and validation cohorts. (A) predicted 3-year OS probability; (B) predicted 5-year OS probability. Illustrated are the expected number of events and actual number of events at 3 and 5 years after diagnosis.

Figure S7: PFS of the R-ISS II patients redistributed by MPSS model

Figure S8: PFS of the R2-ISS III patients redistributed by MPSS model

Financial Disclosure

This work was supported by the grants to L.Q. from the Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (2022-I2M-1-022) and the National Natural Science Foundation of China (81920108006), to G.A. from the Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (2021-I2M-1-041) and the National Natural Science Foundation of China (82270218 and U22A20291). The study was also supported by grants to FZ from the National Cancer Institute 1R01CA236814-01A1 (FZ), 3R01-CA236814-03S1 (FZ), and U54CA272691-01 (FZ and JDS), U.S. Department of Defense (DoD) CA180190 (FZ) as well as funding from the Myeloma Crowd Research Initiative Award (FZ) and the Paula and Rodger Riney Foundation (FZ), and UAMS Winthrop P. Rockefeller Cancer Institute (WRCRI) Fund (FZ).

Footnotes

Conflict of Interests Disclosures: The authors declare no conflict of financial interests.

Reference

  • 1.Palumbo A, Anderson K. Multiple myeloma. N Engl J Med. Mar 17 2011;364(11):1046–60. doi: 10.1056/NEJMra1011442 [DOI] [PubMed] [Google Scholar]
  • 2.Greipp PR, Miguel JS, Durie BGM, et al. International Staging System for Multiple Myeloma. Journal of Clinical Oncology. 2005;23(15):3412–3420. doi: 10.1200/jco.2005.04.242 [DOI] [PubMed] [Google Scholar]
  • 3.Palumbo A, Avet-Loiseau H, Oliva S, et al. Revised International Staging System for Multiple Myeloma: A Report From International Myeloma Working Group. Journal of Clinical Oncology. 2015;33(26):2863–2869. doi: 10.1200/jco.2015.61.2267 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.D’Agostino M, Cairns DA, Lahuerta JJ, et al. Second Revision of the International Staging System (R2-ISS) for Overall Survival in Multiple Myeloma: A European Myeloma Network (EMN) Report Within the HARMONY Project. J Clin Oncol. May 23 2022:Jco2102614. doi: 10.1200/jco.21.02614 [DOI] [PubMed] [Google Scholar]
  • 5.Abdallah NH, Binder M, Rajkumar SV, et al. A simple additive staging system for newly diagnosed multiple myeloma. Blood Cancer J. Jan 31 2022;12(1):21. doi: 10.1038/s41408-022-00611-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Avet-Loiseau H, Hulin C, Campion L, et al. Chromosomal Abnormalities Are Major Prognostic Factors in Elderly Patients With Multiple Myeloma: The Intergroupe Francophone du Myélome Experience. Journal of Clinical Oncology. 2013;31(22):2806–2809. doi: 10.1200/jco.2012.46.2598 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Rajan AM, Rajkumar SV. Interpretation of cytogenetic results in multiple myeloma for clinical practice. Blood cancer journal. 2015;5(e365)doi: 10.1038/bcj.2015.92 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Schmidt TM, Barwick BG, Joseph N, et al. Gain of Chromosome 1q is associated with early progression in multiple myeloma patients treated with lenalidomide, bortezomib, and dexamethasone. Blood Cancer J. Nov 25 2019;9(12):94. doi: 10.1038/s41408-019-0254-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Du C, Mao X, Xu Y, et al. 1q21 gain but not t(4;14) indicates inferior outcomes in multiple myeloma treated with bortezomib. Leukemia & Lymphoma. 2019:1–10. doi: 10.1080/10428194.2019.1700503 [DOI] [PubMed] [Google Scholar]
  • 10.Boyd KD, Ross FM, Chiecchio L, et al. A novel prognostic model in myeloma based on co-segregating adverse FISH lesions and the ISS: analysis of patients treated in the MRC Myeloma IX trial. Leukemia. Feb 2012;26(2):349–55. doi: 10.1038/leu.2011.204 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Binder M, Rajkumar SV, Ketterling RP, et al. Prognostic implications of abnormalities of chromosome 13 and the presence of multiple cytogenetic high-risk abnormalities in newly diagnosed multiple myeloma. Blood Cancer J. Sep 1 2017;7(9):e600. doi: 10.1038/bcj.2017.83 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Sergentanis TN, Kastritis E, Terpos E, Dimopoulos MA, Psaltopoulou T. Cytogenetics and Survival of Multiple Myeloma: Isolated and Combined Effects. Clinical lymphoma, myeloma & leukemia. Jun 2016;16(6):335–40. doi: 10.1016/j.clml.2016.03.006 [DOI] [PubMed] [Google Scholar]
  • 13.Rodríguez-Otero P, Paiva B, Engelhardt M, Prósper F, San Miguel JF. Is immunotherapy here to stay in multiple myeloma? Haematologica. Mar 2017;102(3):423–432. doi: 10.3324/haematol.2016.152504 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Steiger S, Lutz R, Prokoph N, et al. Bone Marrow Immune Signatures in Multiple Myeloma Are Linked to Tumor Heterogeneity and Treatment Outcome. Blood. 2022;140(Supplement 1):2083–2085. doi: 10.1182/blood-2022-168685 [DOI] [Google Scholar]
  • 15.Anderson KC, Carrasco RD. Pathogenesis of myeloma. Annu Rev Pathol. 2011;6:249–74. doi: 10.1146/annurev-pathol-011110-130249 [DOI] [PubMed] [Google Scholar]
  • 16.Edwards CV, Hassan H, Yildirim C, et al. Peripheral Blood Monocyte Count is a Dynamic Prognostic Biomarker in Multiple Myeloma. Blood Adv. Nov 21 2022;doi: 10.1182/bloodadvances.2022008021 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.NCCN Guidelines® Insights: Prostate Cancer, Version 1.2023: Featured Updates to the NCCN Guidelines. [Google Scholar]
  • 18.Kuang C, Xia M, An G, et al. Excessive serine from the bone marrow microenvironment impairs megakaryopoiesis and thrombopoiesis in Multiple Myeloma. Nat Commun. Apr 13 2023;14(1):2093. doi: 10.1038/s41467-023-37699-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Barlogie B, Tricot GJ, van Rhee F, et al. Long-term outcome results of the first tandem autotransplant trial for multiple myeloma. Br J Haematol. Oct 2006;135(2):158–64. doi: 10.1111/j.1365-2141.2006.06271.x [DOI] [PubMed] [Google Scholar]
  • 20.Barlogie B, Tricot G, Anaissie E, et al. Thalidomide and hematopoietic-cell transplantation for multiple myeloma. N Engl J Med. Mar 9 2006;354(10):1021–30. doi: 10.1056/NEJMoa053583 [DOI] [PubMed] [Google Scholar]
  • 21.Pineda-Roman M, Zangari M, Haessler J, et al. Sustained complete remissions in multiple myeloma linked to bortezomib in total therapy 3: comparison with total therapy 2. Br J Haematol. Mar 2008;140(6):625–34. doi: 10.1111/j.1365-2141.2007.06921.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Xiong W, Wu X, Starnes S, et al. An analysis of the clinical and biologic significance of TP53 loss and the identification of potential novel transcriptional targets of TP53 in multiple myeloma. Blood. Nov 15 2008;112(10):4235–46. doi: 10.1182/blood-2007-10-119123 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Zhan F, Huang Y, Colla S, et al. The molecular classification of multiple myeloma. Blood. Sep 15 2006;108(6):2020–8. doi: 10.1182/blood-2005-11-013458 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Zhou Y, Zhang Q, Stephens O, et al. Prediction of cytogenetic abnormalities with gene expression profiles. Blood. May 24 2012;119(21):e148–50. doi: 10.1182/blood-2011-10-388702 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Dimopoulos MA, Barlogie B, Smith TL, Alexanian R. High serum lactate dehydrogenase level as a marker for drug resistance and short survival in multiple myeloma. Annals of internal medicine. Dec 15 1991;115(12):931–5. doi: 10.7326/0003-4819-115-12-931 [DOI] [PubMed] [Google Scholar]
  • 26.Terpos E, Katodritou E, Roussou M, et al. High serum lactate dehydrogenase adds prognostic value to the international myeloma staging system even in the era of novel agents. European journal of haematology. Aug 2010;85(2):114–9. doi: 10.1111/j.1600-0609.2010.01466.x [DOI] [PubMed] [Google Scholar]
  • 27.Moreau P, Cavo M, Sonneveld P, et al. Combination of international scoring system 3, high lactate dehydrogenase, and t(4;14) and/or del(17p) identifies patients with multiple myeloma (MM) treated with front-line autologous stem-cell transplantation at high risk of early MM progression-related death. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. Jul 10 2014;32(20):2173–80. doi: 10.1200/jco.2013.53.0329 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Sonneveld P, Avet-Loiseau H, Lonial S, et al. Treatment of multiple myeloma with high-risk cytogenetics: a consensus of the International Myeloma Working Group. Blood. Jun 16 2016;127(24):2955–62. doi: 10.1182/blood-2016-01-631200 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Chng WJ, Dispenzieri A, Chim CS, et al. IMWG consensus on risk stratification in multiple myeloma. Leukemia. Feb 2014;28(2):269–77. doi: 10.1038/leu.2013.247 [DOI] [PubMed] [Google Scholar]
  • 30.An G, Xu Y, Shi L, et al. Chromosome 1q21 gains confer inferior outcomes in multiple myeloma treated with bortezomib but copy number variation and percentage of plasma cells involved have no additional prognostic value. Haematologica. 2014;99(2):353–359. doi: 10.3324/haematol.2013.088211 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Weinhold N, Salwender HJ, Cairns DA, et al. Chromosome 1q21 abnormalities refine outcome prediction in patients with multiple myeloma - a meta-analysis of 2,596 trial patients. Haematologica. 2021;106(10):2754–2758. doi: 10.3324/haematol.2021.278888 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Greenberg AJ, Rajkumar SV, Therneau TM, Singh PP, Dispenzieri A, Kumar SK. Relationship between initial clinical presentation and the molecular cytogenetic classification of myeloma. Leukemia. 2014/02/01 2014;28(2):398–403. doi: 10.1038/leu.2013.258 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Schavgoulidze A, Perrot A, Cazaubiel T, et al. Prognostic impact of translocation t(14;16) in multiple myeloma according to the presence of additional genetic lesions. Blood Cancer J. Oct 26 2023;13(1):160. doi: 10.1038/s41408-023-00933-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Perrot A, Lauwers-Cances V, Tournay E, et al. Development and Validation of a Cytogenetic Prognostic Index Predicting Survival in Multiple Myeloma. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. May 15 2019:Jco1800776. doi: 10.1200/jco.18.00776 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.D’Agostino M, Lahuerta J-J, Wester R, et al. A New Risk Stratification Model (R2-ISS) in Newly Diagnosed Multiple Myeloma: Analysis of Mature Data from 7077 Patients Collected By European Myeloma Network within Harmony Big Data Platform. Blood. 2020;136(Supplement 1):34–37. doi: 10.1182/blood-2020-137021 [DOI] [Google Scholar]
  • 36.Zhou Z, Sehn LH, Rademaker AW, et al. An enhanced International Prognostic Index (NCCN-IPI) for patients with diffuse large B-cell lymphoma treated in the rituximab era. Blood. Feb 6 2014;123(6):837–42. doi: 10.1182/blood-2013-09-524108 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.An international prognostic index for patients with chronic lymphocytic leukaemia (CLL-IPI): a meta-analysis of individual patient data. Lancet Oncol. Jun 2016;17(6):779–790. doi: 10.1016/s1470-2045(16)30029-8 [DOI] [PubMed] [Google Scholar]
  • 38.Palumbo A, Avet-Loiseau H, Oliva S, et al. Revised International Staging System for Multiple Myeloma: A Report From International Myeloma Working Group. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. Sep 10 2015;33(26):2863–9. doi: 10.1200/jco.2015.61.2267 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Durie BG, Hoering A, Abidi MH, et al. Bortezomib with lenalidomide and dexamethasone versus lenalidomide and dexamethasone alone in patients with newly diagnosed myeloma without intent for immediate autologous stem-cell transplant (SWOG S0777): a randomised, open-label, phase 3 trial. Lancet (London, England). Feb 4 2017;389(10068):519–527. doi: 10.1016/s0140-6736(16)31594-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Kastritis E, Terpos E, Roussou M, et al. Evaluation of the Revised International Staging System in an independent cohort of unselected patients with multiple myeloma. Haematologica. Mar 2017;102(3):593–599. doi: 10.3324/haematol.2016.145078 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Chen H, Zhou N, Hu X, et al. The applicability of the Second Revision of the International Staging System for patients with multiple myeloma receiving immunomodulatory drugs or proteasome inhibitor-based regimens as induction treatment: A real-world analysis. Hematol Oncol. Feb 2023;41(1):139–146. doi: 10.1002/hon.3090 [DOI] [PubMed] [Google Scholar]
  • 42.Terebelo HR, Abonour R, Gasparetto CJ, et al. Development of a prognostic model for overall survival in multiple myeloma using the Connect((R)) MM Patient Registry. Br J Haematol. Aug 5 2019;doi: 10.1111/bjh.16139 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Usmani SZ, Crowley J, Hoering A, et al. Improvement in long-term outcomes with successive Total Therapy trials for multiple myeloma: are patients now being cured? Leukemia. Jan 2013;27(1):226–32. doi: 10.1038/leu.2012.160 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supinfo

Figure S1: Prognostic impact of different copy numbers of chromosome1q21. (A, B) PFS and OS of different copy numbers of chromosome 1q21

Figure S2: Prognosis of different molecular risk groups. (A, B) PFS and OS of patients with different number HRA; (C, D) PFS and OS of patients with different co-existed HRA.

Figure S3: Prognosis for training cohort within the MPSS score segment. (A, B) PFS and OS of different MPSS scores

Figure S4: Predictive performances of MPSS model in different age subgroups. (A) OS prediction by MPSS in young patients (age < 65 years old); (B) OS prediction by MPSS in elder patients (age ≥ 65 years old).

Figure S5: Predictive performances of MPSS model in diverse therapeutic eras subgroups. (A) OS prediction by MPSS in patients diagnosed from 1999 to 2013; (B) OS prediction by MPSS in patients diagnosed from 2014 to 2017; (C) OS prediction by MPSS in patients diagnosed from 2018 to 2020.

Figure S6: MPSS model calibration in training and validation cohorts. (A) predicted 3-year OS probability; (B) predicted 5-year OS probability. Illustrated are the expected number of events and actual number of events at 3 and 5 years after diagnosis.

Figure S7: PFS of the R-ISS II patients redistributed by MPSS model

Figure S8: PFS of the R2-ISS III patients redistributed by MPSS model

NIHMS1958049-supplement-Supinfo.pdf (3.6MB, pdf)

RESOURCES