A prognostic grading system for extramedullary multiple myeloma

Abstract

The application discloses a kind of extramedullary multiple myeloma prognosis grading evaluation systems, it includes medical record information acquisition and judging module, image acquisition and judging module, biological information acquisition and judging module, calculation and evaluation module;Wherein calculation and evaluation module according to the principle that EMD-RS total accumulative score is higher, patient prognosis risk level is higher, extramedullary multiple myeloma prognosis stratified risk is divided into I~IV level, the, indicates the score value corresponding to the i-th independent adverse prognostic factor, if patient has the i-th independent adverse prognostic factor, then 1, if not, then 0.Compared with existing staging system, based on secondary EMD, EME phenotype, maximum lesion diameter ≥5cm, del (17p), del (13q) and LDH elevation, the accuracy of the system constructed is improved.

Description

Technical Field

[0001] This invention belongs to the field of biomedical technology, and more specifically, relates to a prognostic grading assessment system for extramedullary multiple myeloma. Background Technology

[0002] Multiple myeloma (MM) is a malignant disease characterized by the abnormal proliferation of clonal plasma cells. It is the second most common hematologic malignancy worldwide, primarily affecting the elderly, and remains incurable. Recent research indicates that extramedullary lesions (EMD) are one of the most aggressive phenotypes of multiple myeloma and are frequently associated with treatment resistance and adverse outcomes (References 1 and 2). Increasing evidence suggests that EMD can independently predict poor prognosis regardless of disease stage, reflecting the inherent aggressiveness of the disease. EMD represents a biologically distinct disease entity, and a growing number of researchers consider extramedullary multiple myeloma to be a novel subgroup of multiple myeloma.

[0003] However, traditional staging systems for multiple myeloma do not incorporate EMD-specific characteristics, thus failing to adequately stratify the risk of this subgroup of extramedullary multiple myeloma. Furthermore, the International Staging System (ISS) and the Revised International Staging System (R-ISS) tend to classify extramedullary multiple myeloma patients into low- or intermediate-risk categories. For example, a large study at the Mayo Clinic reported that most patients with extramedullary multiple myeloma were initially classified as low- or intermediate-risk (ISSI–II: 72%; R-ISSI–II: 64%) (Reference 3).

[0004] Therefore, the traditional risk stratification system for multiple myeloma cannot fully capture its characteristics. There is an urgent need to establish a prognostic risk stratification system for extramedullary multiple myeloma to improve the accuracy of predicting clinical outcomes for patients with extramedullary multiple myeloma and to help clinicians optimize treatment strategies.

[0005] Source of literature: 1.D'AgostinoM,CairnsDA,LahuertaJJ,etal.SecondRevisionoftheInternationalStagingSystem(R2-ISS)forOverallSurvivalinMultipleMyeloma:AEuropeanMyelomaNetwork(EMN)ReportWithintheHARMONYProject.JClinOncol.2022;40(29):3406-3418. 2.HoM,ParuzzoL,MinehartJ,etal.ExtramedullaryMultipleMyeloma:ChallengesandOpportunities.CurrOncol.2025;32(3): 3.ZanwarS,HoM,LinY,etal.Natural history,predictors of development ofextramedullarydisease,andtreatmentoutcomesforpatientswithextramedullarymultiplemyeloma.AmJHematol.2023;98(10):1540-1549. Summary of the Invention

[0006] To address the aforementioned technical deficiencies or improvement needs, this invention provides a prognostic grading and assessment system for extramedullary multiple myeloma. Its purpose is to identify secondary EMD, EME phenotype, maximum lesion diameter ≥5cm, del(17p), del(13q), and elevated LDH as independent adverse prognostic factors for extramedullary multiple myeloma. Based on these independent adverse prognostic factors, an EMD-RS system is constructed, capable of classifying the prognostic risk of extramedullary multiple myeloma patients into grades I to IV. This solves the technical problem that existing multiple myeloma staging systems (ISS, R-ISS, and R2-ISS) do not incorporate EMD-specific characteristics, making accurate prognostic grading of extramedullary multiple myeloma patients difficult and resulting in poor accuracy in predicting patient prognostic risk.

[0007] To achieve the above objectives, according to the first aspect of the present invention, a prognostic grading assessment system for extramedullary multiple myeloma is provided, which predicts the prognostic risk of extramedullary multiple myeloma based on bioinformatics and imaging information, and includes a medical record information acquisition and judgment module, an image acquisition and judgment module, a bioinformatics acquisition and judgment module, and a calculation and assessment module. The medical record information acquisition and judgment module is used to acquire the medical record data and clinical course information of patients with extramedullary multiple myeloma to be evaluated. If the patient has extramedullary lesions at the time of initial diagnosis of multiple myeloma, it is judged as primary EMD; if the patient does not have extramedullary lesions at the time of initial diagnosis, but develops extramedullary lesions during subsequent treatment or disease progression, it is judged as secondary EMD; and the judgment result is submitted to the calculation and evaluation module. The image acquisition and judgment module is used to acquire whole-body PET / CT image data of patients with extramedullary multiple myeloma to be evaluated before treatment, identify and analyze the imaging characteristics of extramedullary lesions, determine whether the patient has the EME phenotype, and determine whether the maximum lesion diameter is ≥5cm based on the size of the largest lesion in the image; and submit the judgment result to the calculation and evaluation module. The bioinformatics acquisition and judgment module is used to acquire the patient's cytogenetic abnormality information and blood biochemical test index information, determine whether the patient has del(17p) and del(13q) abnormalities, and acquire the lactate dehydrogenase (LDH) test value and its corresponding reference range when the patient has EMD. When the LDH test result is higher than the upper limit of the normal reference range, it is judged as LDH elevation; otherwise, it is judged as LDH not elevation; and the judgment result is submitted to the calculation and evaluation module. The calculation and assessment module performs prognostic grading according to the principle that the higher the total cumulative EMD-RS score, the higher the patient's prognostic risk level. The total cumulative EMD-RS score is calculated according to the following formula: ; In the formula This represents the score corresponding to the i-th independent adverse prognostic factor, which includes secondary EMD, EME phenotype, maximum lesion diameter ≥5cm, del(17p), del(13q), and elevated LDH; among which The regression coefficients of the multivariate Cox model constructed from the independent adverse prognostic factors are obtained by scaling the coefficients proportionally, rounding them to the nearest 0.5 times, or by approximate merging. This indicates whether the patient has the i-th independent adverse prognostic factor. If the patient has this independent adverse prognostic factor, then... The value is 1; if it does not exist, then... The value is 0.

[0008] Preferably, the system, wherein the Standardized weights are assigned using a minimum scoring unit of 0.5. The standardized weights are assigned by discretization according to fixed rules. Calculate using the following formula: ; In the formula, The standardized weight of the i-th independent adverse prognostic factor; Let be the regression coefficient of the i-th independent adverse prognostic factor; The regression coefficient of the independent adverse prognostic factors used as a reference among the aforementioned independent adverse prognostic factors is assigned a score as follows: Standardized weights ≥1.25, The score is 1.5 points; 0.75 ≤ Standardized weight <1.25, The score is 1 point; 0.25 ≤ Standardized weight <0.75, The score is 0.5 points; Standardized weights <0.25, The score is 0.

[0009] Preferably, the system, wherein the Calculated using the following formula: ; In the formula, This represents the hazard ratio of the i-th independent adverse prognostic factor in a multivariate Cox model.

[0010] Preferably, in the system, the reference independent adverse prognostic factor is del(17p). The scores were assigned according to the following ratio: secondary EMD: EME phenotype: maximum lesion diameter ≥5cm: del(17p): del(13q): LDH elevation = 1.5:1:1:1:0.5:0.5.

[0011] Preferably, the system, wherein the Scoring will be conducted as follows: The first independent adverse prognostic factor was secondary EMD, with an S1 score of 1.5. The second independent adverse prognostic factor was the EME phenotype, with an S2 score of 1. The third independent adverse prognostic factor was the largest lesion diameter ≥5cm, with an S3 score of 1 point; The fourth independent adverse prognostic factor was del (17p), and S4 was assigned a score of 1. The fifth independent adverse prognostic factor was del(13q), with an S5 score of 0.5. The sixth independent adverse prognostic factor was elevated LDH, with an S6 score of 0.5.

[0012] Preferably, in the system, the calculation and evaluation module assesses the patient's prognostic grading based on the calculated total cumulative EMD-RS score using the following method: If the patient's total cumulative EMD-RS score is 0 to 0.5, their prognostic classification is assessed as Grade I (low risk). If a patient's total cumulative EMD-RS score is 1 to 1.5, their prognostic classification is assessed as Grade II (low to intermediate risk). If a patient's total cumulative EMD-RS score is 2 to 2.5, their prognostic classification is assessed as Grade III (intermediate to high risk). If a patient's total cumulative EMD-RS score is 3 to 5.5, their prognostic classification is assessed as Grade IV (high risk).

[0013] Preferably, in the system, the calculation and evaluation module further predicts the patient's overall survival based on prognostic grading: If the patient's prognostic classification is grade I, the predicted overall survival is 60.24-NR months; If the patient's prognostic grade is II, the predicted overall survival is 35.62-55.98 months; If the patient's prognostic grade is III, the predicted overall survival is 17.87-28.96 months; If a patient's prognostic grade is IV, their predicted overall survival is 2.94-10.18 months.

[0014] Preferably, in the system, the calculation and evaluation module predicts the patient's overall survival according to the following method: If the patient's prognostic classification is grade I, the predicted median overall survival is not reached. If the patient's prognostic grade is II, the predicted median overall survival is 45.75 months; If the patient's prognostic grade is III, the predicted median overall survival is 23.00 months; If a patient's prognostic grade is IV, the predicted median overall survival is 6.63 months.

[0015] Preferably, in the system, the calculation and evaluation module further predicts the patient's progression-free survival based on prognostic grading: If the patient's prognostic grading is grade I, the predicted progression-free survival is 43.89–69.23 months; If the patient's prognostic grading is grade II, the predicted progression-free survival is 13.41–22.65 months; If the patient's prognostic grading is grade III, the predicted progression-free survival is 7.66-15.24 months; If a patient's prognostic grade is IV, their predicted progression-free survival is 2.42-4.17 months.

[0016] Preferably, in the system, the calculation and evaluation module predicts the patient's progression-free survival according to the following method: If the patient's prognostic classification is grade I, the predicted median progression-free survival is 56.50 months. If the patient's prognostic classification is grade II, the predicted median progression-free survival is 17.7 months; If the patient's prognostic grade is III, the predicted median progression-free survival is 11.18 months; If a patient's prognostic grade is IV, the predicted median progression-free survival is 3.27 months.

[0017] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: This invention identifies multiple independent adverse prognostic factors for patients with extramedullary multiple myeloma, including secondary EMD, EME phenotype, maximum lesion diameter ≥5cm, del(17p), del(13q), and elevated LDH. Based on these identified independent adverse prognostic factors, a specific prognostic grading system for extramedullary multiple myeloma is constructed. This system integrates clinical course information, PET / CT imaging features, and biological indicators to uniformly quantify and assess independent adverse prognostic factors such as secondary EMD, EME phenotype, maximum lesion diameter, cytogenetic abnormalities (del(17p) and del(13q)), and elevated LDH. The constructed EMD-RS model shows an average net benefit higher than the corresponding traditional baseline staging systems (R-ISS, ISS, R2-ISS) across all risk threshold ranges, improving the specificity, comprehensiveness, and accuracy of prognostic staging for extramedullary multiple myeloma. Meanwhile, the EMD-RS total cumulative score calculation of this invention is simple and reproducible, making it easy to achieve standardized application. It helps in the prognostic risk stratification, individualized treatment decision-making, and survival outcome prediction of patients with extramedullary multiple myeloma, and has good clinical practical value and prospects for promotion and application. Attached Figure Description

[0018] Figure 1 The figures show the overall survival and progress-free survival of the total queue, training queue, and validation queue. In the figure, (A) is the Kaplan-Meier survival curve for the overall survival (OS) of the total queue, training queue, and validation queue; and (B) is the Kaplan-Meier survival curve for progress-free survival (PFS) of the total queue, training queue, and validation queue.

[0019] Figure 2 This is a multivariate analysis of prognostic factors for overall survival (OS) and progression-free survival (PFS) in the training cohort. The forest plot shows the adjusted hazard ratios (HRs) and their 95% confidence intervals derived from the final multivariate Cox regression model, where (A) is OS and (B) is PFS.

[0020] Figure 3This is the prognostic stratification of the EMD-RS model. Kaplan–Meier survival curves show the overall survival (OS) and progression-free survival (PFS) stratified by EMD-RSI–IV stage in the training cohort (A, B), validation cohort 1 (C, D), and validation cohort 2 (E, F). (G, H) are subgroup analyses of OS stratified by age and autologous stem cell transplantation (ASCT) status.

[0021] Figure 4 This is a performance evaluation of the EMD-RS model. In the figure, A and B are the time-dependent ROC curves of total survival (OS) in the training and validation cohorts at 12, 24, and 36 months, while C and D are the calibration curves of predicted OS and actual observed OS.

[0022] Figure 5 This refers to the predictive performance of EMD-RS for progression-free survival (PFS) in key subgroups. Figure 5 In the table, A represents the predictive performance of EMD-RS for progression-free survival (PFS) in patients ≤65 years of age, B represents the predictive performance of EMD-RS for progression-free survival (PFS) in patients ≥65 years of age, C represents the predictive performance of EMD-RS for progression-free survival (PFS) in patients who have received autologous stem cell transplantation (ASCT), and D represents the predictive performance of EMD-RS for progression-free survival (PFS) in patients who have not received autologous stem cell transplantation (ASCT).

[0023] Figure 6 The diagram shows the results of the decision curve analysis (DCA). In the training cohort (A, B), validation cohort 1 (C, D), and validation cohort 2 (E, F), the decision curve analysis (DCA) results show that at 12 and 24 months, the net benefit (NB) of EMD-RS is consistently better than that of the ISS, R-ISS, and R2-ISS scoring systems.

[0024] Figure 7 This figure shows the survival outcomes of extramedullary multiple myeloma (EMM) patients stratified by risk groups: ISS, R-ISS, and R2-ISS. Kaplan-Meier survival curves illustrate overall survival (OS; A to C) and progression-free survival (PFS; D to F) stratified by ISS, R-ISS, and R2-ISS. The concordance index (C-index) calculated by the Cox proportional hazards model is shown below the legend. Each panel provides the hazard ratio (HR) and its 95% confidence interval (CI) for pairwise comparisons, with only comparisons explicitly marked as not statistically significant (P ≥ 0.05); all other pairwise comparisons are statistically significant (P < 0.05). The median OS or median PFS for each stage is indicated in the figure. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0026] Terminology Explanation: Extramedullary disease (EMD): Defined as clonal plasma cell infiltration beyond the bone marrow microenvironment, it is one of the most aggressive manifestations of multiple myeloma (MM). Extramedullary disease (EMD) occurs when clonal plasma cells escape the bone marrow (BM) microenvironment and infiltrate soft tissues, distant organs, or periosseous tissues.

[0027] Primary EMD: Extramedullary lesions are present at the time of initial diagnosis of multiple myeloma (MM), with no prior history of MM.

[0028] Secondary EMD: When multiple myeloma (MM) is initially diagnosed, there are no extramedullary lesions (i.e., non-EMD-MM), but extramedullary lesions appear after relapse or progression.

[0029] Extramedullary multiple myeloma (EMM) refers to tumors formed when multiple myeloma cells breach the bone marrow barrier and metastasize to soft tissues or organs outside the bone marrow. It is one of the most aggressive forms of MM and is now considered a high-risk subtype. Based on its relationship to the skeleton, EMM can be further divided into two categories: Extramedullary bone-related lesions (EMB) refer to lesions caused by multiple myeloma cells breaching the bone cortex (bone surface) and directly invading the surrounding soft tissues or adjacent anatomical structures. In other words, these lesions remain adjacent to the primary bone lesion.

[0030] Non-bone-related extramedullary lesions (EME) refer to independent tumors formed by multiple myeloma cells spreading through the bloodstream to soft tissues, skin, or organs that are completely independent of the bone.

[0031] In this invention, the scope of extramedullary multiple myeloma includes bone-related extramedullary lesions (EMB) and non-bone-related extramedullary lesions (EME).

[0032] Overall survival (OS): defined as the time from EMD diagnosis to death from any cause; calculation method: usually expressed as "median OS", which means that 50% of patients die before this time point.

[0033] Progression-free survival (PFS): defined as the time from diagnosis of EMD to disease progression (such as hematological progression, tumor enlargement, or the appearance of new lesions) or death (regardless of the cause of death); calculation method: usually expressed as "median PFS".

[0034] del(17p): Deletion of the short arm of chromosome 17. The most important gene in this region is TP53.

[0035] del(13q): The long arm of chromosome 13 is deleted. This region contains multiple tumor suppressor genes, including RB1 (retinoblastoma gene).

[0036] This invention is based on a retrospective multicenter cohort study that included patients with extramedullary multiple myeloma (EMM), including EMB and EME, who visited three tertiary medical centers between January 2017 and July 2024. Univariate Cox regression analysis identified several variables significantly associated with overall survival (OS) and progression-free survival (PFS). These variables were incorporated into a LASSO Cox regression model. Through 10-fold cross-validation, LASSO selected a streamlined set of predictive factors, which were then incorporated into a multivariate Cox regression model. Based on the multivariate Cox regression model, we identified six independent adverse prognostic factors predicting OS: secondary EMD, EME phenotype, maximum lesion diameter ≥5 cm, del (17p), del (13q), and elevated LDH. Based on these factors, we established an EMD-RS scoring system and constructed a prognostic grading assessment system for extramedullary multiple myeloma.

[0037] Based on this, the present invention provides a prognostic grading assessment system for extramedullary multiple myeloma, which predicts the prognostic risk of extramedullary multiple myeloma based on bioinformatics and imaging information, including a medical record information acquisition and judgment module, an image acquisition and judgment module, a bioinformatics acquisition and judgment module, and a calculation and evaluation module. The medical record information acquisition and judgment module is used to acquire the medical record data and clinical course information of patients with extramedullary multiple myeloma to be evaluated. If the patient has extramedullary lesions at the time of initial diagnosis of multiple myeloma, it is judged as primary EMD; if the patient does not have extramedullary lesions at the time of initial diagnosis, but develops extramedullary lesions during subsequent treatment or disease progression, it is judged as secondary EMD; and the judgment result is submitted to the calculation and evaluation module. The image acquisition and judgment module is used to acquire whole-body PET / CT image data of patients with extramedullary multiple myeloma to be evaluated before treatment, identify and analyze the imaging characteristics of extramedullary lesions, determine whether the patient has the EME phenotype, and determine whether the maximum lesion diameter is ≥5cm based on the size of the largest lesion in the image; and submit the judgment result to the calculation and evaluation module. The bioinformatics acquisition and judgment module is used to acquire cytogenetic abnormality information and blood biochemical test index information of patients with extramedullary multiple myeloma to be evaluated, determine whether the patient has del(17p) and del(13q) abnormalities, and acquire the lactate dehydrogenase (LDH) test value and its corresponding normal reference range when the patient has EMD. When the LDH test result is higher than the upper limit of the normal reference range, it is judged as LDH elevation; otherwise, it is judged as LDH not elevation; and the judgment result is submitted to the calculation and evaluation module. The calculation and evaluation module classifies the prognostic risk level according to the principle that the higher the total cumulative EMD-RS score, the higher the prognostic risk level of the patient. In some embodiments, the prognostic risk level of extramedullary multiple myeloma is divided into four levels: Level I, Level II, Level III and Level IV, where Level I is low risk, Level II is low-intermediate risk, Level III is intermediate-high risk and Level IV is high risk.

[0038] The total cumulative score of EMD-RS is calculated according to the following formula: ; In the formula This represents the score corresponding to the i-th independent adverse prognostic factor; This indicates whether the patient has the i-th independent adverse prognostic factor. If the patient has this independent adverse prognostic factor, then... The value is 1; if it does not exist, then... The value is 0; the independent adverse prognostic factors include secondary EMD, EME phenotype, maximum lesion diameter ≥5cm, del (17p), del (13q) and elevated LDH; in The regression coefficients of the multivariate Cox model constructed from the independent adverse prognostic factors are obtained by scaling, rounding, or approximating.

[0039] The Standardized weights are assigned using a minimum scoring unit of 0.5. The standardized weights are obtained by rounding or approximate merging. Calculate using the following formula: ; In the formula, The standardized weight of the i-th independent adverse prognostic factor; Let be the regression coefficient of the i-th independent adverse prognostic factor; It is the regression coefficient of the independent adverse prognostic factors used as a reference.

[0040] In some embodiments Calculated using the following formula: ; In the formula, This represents the hazard ratio of the i-th independent adverse prognostic factor in a multivariate Cox model.

[0041] In some embodiments, the reference independent adverse prognostic factor is del(17p). The scores were assigned according to the following ratio: secondary EMD: EME phenotype: maximum lesion diameter ≥5cm: del(17p): del(13q): LDH elevation = 1.5:1:1:1:0.5:0.5.

[0042] For example, taking del(17p) as the reference independent adverse prognostic factor, let its regression coefficient be β. ref Then the relative weight W of the i-th risk factor i Calculate using the following formula: ; Using the above method, variables with different dimensions and effect sizes can be uniformly converted into relative weights W. i This facilitates subsequent integration and scoring. The present invention further discretizes the continuous standardized weights to form integer or half-integer integral values. In a preferred embodiment, 0.5 points is used as the smallest scoring unit for the standardized weights. Rounding or approximation is performed to obtain the final score. .

[0043] For example, standardized weights ≥1.25, Assign a value of 1.5 points; 0.75 ≤ Standardized weight <1.25, Assign 1 point; 0.25 ≤ Standardized weight <0.75, Assign a value of 0.5 points; Standardized weights <0.25, Assign a score of 0.

[0044] In some embodiments The assignment is as follows: The first independent adverse prognostic factor was secondary EMD. The value assigned is 1.5 points; The second independent adverse prognostic factor was the EME phenotype. The value assigned is 1 point; The third independent adverse prognostic factor was a maximum lesion diameter ≥5cm. The value assigned is 1 point; The fourth independent adverse prognostic factor was del (17p). The value assigned is 1 point; The fifth independent adverse prognostic factor was del(13q). The value assigned is 0.5 points; The sixth independent adverse prognostic factor was elevated LDH. The value assigned is 0.5 points.

[0045] For any patient to be tested, each of the above independent adverse prognostic factors is scored according to whether the patient possesses them, and all corresponding scores are summed to obtain the total EMD-RS score. The formula for calculating the total EMD-RS score is as follows: ; in, This represents the integral value corresponding to the i-th risk factor; This indicates whether the patient has the i-th risk factor. The value is 1 if the patient has the risk factor, and 0 otherwise. n represents the total number of risk factors included in the scoring system, and n=6.

[0046] The total cumulative EMD-RS score is calculated based on the scores of the above items (whether it is secondary EMD, EME phenotype, maximum lesion diameter ≥5cm, del (17p), del (13q) and elevated LDH), and the scores are assigned as follows: The first independent adverse prognostic factor was secondary EMD, with an S1 score of 1.5. The second independent adverse prognostic factor was the EME phenotype, with an S2 score of 1. The third independent adverse prognostic factor was the largest lesion diameter ≥5cm, with an S3 score of 1 point; The fourth independent adverse prognostic factor was del (17p), and S4 was assigned a score of 1. The fifth independent adverse prognostic factor was del(13q), with an S5 score of 0.5. The sixth independent adverse prognostic factor was elevated LDH, with an S6 score of 0.5.

[0047] The calculation is performed according to whether the patient has secondary EMD, X1 is 1, and if not, X1 is 0; The calculation is performed based on whether the patient has the EME phenotype, X2 is 1, and if not, X2 is 0. The calculation is performed as follows: if the patient's maximum lesion diameter is ≥5cm, X3 is 1; if the maximum lesion diameter is <5cm, X3 is 0. The calculation is performed according to the following: if the patient is del(17p), X4 is 1; if not del(17p), X4 is 0. The calculation is performed according to the following: if the patient is del(13q), X5 is 1; if not del(13q), X5 is 0. The calculation is performed based on whether the patient has elevated LDH, x6 is 1, and if not, x6 is 0.

[0048] The calculation and assessment module assesses the patient's prognostic grading based on the calculated total cumulative EMD-RS score using the following method: If the patient's total cumulative EMD-RS score is 0 to 0.5, their prognostic classification is assessed as Grade I (low risk). If a patient's total cumulative EMD-RS score is 1 to 1.5, their prognostic classification is assessed as Grade II (low to intermediate risk). If a patient's total cumulative EMD-RS score is 2 to 2.5, their prognostic classification is assessed as Grade III (intermediate to high risk). If a patient's total cumulative EMD-RS score is 3 to 5.5, their prognostic classification is assessed as Grade IV (high risk).

[0049] Preferably, the calculation and evaluation module also predicts the patient's overall survival based on prognostic grading: If the patient's prognostic classification is grade I, the predicted overall survival is 60.24-NR months; If the patient's prognostic grade is II, the predicted overall survival is 35.62-55.98 months; If the patient's prognostic grade is III, the predicted overall survival is 17.87-28.96 months; If a patient's prognostic grade is IV, their predicted overall survival is 2.94-10.18 months.

[0050] In some embodiments, the calculation and evaluation module predicts patient overall survival according to the following method: If the patient's prognostic classification is grade I, the predicted median overall survival is not reached. If the patient's prognostic grade is II, the predicted median overall survival is 45.75 months; If the patient's prognostic grade is III, the predicted median overall survival is 23.00 months; If a patient's prognostic grade is IV, the predicted median overall survival is 6.63 months.

[0051] More preferably, the calculation and assessment module also predicts progression-free survival based on prognostic grading: If the patient's prognostic grading is grade I, the predicted progression-free survival is 43.89–69.23 months; If the patient's prognostic grading is grade II, the predicted progression-free survival is 13.41–22.65 months; If the patient's prognostic grading is grade III, the predicted progression-free survival is 7.66-15.24 months; If a patient's prognostic grade is IV, their predicted progression-free survival is 2.42-4.17 months.

[0052] In some embodiments, the calculation and evaluation module predicts patient progression-free survival according to the following method: If the patient's prognostic classification is grade I, the predicted median progression-free survival is 56.50 months. If the patient's prognostic classification is grade II, the predicted median progression-free survival is 17.7 months; If the patient's prognostic grade is III, the predicted median progression-free survival is 11.18 months; If a patient's prognostic grade is IV, the predicted median progression-free survival is 3.27 months.

[0053] In some embodiments, the image acquisition and judgment module is used to acquire whole-body PET / CT image data of patients with extramedullary multiple myeloma at initial diagnosis, extract all individual extramedullary lesion regions from the PET / CT images, and count the number and location of extramedullary lesions; and determine whether it is a secondary EMD or EME phenotype according to the following method: If the number of extramedullary neoplastic lesions is 0 at the initial diagnosis and EMD only appears after treatment, it is judged as secondary EMD. If the number of extramedullary lesions is ≥1 at the initial diagnosis, it is judged as primary EMD. If an extramedullary lesion presents with non-periosteal infiltrating lesions, it is judged as an EME phenotype; if an extramedullary lesion is only a soft tissue lesion directly continuous with the bone cortex, it is judged as a non-EME phenotype, and the judgment result is submitted to the calculation and evaluation module. Based on the obtained maximum lesion diameter, determine whether the maximum lesion diameter is ≥5cm; and submit the determination result to the calculation and evaluation module; The bioinformatics acquisition and judgment module determines whether LDH levels are elevated using the following method: When EMD occurs, the LDH test result is higher than the upper limit of the normal reference range, indicating elevated LDH.

[0054] The following are examples. This retrospective, multicenter cohort study included extramedullary multiple myeloma (EMM) patients who were treated at three tertiary medical centers between January 2017 and July 2024: Shanghai Changzheng Hospital (training cohort), Renji Hospital affiliated with Shanghai Jiao Tong University School of Medicine, and Huadong Hospital affiliated with Fudan University (validation cohort).

[0055] Inclusion criteria include: (1) Patients with extramedullary multiple myeloma (EMM) aged ≥18 years; (2) Newly diagnosed or relapsed MM according to the International Myeloma Working Group (IMWG) criteria (Rajkumar SV, Dimopoulos MA, Palumbo A, et al. International Myeloma Working Group updatedcriteria for the diagnosis of multiple myeloma. The Lancet Oncology. 2014;15(12):e538-548.). (3) Extramedullary lesions (EMD), including bone-related extramedullary lesions (EMB) and non-bone-related extramedullary lesions (EME), are confirmed by biopsy or positron emission tomography (PET / CT). (4) Receive standard treatment (referring to standardized full-course systemic anti-myeloma treatment implemented according to IMWG and the Chinese guidelines for the diagnosis and treatment of multiple myeloma. The treatment regimen is based on a three-drug combination regimen of proteasome inhibitor, immunomodulatory agent, and dexamethasone, including induction chemotherapy, autologous hematopoietic stem cell transplantation for eligible patients, post-transplant consolidation therapy, and long-term maintenance therapy; relapsed patients are given salvage therapy with drugs of different mechanisms of action, and basic supportive and symptomatic treatment such as bone disease intervention, renal protection, correction of anemia, and prevention of thrombosis and infection are provided throughout the course). (5) Complete baseline clinical and laboratory data are available before treatment begins.

[0056] Exclusion criteria include: (1) Plasma cell leukemia, solitary plasma cell tumor, combined with active malignant tumor, or missing key baseline information.

[0057] (2) Disease progression at enrollment was defined according to the 2021 IMWG criteria, that is, the patient experienced biochemical or clinical relapse after previously achieving therapeutic efficacy (Moreau P, Kumar SK, San Miguel J, et al. Treatment of relapsed and refractory multiple myeloma: recommendations from the International Myeloma Working Group. The Lancet Oncology. 2021;22(3):e105-e118.).

[0058] All clinical data were extracted from electronic medical records and independently verified by a second researcher. Follow-up was updated to July 31, 2025, ensuring that all patients were observed for ≥12 months. This study was approved by the ethics committees of Shanghai Changzheng Hospital (2016SL019A), Renji Hospital (KY2020-191), and Huadong Hospital (2022K117). All participants signed written informed consent forms, and the study process followed the Declaration of Helsinki.

[0059] Example 1: Construction and Verification of the EMD-RS System This study included 287 patients with extramedullary multiple myeloma (EMM), and their baseline characteristics are shown in Table 1. Among them, 200 EMM patients collected from Changhai Changzheng Hospital served as the training cohort, and 87 EMM patients collected from Renji Hospital affiliated with Shanghai Jiao Tong University School of Medicine and Huadong Hospital affiliated with Fudan University served as the validation cohort (not the same group of patients as the training cohort). The overall survival (OS) and progression-free survival (PFS) of EMM patients in the training and validation cohorts are as follows: Figure 1 As shown.

[0060] Table 1. Clinical characteristics (baseline characteristics) of EMM patients (before treatment)

[0061]

[0062] Table 1 shows the age, bone marrow plasma cell ratio, and laboratory indicators based on baseline data at or closest to the time of EMD diagnosis. The determination of EMB and EME is primarily based on PET / CT imaging, with pathological biopsy used when necessary. EMB is defined as a soft tissue lesion directly continuous with the bone cortex, while EME is defined as involvement of extraosseous soft tissues or organs not directly continuous with the bone cortex. Primary EMD refers to the presence of extramedullary lesions at the initial diagnosis of multiple myeloma, while secondary EMD refers to the absence of extramedullary lesions at initial diagnosis but the appearance of extramedullary lesions during subsequent relapse or progression. The maximum lesion diameter is determined based on imaging measurements. Cytogenetic abnormalities include 1q21 gain / amplification, del(13q), del(17p), t(4;14), and t(11;14), all detected using bone marrow plasma cell fluorescence in situ hybridization (FISH). M protein was detected by serum protein electrophoresis / immunofixation electrophoresis; hemoglobin and platelets were detected by peripheral blood cell analysis; albumin, creatinine, lactate dehydrogenase (LDH), and calcium were detected by serum biochemistry; β2-microglobulin (β2-MG) was detected by serum immunological methods. Elevated LDH was defined as baseline serum LDH exceeding the upper limit of the normal reference range of the corresponding laboratory.

[0063] Continuous variables are expressed as median (interquartile range, IQR), and categorical variables are expressed as the percentage of cases. For comparisons between the training and validation cohorts, the Wilcoxon rank-sum test was used for continuous variables, and the Pearson chi-square test was used for categorical variables.

[0064] Table 1 shows that there were no significant differences between the training and validation cohorts in baseline demographic, clinical, cytogenetic, or laboratory characteristics. The median age at EMD diagnosis was 62 years (IQR, 54–68), with 55.4% being male. According to the ISS staging, 65.9% of patients were in stage I–II; according to the R-ISS staging, 85.4% were in stage I–II; and according to the R2-ISS staging, 31.0% were in stage I–II. 47.7% of patients had only one EMD lesion, while 52.3% had ≥2 EMD lesions; 36.9% of patients had a maximum lesion diameter ≥5 cm. 71.8% of patients had primary EMD at initial diagnosis, while 28.2% developed secondary EMD during disease progression. Overall, 68.6% of patients had high-risk cytogenetic features, with 55.4%, 10.5%, and 2.8% carrying one, two, and ≥ three high-risk abnormalities, respectively. The specific abnormality distribution was as follows: t(4;14) accounted for 12.2%, t(11;14) for 12.5%, t(14;16) for 2.4%, del(17p) for 7.3%, del(13q) for 28.9%, and 1q21 gain / amplification for 63.4%.

[0065] Of the 287 patients with extramedullary multiple myeloma (EMM) included, 92 patients (32.1%) had EME and 195 patients (67.9%) had EMB. The distribution of affected sites in these 92 EME patients is shown in Table S1.

[0066] Table S1 Distribution of affected sites in EME patients As shown in Table S1, the most common sites of involvement in the 92 EME patients were the skin or soft tissue (25.0%) and pleura (23.9%), followed by lymph nodes (18.5%) and the central nervous system (13.0%).

[0067] Table S2 Univariate Cox regression analysis of OS and PFS in extramedullary multiple myeloma Abbreviations: OS, overall survival; PFS, progression-free survival; HR, hazard ratio; CI, confidence interval; EMB, extramedullary bone-related lesion; EME, extramedullary bone lesion; ASCT, autologous stem cell transplantation; M protein, monoclonal protein; EMD, extramedullary lesion; EMD number, number of extramedullary lesions; Max lesion diameter ≥ 5 cm; t(4;14), translocation of chromosomes 4 and 14; t(11;14), translocation of chromosomes 11 and 14; t(14;16), translocation of chromosomes 14 and 16; del(17p), deletion of the short arm of chromosome 17; del(13q), deletion of the long arm of chromosome 13; 1q21 gain / amplification, increase or amplification of 1q21 site copy number; Hb, hemoglobin; Alb, albumin; β2-microglobulin, serum β2-microglobulin; Cr, serum creatinine; LDH, lactate dehydrogenase; Ca, serum calcium; Plt, platelet count.

[0068] Through 10-fold cross-validation, LASSO selected a set of concise predictors, which were then incorporated into a multivariate Cox model. For overall survival (OS), six variables retained independent prognostic significance: EME phenotype (HR 2.10, 95% CI 1.31–3.37), elevated LDH (HR 1.61, 95% CI 1.03–2.51), del(13q) (HR 1.80, 95% CI 1.14–2.83), del(17p) (HR 2.48, 95% CI 1.22–5.01), maximum lesion diameter ≥5 cm (HR 2.08, 95% CI 1.36–3.16), and secondary EMD (HR 3.76, 95% CI 2.33–6.07). Figure 2 A.

[0069] For PFS, seven variables had independent predictive significance: EME phenotype (HR 1.77, 95% CI 1.18–2.64), elevated LDH (HR 2.21, 95% CI 1.53–3.19), autologous hematopoietic stem cell transplantation (ASCT) (HR 0.48, 95% CI 0.33–0.72), del(13q) (HR 1.53, 95% CI 1.03–2.25), del(17p) (HR 2.19, 95% CI 1.18–4.06), largest lesion diameter ≥5 cm (HR 1.50, 95% CI 1.03–2.17), and secondary EMD (HR 2.95, 95% CI 1.97–4.41). Figure 2 B.

[0070] Six independent adverse prognostic factors remained in both endpoints: secondary EMD, EME, maximum lesion diameter ≥5 cm, del(17p), del(13q), and elevated LDH. The robustness of the model was confirmed through 1000 bootstrap iterations, and no violation of the proportional hazards assumption was observed. Based on a multivariate Cox regression model, we identified six independent adverse prognostic factors predicting overall survival (OS) (secondary EMD, EME, maximum lesion diameter ≥5 cm, del(17p), del(13q), and elevated LDH), and established a predictive model for the extramedullary lesion risk score (EMD-RS) accordingly. Details are as follows: This embodiment constructs a risk scoring system (EMD-RS) based on the clinical and biological characteristics of patients with extramedullary multiple myeloma. Using the Cox proportional hazards regression model results from the training cohort as a foundation, it converts the hazard ratios (HRs) corresponding to each independent prognostic factor into regression coefficients, and further standardizes and discretizes them to form a weighted integral system that can be used for clinical stratification assessment. Specifically: (a) Determining the modeling object and candidate variables First, based on the case data of the training cohort, information such as the clinical characteristics, lesion characteristics, cytogenetic abnormalities and laboratory indicators of the subjects were collected, and overall survival (OS) and / or progression-free survival (PFS) were used as the study endpoints.

[0071] Based on this, a univariate Cox proportional hazards regression analysis was performed on the candidate variables to screen variables that were significantly associated with the prognostic outcome, as follows: ① Data collection: Obtain the survival time, outcome event status (e.g., survival / death) and candidate feature variables (including demographic characteristics, disease characteristics, baseline stage, cytogenetic characteristics and laboratory parameters) of the training cohort samples.

[0072] ② Data processing: Deleting or imputing missing data (such as mean imputation, KNN imputation) to ensure the completeness of the analysis.

[0073] ③ Constructing a Surv object: Use the Surv() function in R to construct a survival analysis object.

[0074] ④ Perform univariate Cox regression: Using iterative modeling, Cox regression analysis is performed separately for each candidate variable. In this embodiment, a for loop is used to iterate through all variables. This process will output the hazard ratio (HR) of each variable, its 95% confidence interval (CI), and the significance level (p-value).

[0075] ⑤ Screening of significant variables: Significant variables were screened based on the p-value threshold (p < 0.05), and these variables will be included in the subsequent LASSO Cox regression analysis model. Univariate Cox analysis identified several variables significantly associated with OS and PFS in extramedullary multiple myeloma (Table S2), and these variables were all included in the LASSO Cox regression model (Table S2).

[0076] Further, variables with statistical or clinical significance were included in the multivariate Cox proportional hazards regression model. The independent adverse prognostic factors identified included secondary extramedullary lesions (EMD), EME phenotype, maximum lesion diameter ≥5cm, del (17p), del (13q), and elevated lactate dehydrogenase (Elevated LDH).

[0077] (II) Determination of Regression Coefficients For each independent adverse prognostic factor retained in the multivariate Cox regression model, the corresponding regression coefficient β is calculated based on its hazard ratio (HR). The regression coefficient corresponding to the i-th independent adverse prognostic factor is determined according to the following formula: ; in, Represents the regression coefficient of the i-th independent adverse prognostic factor; Let represent the hazard ratio of the i-th independent adverse prognostic factor in the Cox model; ln represents the natural logarithm. Through the above transformation, the relative risk effects of each independent adverse prognostic factor can be uniformly mapped into a linearly additive coefficient form, thus providing a basis for the subsequent construction of an integral model.

[0078] (III) Construction of Standardized Weights To improve the interpretability and operability of the scoring system, this embodiment further standardizes the regression coefficients of each independent adverse prognostic factor. Specifically, a pre-defined reference independent adverse prognostic factor is selected as the benchmark, its weight is defined as 1, and the standardized weights of other independent adverse prognostic factors relative to this benchmark are calculated. For example, del(17p) is selected as the reference independent adverse prognostic factor, and its regression coefficient is set to β. ref Then the relative weight W of the i-th independent adverse prognostic factor i Calculate using the following formula: ; in, The standardized weight of the i-th independent adverse prognostic factor; Let be the regression coefficient of the i-th independent adverse prognostic factor; The regression coefficients are used as a reference for independent adverse prognostic factors. Using the above method, variables with different dimensions and effect sizes can be uniformly converted into relative weights, facilitating subsequent integration and scoring.

[0079] (iv) Discretization of integral values Considering the need for convenient, stable, and easily calculated and stratified scoring tools in clinical applications, this invention further discretizes the continuous standardized weights to form integer or half-integer integral values. Using 0.5 points as the smallest scoring unit, scores are assigned according to fixed rules: standardized weight ≥ 1.25, 1.5 points; 0.75 ≤ standardized weight < 1.25, 1 point; 0.25 ≤ standardized weight < 0.75, 0.5 points; standardized weight < 0.25, 0 points.

[0080] Therefore, the EMD-RS scores corresponding to each independent adverse prognostic factor were constructed. In this embodiment, based on the OS multivariate Cox regression model results of 200 patients in the training cohort, the HR values ​​of each independent adverse prognostic factor were calculated, and the regression coefficients were converted, standardized, and discretized according to the above method. The results are as follows: Secondary extramedullary lesions (EMD): HR=3.76, according to calculate With del(17p) as a reference, the relative weight after standardization is approximately 1.324 / 0.908≈1.46, and after discretization, it is assigned a score of 1.5.

[0081] EME phenotype: HR=2.10, according to calculate With del(17p) as a reference, the relative weight after standardization is approximately 0.742 / 0.908≈0.820, and the score after discretization is 1 point.

[0082] Maximum lesion diameter ≥5cm: HR=2.08, according to calculate With del(17p) as a reference, the relative weight after standardization is approximately 0.732 / 0.908≈0.81, and the score after discretization is 1 point.

[0083] del(17p): HR=2.48, according to calculate As a benchmark item, its standardized weight is defined as 1, and it is assigned a score of 1.

[0084] del(13q): HR=1.80, according to calculate With del(17p) as a reference, the relative weight after standardization is approximately 0.588 / 0.908≈0.65, and the score after discretization is 0.5.

[0085] Elevated lactate dehydrogenase (LDH): HR = 1.61, according to... calculate With del(17p) as a reference, the relative weight after standardization is approximately 0.476 / 0.908≈0.52, and the score after discretization is 0.5.

[0086] Therefore, the following scoring system is formed: Secondary EMD: 1.5 points, EME phenotype: 1 point, maximum lesion diameter ≥5cm: 1 point, del (17p): 1 point, del (13q): 0.5 points, elevated LDH: 0.5 points.

[0087] (v) Calculate the total cumulative score For any patient to be tested, scores are assigned to each of the aforementioned independent adverse prognostic factors based on whether they possess them. All corresponding scores are then summed to obtain the total cumulative score, i.e., the overall risk score. The total cumulative score is calculated using the following formula: ; in, This represents the score corresponding to the i-th independent adverse prognostic factor; This indicates whether the patient has the i-th independent adverse prognostic factor. If the patient has this independent adverse prognostic factor, then... The value is 1, otherwise The value is 0; n represents the total number of independent adverse prognostic factors included in the scoring system. In this example, n=6, and the final score is composed of: secondary EMD (1.5 points), del (17p) (1 point), EME phenotype (1 point), maximum lesion diameter ≥5 cm (1 point), del (13q) (0.5 points), and elevated LDH (0.5 points), with a total cumulative score range of 0-5.5 points (Table 2).

[0088] Table 2. EMD-RS scoring definition based on training queue (n = 200) Rating value The coefficients of the multifactor Cox model were obtained by scaling and rounding, with the coefficients of del(17p) as a reference (1 point).

[0089] Abbreviations: EMD, extramedullary lesion; EME, extra-bone extramedullary myeloma; del, deletion; LDH, lactate dehydrogenase; N, number of cases; PFS, progression-free survival; OS, overall survival; HR, hazard ratio; NR, not reached; CI, confidence interval.

[0090] Furthermore, to determine the most discriminative and feasible risk stratification schemes, we performed grid search optimization (step size = 0.5) in the training cohort. All feasible 2-, 3-, and 4-group schemes were compared using Harrell's C-index, global log-rank p-value, and IBS, while requiring a minimum patient proportion of ≥15% in each stratum. The 4-group scheme (cutoff point = 0.5 / 1.5 / 2.5) achieved the highest discriminative power (C-index = 0.795) and showed significant separation between strata (P < 0.001), while maintaining a relatively balanced sample size among groups (36.0%, 27.0%, 15.5%, 21.5%) (Tables S3 and S4-1 to S4-3).

[0091] Table S3 Direct Comparison of Optimal 2-Group, 3-Group, and 4-Group Plans Abbreviations: OS, overall survival; n1 to n4, sample size of each group; Prop, proportion; Min prop, minimum proportion of each layer; P, log-rank P-value; C-index, Harrell consistency index; NA, not applicable.

[0092] Table S3 summarizes the optimal candidate schemes for each grouping framework (2, 3, and 4 groups) determined by exhaustive grid search. Each scheme is presented with Harrell's C-index, global log-rank p-value, and sample distribution (n1–n4, proportions). A minimum proportion threshold (≥0.15) was used to exclude splitting schemes with insufficient samples. The 4-group scheme (0.5 / 1.5 / 2.5) achieved the best discriminative power (C-index = 0.795) while maintaining good inter-group balance.

[0093] Table S4-1 All candidate schemes and feasibility markers for two-component stratification Table S4-2 All candidate schemes and feasibility markers for the three-part stratification

[0094] Table S4-3: All candidate schemes and feasibility markers for the four-group stratification

[0095] Table S4-4: All candidate schemes and feasibility markers for four-group stratification (continued from Table S4-3)

[0096]

[0097] Each table lists all tested cut-off point combinations, displaying the corresponding C-index, log-rank P-value, group size, proportion, and feasibility marker (“Yes / No”) based on a minimum proportion of ≥0.15 in each stratum. Candidate protocols were sorted according to a uniform selection rule. The selected configurations (0.5 / 1.5 / 2.5) were marked as the final 4-group EMD-RS stratification protocols. Therefore, extramedullary multiple myeloma patients were divided into 4 risk groups: low risk (0–0.5 points), low-intermediate risk (1–1.5 points), intermediate-high risk (2–2.5 points), and high risk (3–5.5 points), as shown in Table 3-1. Specifically: Extramedullary multiple myeloma patients were scored based on the presence of secondary EMD, EME, maximum lesion diameter ≥5 cm, del (17p), del (13q), and elevated LDH. The scores of these six independent adverse prognostic factors were summed to obtain the total EMD-RS score. Based on the total EMD-RS score, the prognosis of extramedullary multiple myeloma patients was divided into four levels, as shown in Table 3. The evaluation results of the constructed EMD-RS model are as follows: Figure 3 As shown in Figures A and B, the constructed EMD-RS model was validated using two validation queues, one with 40 EMMs and the other with 47 EMMs. The validation results are as follows. Figure 3 As shown in C to F, the validation cohort was further divided into two subgroups: patients who received autologous stem cell transplantation (ASCT) and those who did not. The results are shown in the figures below. Figure 3 As shown in G and H. The prognostic stratification results of the training set and the two batches of validation queues are shown in Table 3.

[0098] Table 3. Prognostic stratification of patients with extramedullary multiple myeloma The median overall survival for the low-risk, low-intermediate-risk, intermediate-high-risk, and high-risk groups was NR (95% confidence interval 60.24–NR), 45.75 months (95% confidence interval 35.62–55.98 months), 23 months (95% confidence interval 17.87–28.96 months), and 6.63 months (95% confidence interval 2.94–10.18 months), respectively. The median progression-free survival for the same groups was 56.5 months (95% confidence interval 43.89–69.23 months), 17.7 months (95% confidence interval 13.41–22.65 months), 11.18 months (95% confidence interval 7.66–15.24 months), and 3.27 months (95% confidence interval 2.42–4.17 months), respectively.

[0099] As shown in the table above, if the total cumulative score is 0 to 0.5, the prognostic stratification of patients with extramedullary multiple myeloma is classified as Grade I (low risk), with a predicted overall survival of 60.24-NR months and a predicted progression-free survival of 43.89-69.23 months. If the total cumulative score is 1 to 1.5, the prognostic stratification of patients with extramedullary multiple myeloma is classified as grade II (low to intermediate risk), with a predicted overall survival of 35.62 to 55.98 months and a predicted progression-free survival of 13.41 to 22.65 months. If the total cumulative score is 2 to 1.5 points, the prognostic stratification of patients with extramedullary multiple myeloma is classified as grade III (intermediate to high risk), with a predicted overall survival of 17.87 to 28.96 months and a predicted progression-free survival of 7.66 to 15.24 months. If the total cumulative score is 3 to 5.5, the prognostic stratification of patients with extramedullary multiple myeloma is classified as grade IV (high risk), with a predicted overall survival of 2.94 to 10.18 months and a predicted progression-free survival of 2.42 to 4.17 months.

[0100] Nested bootstrap resampling (1000 iterations) further confirmed the robustness of the hierarchical scheme. In 1000 iterations, the four schemes were selected 904 times, showing the lowest OOB IBS (0.1715) and the highest OOB average AUC (0.8558), with minimal optimism bias (ΔIBS = +0.0142; ΔAUC = −0.0201) (Table S5). These results support the reproducibility and generalization ability of the identified thresholds (0.5, 1.5, and 2.5), and therefore they were fixed for subsequent validation analyses. The AUCs obtained from time-dependent ROC analyses at 12, 24, and 36 months were 0.895, 0.858, and 0.838, respectively. Figure 4 The Brier scores for the models (A, B, C, and D) were 0.119, 0.220, and 0.350, respectively, indicating that the models have good time discrimination and calibration (Table S6).

[0101] Table S5. EMD-RS grouping stability and bias correction analysis based on nested bootstrap (1000 times) Abbreviations: IBS, Integrated Brier Score; OOB, Out-of-Bag; AUC, Area Under the ROC Curve; Δ, Difference Between the Training Set and the OOB Estimate; EMD-RS, Extramedullary Lesion Risk Score.

[0102] Each iteration includes an inner bootstrap (for cutpoint selection) and an outer out-of-bag (OOB) evaluation. The table reports the selection frequency, the average Integrated Brier Score (IBS) for the training and OOB sets, the average time-dependent AUC, and the optimism bias (Δ = training queue − OOB). The four models were repeatedly selected 904 times in 1000 runs, achieving the lowest OOB IBS and the highest OOB AUC, demonstrating their robustness.

[0103] Table S6 Time-dependent performance metrics of the four selected schemes (0.5 / 1.5 / 2.5) Abbreviations: AUC, Area Under the ROC Curve; Brier score, a measure of calibration error.

[0104] The time-dependent AUC and Brier score were calculated at 12, 24, and 36 months using the inverse probability weighting (IPCW) method. A higher AUC and a lower Brier score indicate that the model has excellent discriminative power and stable calibration over the time dimension. The overall Brier score (12–36 months) is approximately 0.2273.

[0105] Model evaluation of EMD-RS: In the training cohort (n = 200), EMD-RS stages I–IV included 72 (36.0%), 54 (27.0%), 31 (15.5%), and 43 (21.5%) patients, respectively. The median overall survival (OS) for stages I–IV was not reached (NR), 45.8 months, 23.0 months, and 6.6 months, respectively. The corresponding 1-year, 3-year, and 5-year OS rates were: Stage I 98.6%, 84.8%, and 76.2%; Stage II 92.6%, 62.3%, and 35.8%; Stage III 73.0%, 28.9%, and 12.9%; and Stage IV 37.2%, 7.5%, and 0%. The median progression-free survival (PFS) for stages I–IV were 56.5, 17.7, 11.2, and 3.3 months, respectively, with significant differences in survival curves (P < 0.001). Figure 4 (AB). The C-index of EMD-RS for OS and PFS is 0.795 and 0.740, respectively.

[0106] Model validation of EMD-RS: In the validation cohort (n = 87), stages I–IV included 38 (43.7%), 21 (24.1%), 11 (12.6%), and 17 (19.5%) patients, respectively. The median overall survival (OS) was 61.6, 58.3, 25.7, and 5.1 months, respectively, while the 1-year OS rate for stage I was 100% (no events observed). The median progression-free survival (PFS) for stages I–IV was 43.4, 22.7, 12.4, and 4.1 months, respectively.

[0107] The time-dependent AUCs of the OS at 12, 24, and 36 months were 0.895, 0.858, and 0.838 for the training queue, and 0.915, 0.951, and 0.936 for the validation queue. Figure 4(A and B). Calibration plots based on the equal frequency decimals show good agreement between predicted and observed probabilities at 12, 24, and 36 months in the training cohort. Calibration metrics show CITL values ​​ranging from −0.02 to 0.01 with a slope of 0.94–1.03 and a weighted MAE of 0.03–0.05, with most calibration points falling within the 95% confidence interval of the Kaplan-Meier estimate. In the validation cohort, model calibration remains acceptable, characterized by slightly negative CITL values ​​(−0.03 to −0.05), a slope between 0.87 and 0.96, and an MAE of 0.05–0.07. Figure 4 (C and D in the middle).

[0108] Example 2: Comparison of EMD-RS System with Existing Traditional Installment Systems Based on the training cohort and two independent validation cohorts, this embodiment compares the predictive performance of the EMD-RS model constructed in Example 1 with existing clinical prognostic models (ISS, R-ISS, and R2-ISS) for prognostic stratification of extramedullary multiple myeloma. The mean and maximum net benefit (NB) of EMD-RS compared to the traditional staging system are shown in Table S7, and the standardized net benefit (sNB) of EMD-RS and the traditional staging system in OS prediction is compared in Table S8. The decision curves (DCA) based on the training cohort and two independent validation cohorts are shown in Table S8. Figure 6 As shown.

[0109] Table S7 Decision Curve Analysis of Overall Survival (OS): EMD-RS vs. Traditional Staged Systems - Average and Maximum Net Benefit (NB)

[0110] Table S8 Comparison of Standardized Net Gain (sNB) of EMD-RS and Traditional Periodic Systems in OS Prediction

[0111] Decision curve analysis results show that, during the 12-month and 24-month OS follow-up periods of the training cohort, validation cohort 1, and validation cohort 2, the EMD-RS model consistently outperformed the corresponding traditional baseline staging systems (R-ISS, ISS, and R2-ISS) across all risk threshold ranges. In the training cohort, at the 12-month follow-up, the average net benefit of EMD-RS was 0.126, an improvement of 0.059 compared to R-ISS (0.067), with a maximum net benefit difference of 0.043. At the 24-month follow-up, the average net benefit of EMD-RS was 0.191, exceeding that of ISS (0.128) by 0.062, with a maximum net benefit difference of 0.015. The external validation cohort results showed stable performance. In validation cohort 1, the average net benefit of EMD-RS at 12 months was 0.210, a difference of 0.059 compared to ISS (0.151); at 24 months, the average net benefit difference compared to R-ISS reached 0.054, with a maximum net benefit advantage of 0.049. In validation cohort 2, the average net benefit difference between EMD-RS and R-ISS at 12 months was 0.055, with a maximum net benefit difference of 0.048; the advantage was most significant at 24 months compared to R2-ISS, with an average net benefit difference of 0.082 and a maximum net benefit difference of 0.025. Overall, the EMD-RS model demonstrated higher clinical net benefits across independent cohorts and different follow-up time points, indicating superior clinical application value, and the external validation results were robust.

[0112] Standardized net benefit analysis further validated the clinical benefit advantage of the model. The overall trend of the results was consistent with the original net benefit, but the numerical level was higher. At 12-month follow-up in the training cohort, the standardized net benefit of the EMD-RS model was 0.856, significantly higher than that of the ISS staging system (0.786), with the difference between the mean and maximum standardized net benefits both being 0.07. At 24-month follow-up, the sNB values ​​of both models were equal, with no difference. In validation cohort 1, at 12-month follow-up, the standardized net benefits of EMD-RS and the ISS staging system were completely consistent, with no difference in benefit. At 24-month follow-up, the sNB of EMD-RS was 0.911, higher than that of ISS (0.844), with a benefit difference of 0.067. The model's advantage was most pronounced in validation cohort 2, with a standardized net benefit of 0.937 at 12-month follow-up, an improvement of 0.097 compared to ISS (0.840), representing the largest benefit difference. At 24-month follow-up, EMD-RS (0.894) remained higher than ISS (0.873), with a benefit difference of 0.021. Overall, across all time points, the EMD-RS model demonstrated superior standardized net benefit for the vast majority of follow-up periods, showing comparable benefit to the traditional staging system only at a few time points, exhibiting no disadvantage. This further confirms that the model's clinical decision-making value is superior to the traditional staging system.

[0113] Figure 6Decision curve analysis (DCA) results showed that at 12 and 24 months, the net benefit (NB) of EMD-RS consistently outperformed the ISS, R-ISS, and R2-ISS scoring systems. It demonstrated a stable advantage in the training cohort, validation cohort 1, and validation cohort 2.

[0114] Furthermore, survival outcomes in extramedullary multiple myeloma (EMM) patients, defined by risk groups according to ISS, R-ISS, and R2-ISS, are as follows: Figure 7 As shown.

[0115] The Concordance Index (C-index) is a core indicator for evaluating the predictive ability of survival analysis models (such as the Cox proportional hazards model). It primarily measures the probability that the model's predictions match actual outcomes. Generally, a C-index of 0.5–0.6 indicates almost no predictive value, meaning poor predictive ability; 0.6–0.7 indicates some predictive ability, but weak, indicating low predictive ability; 0.7–0.8 indicates acceptable and practical discriminative ability, indicating good predictive ability; 0.8–0.9 indicates very strong discriminative ability, indicating excellent predictive ability; and 0.9–1.0 (very rare) indicates extremely strong predictive ability, indicating very excellent predictive ability. A model with a C-index above 0.7 is generally considered to have clinical value, and a C-index between 0.75 and 0.85 is considered ideal.

[0116] contrast Figure 3 and Figure 7 It can be seen that the C-index for OS prediction in existing traditional staging systems (ISS, R-ISS, and R2-ISS) are 0.587, 0.620, and 0.613, respectively; while the C-index for OS prediction in the EMD-RS staging system constructed in this invention is 0.795 (training queue), 0.789 (validation queue 1), and 0.796 (validation queue 2). This indicates that the EMD-RS constructed in this invention has a significantly better OS prediction capability for EMM than existing traditional staging systems (ISS, R-ISS, and R2-ISS).

[0117] In existing traditional staging systems (ISS, R-ISS, and R2-ISS), the C-index for PFS prediction is 0.576, 0.608, and 0.613, respectively; while the C-index for PFS prediction of the EMD-RS staging system constructed in this invention is 0.740 (training queue), 0.732 (validation queue 1), and 0.80 (validation queue 2). This indicates that the EMD-RS constructed in this invention also significantly outperforms existing traditional staging systems (ISS, R-ISS, and R2-ISS) in predicting PFS for EMM.

[0118] Furthermore, the EMD-RS staging system constructed in this invention has a C-index of 0.787 for predicting overall survival (OS) in patients who have received autologous stem cell transplantation (ASCT) and a C-index of 0.813 for patients who have not received ASCT, indicating that the EMD-RS staging system has good predictive ability for overall survival (OS) regardless of whether extramedullary multiple myeloma patients receive autologous stem cell transplantation (ASCT).

[0119] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A prognostic grading system for extramedullary multiple myeloma, characterized in that, It includes modules for acquiring and judging medical records, acquiring and judging images, acquiring and judging bioinformatics, and calculating and evaluating data. The medical record information acquisition and judgment module is used to acquire the medical record data and clinical course information of the patient with extramedullary multiple myeloma to be evaluated, determine whether the patient has secondary EMD, and submit the judgment result to the calculation and evaluation module. The image acquisition and judgment module is used to acquire whole-body PET / CT image data of patients with extramedullary multiple myeloma to be evaluated before treatment, identify and analyze the imaging characteristics of extramedullary lesions, determine whether the patient has the EME phenotype, and determine whether the maximum lesion diameter is ≥5cm based on the size of the largest lesion in the image; and submit the judgment result to the calculation and evaluation module. The bioinformatics acquisition and judgment module is used to acquire cytogenetic abnormality information of patients with extramedullary multiple myeloma to be evaluated, determine whether the patient has del(17p) and del(13q) abnormalities, acquire the LDH detection value when the patient has EMD, and determine whether LDH is elevated; and submit the judgment results to the calculation and evaluation module. The calculation and assessment module classifies extramedullary multiple myeloma prognostic risks into levels I to IV based on the principle that a higher total EMD-RS score indicates a higher prognostic risk level. The total EMD-RS score is calculated according to the following formula: ； In the formula The score represents the score corresponding to the i-th independent adverse prognostic factor, which includes secondary EMD, EME phenotype, maximum lesion diameter ≥5cm, del (17p), del (13q) and elevated LDH; The regression coefficients of the multivariate Cox model constructed from the independent adverse prognostic factors are obtained by scaling the coefficients proportionally, rounding them to the nearest 0.5, or by approximate aggregation; if a patient has the i-th independent adverse prognostic factor, then... The value is 1; if it does not exist, then... The value is 0.

2. The system as described in claim 1, characterized in that, The Standardized weights are assigned using a minimum scoring unit of 0.

5. The standardized weights are assigned by discretization according to fixed rules. Calculate using the following formula: ； In the formula, The standardized weight of the i-th independent adverse prognostic factor; Let be the regression coefficient of the i-th independent adverse prognostic factor; The regression coefficients of the independent adverse prognostic factors mentioned above, which serve as reference independent adverse prognostic factors, are scored according to the following method: Standardized weights ≥1.25, The assigned score is 1.5 points; 0.75 ≤ standardized weight <1.25, The score is 1 point; 0.25 ≤ Standardized weight <0.75, The score is 0.5 points; Standardized weights <0.25, The score is 0.

3. The system as described in claim 2, characterized in that, The reference independent adverse prognostic factor was del(17p) abnormality.

4. The system as described in claim 3, characterized in that, The The scores were assigned according to the following ratio: secondary EMD: EME phenotype: maximum lesion diameter ≥5cm: del(17p): del(13q): LDH elevation = 1.5:1:1:1:0.5:0.

5.

5. The system as described in claim 4, characterized in that, The Scoring will be conducted as follows: The first independent adverse prognostic factor was secondary EMD, with an S1 score of 1.

5. The second independent adverse prognostic factor was the EME phenotype, with an S2 score of 1. The third independent adverse prognostic factor was the largest lesion diameter ≥5cm, with an S3 score of 1 point; The fourth independent adverse prognostic factor was del (17p), and S4 was assigned a score of 1. The fifth independent adverse prognostic factor was del(13q), with an S5 score of 0.

5. The sixth independent adverse prognostic factor was elevated LDH, with an S6 score of 0.

5.

6. The system as described in claim 5, characterized in that, The calculation and assessment module assesses the patient's prognostic grading based on the calculated total cumulative EMD-RS score using the following method: If the patient's total cumulative EMD-RS score is 0-0.5, the prognostic grade is assessed as Grade I. If the patient's total cumulative EMD-RS score is 1 to 1.5, the prognostic grade is assessed as Grade II. If the patient's total cumulative EMD-RS score is 2 to 2.5, the prognostic grade is assessed as Grade III. If a patient's total cumulative EMD-RS score is 3 to 5.5, their prognostic grade is assessed as IV.

7. The system as described in claim 6, characterized in that, The calculation and assessment module also predicts overall patient survival based on prognostic grading: If the patient's prognostic classification is grade I, the predicted overall survival is 60.24-NR months; If the patient's prognostic grade is II, the predicted overall survival is 35.62-55.98 months; If the patient's prognostic grade is III, the predicted overall survival is 17.87-28.96 months; If a patient's prognostic grade is IV, their predicted overall survival is 2.94-10.18 months.

8. The system as described in claim 7, characterized in that, The calculation and evaluation module predicts the patient's overall survival according to the following method: If the patient's prognostic classification is grade I, the predicted median overall survival is not reached. If the patient's prognostic grade is II, the predicted median overall survival is 45.75 months; If the patient's prognostic grade is III, the predicted median overall survival is 23.00 months; If a patient's prognostic grade is IV, the predicted median overall survival is 6.63 months.

9. The system as described in claim 7 or 8, characterized in that, The calculation and assessment module also predicts progression-free survival based on prognostic grading: If the patient's prognostic grading is grade I, the predicted progression-free survival is 43.89–69.23 months; If the patient's prognostic grading is grade II, the predicted progression-free survival is 13.41–22.65 months; If the patient's prognostic grading is grade III, the predicted progression-free survival is 7.66-15.24 months; If a patient's prognostic grade is IV, their predicted progression-free survival is 2.42-4.17 months.

10. The system as described in claim 9, characterized in that, The calculation and evaluation module predicts according to the following method: If the patient's prognostic classification is grade I, the predicted median progression-free survival is 56.50 months. If the patient's prognostic grade is II, the predicted median progression-free survival is 17.7 months; If the patient's prognostic grading is grade III, the predicted median progression-free survival is 11.18 months; If a patient's prognostic grade is IV, the predicted median progression-free survival is 3.27 months.

Images