Prognosis of dedifferentiated liposarcoma based on surveillance, epidemiology, and end results database Authors: Linna Zhu, Xianfei Duan, Liehua Deng Published: 2025 ABSTRACT AbstractIntroductionThis research aimed to develop and validate a novel nomogram for predicting overall survival (OS) in Dedifferentiated liposarcoma (DDLPS) patients.MethodsA retrospective cohort of 1,155 DDLPS cases diagnosed from 2004 to 2015 was identified through the Surveillance, Epidemiology, and End Results (SEER) registry. The prognostic model was subsequently developed by incorporating covariates independently associated with survival outcomes, as determined through multivariate analysis. To assess the model's predictive performance and clinical utility, we employed a suite of validation metrics.ResultsCox regression revealed age, tumor size, laterality, AJCC stage, combined summary stage, radiation, chemotherapy, and surgery as independent predictive variables for DDLPS. The C-index of the nomogram in the training and validation cohorts was 0.708 and 0.704, which was higher than that of the AJCC system. All AUC values、NRI (> 50%), IDI (> 10%), calibration and DCA curves showed that the new model had an excellent ability to predict OS.ConclusionsOur study represents the initial effort to develop a nomogram for predicting OS in DDLPS patients based on the SEER database. Significantly, the nomogram showed a greater potential for accurately predicting survival as opposed to the AJCC system.Keywords: Dedifferentiated liposarcoma, Nomogram, Overall survival, Prognosis, SEER Introduction Dedifferentiated liposarcoma (DDLPS) is a highly malignant subtype of liposarcoma, considered one of the most common types. It is believed to exist on a continuum with its corresponding well-differentiated liposarcoma, accounting for 40–45% of all liposarcomas. This disease typically affects middle-aged to elderly individuals, with no significant gender differences [1, 2]. DDLPS may originate from adipocytes present in subcutaneous or deep soft tissues. The primary site of DDLPS is retroperitoneal, followed by the extremities, pretesticular region, head and neck, and thorax, although it is rare in superficial soft tissues [3, 4]. The recent clinical data indicates that the incidence of DDLPS is rising, signifying that diagnostic and therapeutic challenges persist. DDLPS typically presents as a single deep-seated mass with slow and progressive growth [5]. The majority of DDLPS cases are sporadic, and the etiology and genetic mechanisms underlying the disease remain unclear. From a genomic perspective, deletions or duplications of certain genes contribute to DDLPS development. For instance, amplification of the CDK4 gene and other genes such as TP53, MDM2, CDKN2A, and CDKN2B are associated with the occurrence of DDLPS [6–8]. DDLPS has a poor prognosis, with a high potential for recurrence and metastasis. The latest statistics show a 5-year survival rate of only 30%. In fact, DDLPS is very susceptible to distant metastasis to the brain, liver, lungs, bone, skin, and soft tissues [9–11]. Local recurrence, observed in 40% of DDLPS patients, further complicates the situation. Full surgical resection is seen to be the best form of treatment for DDLPS, and routine follow-up and imaging examinations are typically required post-surgery to monitor for recurrence or metastasis. Adjuvant treatment measures, such as radiotherapy and chemotherapy can also be helpful. Recently, molecular targeted therapy and immunotherapy have emerged as new directions in DDLPS treatment, but they require more clinical trials to validate their efficacy and safety. Numerous factors affect the survival rate of DDLPS patients, including tumor size, location, depth, grading, staging, histological subtypes, pathological grading, surgical methods, radiotherapy, and chemotherapy, among others [12–14]. These factors make it challenging to predict survival accurately. Clinicians frequently face questions about the prognosis and chances of recovery for tumor patients, and their predictions significantly influence treatment, follow-up, and the patient's quality of life. The AJCC staging manual, which includes DDLPS, is a crucial resource for recommending treatment options and classifying patients for clinical research. However, its limitations include its dependence on tumor size and extent of spread, which may not fully capture the heterogeneity of DDLPS and its prognostic factors. Moreover, it does not consider other important clinical characteristics or treatment information that could impact patient outcomes, potentially leading to inaccurate survival predictions. Nomograms are visualization tools that integrate multiple predictive factors into a simple chart. By following the scale lines of each predictive factor on the chart, clinicians can directly read out the predicted outcome for a patient [15, 16]. Although nomograms have been applied to depict a variety of diseases, including breast cancer, lung cancer, and liver cancer [17, 18], the low incidence of DDLPS and small sample sizes have prevented the development of an established predictive model that accurately forecasts the overall survival of DDLPS patients. This study contributes a comprehensive nomogram by analyzing the Surveillance, Epidemiology, and End Results database, which includes demographic variables, clinical features, treatment modalities and other relevant characteristics, surpassing the AJCC staging system in clinical application for more accurately forecasting the OS of DDLPS patients. The developed nomogram can facilitate informed clinical decision-making, improve patient prognosis, and treatment strategies. Methods Patients Using SEER*Stat software (version 8.4.0.1), we retrieved de-identified clinical data of dedifferentiated liposarcoma (DDLPS) cases from the Surveillance, Epidemiology, and End Results (SEER) database (SEER Research Plus Data, 18 Registries, 2004–2015). Cases were identified using ICD-O-3 (International Classification of Diseases for Oncology, 3rd Edition) morphology code 8858/3, which specifies DDLPS histology and malignant behavior. The dataset encompassed comprehensive sociodemographic variables, clinicopathological characteristics, and survival outcomes. A predictive nomogram was constructed using a training cohort and subsequently validated in an independent validation cohort. This study utilized publicly available SEER data with no personal identifiers; thus, institutional review board approval and informed consent were waived in accordance with the database's access guidelines. Variable This study considered multiple factors that may affect the prognosis of DDLPS patients, including demographic variables such as age, gender, race, and marital status; clinical features, such as grade, AJCC stage (8th edition), location, tumor size, and combined summary stage; and treatment modalities such as radiation, chemotherapy, and surgery. In this study, the outcomes were survival status and time, where OS survival time was the duration between the initial diagnosis and death from any cause. Nomogram The cohort was randomly partitioned into training and validation subsets at a 7:3 ratio following confirmation of inter-group homogeneity (all P > 0.05 for baseline characteristic comparisons). We first identified the independent prognostic factors of DDLPS patients through multivariate Cox regression analysis, then used these factors to construct a new nomogram. Ultimately, a variety of validation techniques were employed in this work to ensure the construct's accuracy and reliability. Higher C-index and AUC values indicate better nomogram predictions, which are measured using these two metrics [19]. For a more thorough and precise comparison, Net reclassification index (NRI) and Integrated discrimination improvement (IDI) were compared. The accuracy of the nomogram's estimation of the survival probability was assessed using calibration plots. The calibration curve had an x axis that displayed expected values and a y axis that displayed actual values. If the ratio of actual to anticipated values went through the origin along a 45° line, the nomogram was deemed to be ideally calibrated; the closer it was to the 45° line, the better the calibration. DCA was used to evaluate the nomogram's potential for clinical application [20]. Statistical analysis Statistical computations were conducted using SPSS (v27.0; IBM Corp), with statistical significance defined as two-tailed P < 0.05. The predictive modeling workflow was executed in R (v4.2.2; R Foundation) through sequential analytical phases. Results Patient characteristics The study cohort comprised 1,155 patients with DDLPS(selected according to the flowchart in Figure 1), who were randomly allocated to training (70%, n = 809) and validation (30%, n = 346) cohorts using a 7:3 ratio. Table 1 summarizes the demographic and clinicopathological variables stratified by cohort. All variables demonstrated nonsignificant intergroup differences (p > 0.05) in comparative analysis, confirming balanced baseline characteristics between the training and validation cohorts. The mean age of patients was 65 years, with a predominance of male gender (65%), white race (85%), and married status (67%). Tumor size distribution was relatively uniform, with tumors primarily located in the retroperitoneum (44%) and overall soft tissue (55%). Approximately 70% of tumors showed no apparent laterality. Most tumors were locational (44%) and regional (45%). Tumor grade was predominantly III and IV (combined approximately 80%), although AJCC cancer stage was mainly stage III (60%). Surgical treatment was performed in the vast majority of cases (93%), but chemotherapy (82%) and radiotherapy (61%) were refused by most patients. Research flowchart. His/behave: SEER Histology/Behavior Code. Represents tumor type and malignancy grade Table 1Patient characteristics and sociodemographicVariablesTesting cohort (n = 808)Validation cohort (n = 347)PAge65 (56, 74)65 (57, 74)0.79Size, n (%)0.96 < 10 cm217 (27)93 (27) 10–20 cm304 (38)128 (37) > 20 cm287 (35)126 (36)Sex, n (%)0.66 Male523 (65)230 (66) Female285 (35)117 (34)Race, n (%)0.99 White683 (84)294 (85) Black48 (6)21 (6) Other77 (10)32 (9)Marital status, n (%)0.83 Married539 (67)231 (67) Single91 (11)43 (12) Other178 (22)73 (21)Site, n (%)0.63 RAPa355 (44)146 (42) CSAOSFb447 (55)197 (57) Other6 (1)4 (1)Laterality, n (%)0.57 Left121 (15)55 (16) Right119 (15)43 (12) No568 (70)249 (72)Grade, n (%)0.64 I69 (9)26 (7) II84 (10)44 (13) III308 (38)126 (36) IV347 (43)151 (44)AJCC, n (%)0.66 I146 (18)63 (18) II85 (10)44 (13) III491 (61)200 (58) IV86 (11)40 (11)Stage, n (%)0.26 Locational356 (44)167 (48) Regional361 (45)137 (40) Distant91 (11)43 (12)Radc, n (%)0.14 Yes315 (39)152 (44) No/Unknown493 (61)195 (56)Chemod, n (%)0.31 Yes145 (18)53 (15) No/Unknown663 (82)294 (85)Surge, n (%)0.93 Yes755(93)323 (93) No/Unknown53(7)24 (7)RAPa Retroperitoneum and peritoneum, CSAOSFb Connective, subcutaneous and other soft tissues, Radc Radiation, Chemod Chemotherapy, Surge Surgery Patient characteristics and sociodemographic RAPa Retroperitoneum and peritoneum, CSAOSFb Connective, subcutaneous and other soft tissues, Radc Radiation, Chemod Chemotherapy, Surge Surgery Identification of prognostic factors Multivariable and univariate Cox regression identified eight independent prognostic factors for overall survival (OS), as detailed in Tables 2 and 3. Significant predictors included age at diagnosis (HR = 1.03, 95% CI [1.02–1.04], p < 0.001), tumors > 20 cm (HR = 1.45, [1.12–1.88], p < 0.01), absence of tumor laterality (HR = 1.35, [1.02–1.79], p < 0.05), advanced AJCC stages (Stage III: HR = 2.69, [1.21–5.98]; Stage IV: HR = 2.59, [1.16–5.80], both p < 0.05), distant combined summary stage (HR = 2.08, [1.52–2.85], p < 0.001), and lack of therapeutic interventions (radiotherapy: HR = 1.35, [1.08–1.68], p < 0.01; chemotherapy: HR = 0.78, [0.62–0.98], p < 0.05; surgery: HR = 3.70, [2.64–5.18], p < 0.001). Figure 2 shows the cumulative risk curve of DDLPS patients. Table 2univariate analysis for overall survivalVariableHR_CIp-valueAge1 (1–1) < 0.001Tumor size < 10 cmReference 10–20 cm1.4 (1.1–1.7)0.009 > 20 cm1.7 (1.3–2.1) < 0.001Sex MaleReference Female0.98 (0.82–1.2)0.870Race WhiteReference Black0.88 (0.6–1.3)0.520 Other0.83 (0.62–1.1)0.230Marital status MarriedReference Single1 (0.79–1.4)0.760 Other1.2 (1–1.5)0.049Site RAPaReference CSAOSFb0.83 (0.7–0.99)0.042 Other0.56 (0.18–1.7)0.320Laterality LeftReference Right1.1 (0.74–1.5)0.740 No1.6 (1.2–2.1)0.000Grade IReference II0.87 (0.56–1.4)0.550 III1.5 (1.1–2.2)0.022 IV1.7 (1.2–2.4)0.004AJCC IReference II1.1 (0.74–1.6)0.650 III2 (1.5–2.6) < 0.001 IV4.4 (3.2–6.1) < 0.001Stage LocationalReference Regional1.3 (1.1–1.6)0.003 Distant3.6 (2.8–4.7) < 0.001Radc YesReference No/unknown1.5 (1.2–1.8) < 0.001Chemod YesReference No/unknown0.62 (0.5–0.77) < 0.001Surge YesReference No/Unknown5.8 (4.2–7.8) < 0.001 univariate analysis for overall survival Table 3Multivariate analysis for overall survivalVariableMultivariable analysisHR(95% CI)PAge1.03(1.02–1.04) < 0.001***Tumor size < 10 cmReference 10–20 cm1.11(0.86–1.42)0.438 > 20 cm1.45(1.11–1.90)0.006**Sex MaleReference Female0.85(0.7–1.04)0.111Race WhiteReference Black1.25(0.86–1.82)0.247 Other1.02(0.73–1.42)0.902Marital status MarriedReference Single1.17(0.87–1.58)0.306 Other1.03(0.82–1.30)0.782Site RAPaReference CSAOSFb1.17(0.95–1.44)0.144 Other0.30(0.04–2.14)0.228Laterality LeftReference Right1.10(0.76–1.60)0.617 No1.35(1.00–1.81)0.047*Grade IReference II1.01(0.61–1.68)0.966 III0.69(0.28–1.71)0.418 IV0.75(0.30–1.87)0.538AJCC IReference II1.75(0.62–4.94)0.292 III2.69(1.02–7.07)0.045* IV2.59(1.03–6.50)0.043*Stage LocationalReference Regional1.18(0.96–1.44)0.109 Distant2.08(1.43–3.01) < 0.001***Radc YesReference No/unknown1.35(1.10–1.65)0.004**Chemod YesReference No/unknown0.78(0.61–1.00)0.048*Surge YesReference No/unknown3.70(2.60–5.25) < 0.001***RAPa Retroperitoneum and peritoneum, CSAOSFb Connective, subcutaneous and other soft tissues, Radc Radiation, Chemod Chemotherapy, Surge Surgery ∗ p < 0.05; ∗ ∗ p < 0.01; ∗ ∗ ∗ p ≤ 0.001, HR Hazard ratio, CI Confidence interval Multivariate analysis for overall survival RAPa Retroperitoneum and peritoneum, CSAOSFb Connective, subcutaneous and other soft tissues, Radc Radiation, Chemod Chemotherapy, Surge Surgery ∗ p < 0.05; ∗ ∗ p < 0.01; ∗ ∗ ∗ p ≤ 0.001, HR Hazard ratio, CI Confidence interval The cumulative risk curve of DDLPS patients. A (Sex), B (Race), C (Marital status), D (Laterality), E (Size), F (Site), G (Stage), H (Grade), I (AJCC), J (Radiation), K (Chemotherapy), M (Surgery) Nomogram for OS prognosis The outcomes of multivariate Cox regression analysis were used to create a nomogram (Fig. 3) that forecasts the probabilities of OS at 3, 5, and 8-years for patients with DDLPS. Age had the greatest influence on prognosis, followed by surgery, AJCC stage, combined summary stage, and other variables. The scoring system assigns a score to each nomogram variable. A vertical line is then drawn downward from the overall score, which represents the probability of 3, 5, and 8-year OS, after summing the scores of all variables. Nomogram predicting 3-, 5-, and 8-year OS probability. AJCC American Joint Committee on Cancer, Stage Combined Summary Stage, Rad Radiotherapy Status, Chemo Chemotherapy Status, Surg Surgery Status Nomogram evaluation The novel nomogram demonstrated superior discriminative performance compared to the AJCC staging system, with concordance indices (C-index) of 0.708 (training cohort) and 0.704 (validation cohort), versus 0.625 and 0.602 for AJCC staging, respectively. Time-dependent receiver operating characteristic (ROC) analysis further validated the nomogram’s enhanced prognostic accuracy (Fig. 4). For 1-, 3-, and 5-year overall survival (OS) predictions, the nomogram achieved area under the curve (AUC) values of 0.761, 0.748, and 0.768 in the training cohort, outperforming AJCC staging (AUC: 0.667, 0.657, and 0.640). Similar superiority was observed in the validation cohort (nomogram AUC: 0.729, 0.737, 0.765 vs. AJCC AUC: 0.602, 0.623, 0.660). Receiver-operating characteristic curves. The AUC values in the ROC curve showed that the nomogram had higher accuracy than the AJCC staging system in predicting 3-, 5-, and 8-year survival of DDLPS patients in both the training (A–C) and validation cohorts (D–F) The nomogram demonstrated superior predictive accuracy over the AJCC staging system, as evidenced by net reclassification improvement (NRI) and integrated discrimination improvement (IDI) analyses. For 3-, 5-, and 8-year survival predictions, the training cohort showed NRIs of 53.78% (95% CI 0.454–0.679), 52.73% (0.420–0.690), and 62.53% (0.469–0.748), respectively, with corresponding validation cohort NRIs of 54.72% (0.280–0.752), 51.59% (0.259–0.739), and 56.24% (0.171–0.817). IDI analysis further confirmed the nomogram’s enhanced discrimination, with statistically significant improvements in both cohorts (training: 11.35%, 12.21%, 11.98%; validation: 10.14%, 10.96%, 10.72%; all p < 0.001). According to the calibration curves (Fig. 5), the uniform distribution of points on both sides of the reference line and their proximity to the line indicate that the model has good predictive performance and accuracy in the calibration curve. Figure 6 shows that DCA revealed net benefit from the nomogram prediction of survival probabilities as compared to the AJCC system, demonstrating the effectiveness of our established nomogram. Furthermore, the DCA revealed that both nomograms have good clinical utility. Calibration curves. This calibration curve graph depicts the concordance between projected probability and observed outcomes for 3-, 5-, and 8-year OS in both the training cohort (A–C) and validation cohort (D–F) Decision curve analysis curves. DCA for the training cohort (A–C) and validation cohort (D–F) 3-, 5-, and 8-year OS probabilities. The higher the net benefit value, the better the model's performance Discussion Previous studies have investigated the predictive factors affecting prognosis in patients with DDLPS, but the prognostic value of sociodemographic characteristics and clinicopathological features in DDLPS patients remains ambiguous. To address this information gap, we conducted a retrospective analysis of 1,155 DDLPS patients with follow-up data to identify risk factors for reduced OS in liposarcoma patients. Our findings indicated that advanced age, larger tumor size, bilateral distribution, higher AJCC stage, no surgery performed, distant metastasis, chemotherapy, and no radiation therapy are associated with poorer prognosis in DDLPS patients. A DDLPS nomogram was later created and validated to calculate 3-, 5-, and 8-year OS probability. The nomogram constructed in this study provides clinicians with an intuitive tool for individualized prediction of DDLPS prognosis. Although the current version is static, we plan to develop a dynamic nomogram tool in future work to further enhance its practical implementation and dissemination in clinical practice. Analysis revealed a significant positive association between increasing age and poor prognosis in DDLPS patients. Model estimates indicated that for each one-year increase in age, the hazard ratio increases by approximately 3%. This clearly demonstrates that older age is adversely associated with survival outcomes [21, 22]. Moreover, our model revealed that the higher the AJCC stage, the worse the prognosis of DDLPS patients. Gronchi et al. discovered that a larger tumor size constituted an adverse prognostic factor for liposarcoma patients [23], which aligns with our study. This observation may be attributed to the relationship between substantial tumor volume and elevated biological malignancy, encompassing regional invasiveness and metastatic potential. In our research, bilateral tumor distribution was identified as a prognostic predictor for low-grade liposarcoma. In our study, bilateral tumor distribution was an adverse prognostic factor for low-grade liposarcoma. Although the evidence in the literature is limited, the underlying mechanisms may involve both anatomical and molecular aspects [21]: Stojadinovic A et al. [24] indicate that bilateral lesions suggest more extensive retroperitoneal infiltration, increasing the difficulty of achieving complete surgical resection indicate currently, studies show that bilateral tumors reflect higher genomic instability (e.g., TP53 mutations, chromothripsis), and their immune microenvironment may promote local immune escape, thereby accelerating recurrence [25].We determined that patients with distant metastasis exhibited a worse prognosis, which is supported by studies conducted by Lahat G et al. [26] and Italiano A et al. [27], both identifying distant metastasis as an independent negative predictor for overall and progression-free survival in DDLPS patients. Existing literature has shown that surgical resection is typically the preferred method for treating DDLPS, particularly when the patient's tumor is in its early stages and is locally resectable [28]. This agrees with our study findings, which reveal that patients who undergo surgery experience significantly improved survival probabilities and prognoses. Radiotherapy and chemotherapy can be used to reduce residual tumors, control recurrence, and prevent distant metastasis, but their survival benefits remain highly controversial. As systematically reviewed by Gahvari et al. [12], DDLPS exhibits low sensitivity to conventional chemotherapy (objective response rate to monotherapy only 7.5%). Although the combination of doxorubicin and ifosfamide improves response rates (26% vs. 14%), it fails to improve OS (HR = 0.83, p = 0.076) although increasing grade 3–4 hematologic toxicity. The STRASS trial demonstrated that radiotherapy improves local control but does not translate into overall survival benefit [29]. In our nomogram, chemotherapy suggested potential adverse effects, likely attributable to limited efficacy of traditional regimens and associated immunotoxicity. Simultaneously, Jelena Casier et al. [24] revealed that efficacy evaluation of chemo-radiotherapy may be biased due to treatment selection bias: since these modalities are predominantly administered to advanced-stage/high-risk patients, the treatment group inherently exhibits poorer baseline prognosis. In addition, patients with poor performance status are often excluded from intensive chemotherapy protocols, resulting in the inclusion of healthier populations in the treatment cohort. As such, a comprehensive evaluation should be performed prior to chemotherapy, considering the patient's individual circumstances, tumor characteristics, and prognostic risks. The effectiveness of chemotherapy and radiotherapy for DDLPS remains uncertain; therefore, these treatment modalities should be carefully considered. We adopted a variety of metrics to assess the performance of the nomogram. This all—around evaluation improves the accuracy and cogency of our research. Both the C-index and AUC are widely utilized indicators for appraising the performance of predictive models. A higher C-index signifies superior predictive accuracy, with values ranging from 0.5 to 1. With values ranging from 0.5 to 1, the AUC values closer to 1 denote improved model performance and stronger discriminatory power. Our training and validation cohort both exhibited C-indices and AUCs exceeding 0.7, significantly surpassing the AJCC staging system and demonstrating the nomogram's exceptional predictive performance [30]. NRI and IDI are employed to determine whether a new predictive model demonstrates significantly improved performance compared to traditional models. Our new model's NRI values were consistently positive and exceeded 50%, showing that it works better than the AJCC method at improving patient classification. In terms of 3-year, 5-year, and 8-year survival probability, the novel model outperforms the AJCC staging approach by 11.35%, 12.21%, and 11.98% in the training cohort and10.14%, 10.96%, and 10.72% in the validation cohort, respectively, according to IDI values. Calibration curves serve as graphical tools for evaluating a model's predictive accuracy and calibration. Ideally, the number of observed values within each predicted probability interval should correspond to the number of predicted values, resulting in the calibration line aligning with the diagonal line. Our calibration curves closely approximate the ideal 45-degree line, further substantiating the nomogram's exceptional performance. DCA plots illustrate the difference in net benefit between using and not using the predictive model at various thresholds [31]. The net advantage of employing our new model consistently surpasses that of using the AJCC staging method or not using a predictive model in all DCA plots for the training and validation cohorts, demonstrating the new model's clinical applicability and usefulness for treatment decision making. Several methodological constraints warrant consideration. Primarily, the retrospective design utilizing population-level registry data introduces inherent selection biases; geographic/temporal heterogeneity in case ascertainment may constrain generalizability to contemporary treatment eras. Second, the clinical information in the SEER database has inherent limitations: although it encompasses fundamental variables such as tumor type and pathological grade, it lacks critical factors influencing DDLPS prognosis. These factors include hypoalbuminemia (a recognized adverse prognostic marker in metastatic Soft Tissue Sarcoma) [27], surgical margin status (where R1/R2 resection can reduce median OS to < 10 months), variations in individual radiotherapy dose and chemotherapy regimen intensity (which may explain heterogeneity in treatment effects), and molecular biomarkers (such as MDM2 amplification, which can reduce disease-specific survival) [8].Consequently, the analysis and predictive outcomes of the model remain primarily at the clinical phenotypic level, failing to delve into the impact mechanisms of these underlying molecular alterations on prognosis and treatment response. This limitation restricts the biological interpretability and clinical utility of the model. Future efforts should integrate multidimensional clinical and molecular data to enhance the precision of predictive models. Finally, the predictive model developed in this study has not yet undergone external validation in an independent cohort, which is a critical step for confirming its generalizability and clinical applicability. The lack of external validation represents a significant limitation of the current model. Future priorities include utilizing independent datasets from different institutions and geographical regions (e.g., our institutional database and the European Sarcoma Database-ESDB) to thoroughly evaluate the predictive performance and robustness of the model. Conclusion This multicenter prognostic study developed the first SEER-derived nomogram incorporating multidimensional clinicopathological variables from the Surveillance, Epidemiology, and End Results (SEER) registry. The risk stratification framework demonstrated superior discriminative accuracy (ΔC-index: + 0.18 vs + 0.15) over conventional AJCC staging, providing actionable stratification to guide therapeutic decision-making in precision oncology paradigms. Acknowledgements We gratefully acknowledge Mr. Jun Lü and Dr. Hai Yu of The First Affiliated Hospital of Jinan University for their valuable contributions. Specifically, we thank Mr. Lü for his instruction in data analysis and Dr. Yu for addressing our technical inquiries. Abbreviations Dedifferentiated liposarcoma Surveillance, epidemiology, and end results The concordance index Area under the time-dependent receiver-operating characteristic curve Net reclassification index Integrated discrimination improvement Decision-curve analysis American joint committee on cancer Overall survival Hazard ratio Confidence interval Linna Zhu was born in Anhui, China, in 1991. She received her Master’s degree in Dermatology in 2018. Currently, she is working at the First People’s Hospital of Zhaoqing City. Her research interests include skin tumors and big data analysis. Author contributions All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication. Funding This work was supported by the Zhaoqing First People's Hospital Institutional Research Fund Graduate Student Research Project [Grant Number YJJ-2023-01-09]. Data availability All data generated or analysed during this study are included in this published article [and its supplementary information files]. Declarations Ethics approval and consent to participate Not applicable. This study utilized pre-existing, de-identified data from open-access databases. All ethical oversight and informed consent procedures were conducted and fulfilled during the original data generation processes, in accordance with the ethical standards of the originating institutions. Informed consent was obtained from all individual participants included in the study. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement All data generated or analysed during this study are included in this published article [and its supplementary information files].