Dynamic survival prognosis and the role of radiotherapy in retroperitoneal liposarcoma

Retroperitoneal liposarcomas (RPLSs) are rare and highly heterogeneous tumors, presenting unique clinical challenges. Despite surgical resection being the primary treatment, achieving complete resection with negative margins is difficult, and the role of adjuvant therapies, such as radiotherapy, remains unclear. The data for this study was sourced from the SEER (Surveillance, Epidemiology, and End Results) database. Conditional survival (CS) analysis was performed to evaluate survival outcomes over time. Machine learning techniques, including Random Survival Forest (RSF) and Least Absolute Shrinkage and Selection Operator (LASSO) regression, were used to identify prognostic factors and develop a CS-based nomogram. Propensity score matching (PSM) was employed to assess the survival benefit of radiotherapy. A total of 1,186 patients were analyzed (830 training, 356 validation). The CS analysis demonstrated improved survival probabilities with extended survival time; the probability of reaching 10 years rose from 41% at diagnosis to 87% after surviving 8 years. Based on machine learning and rigorous variable selection, the optimal prognostic indicator combination consists of age, histology, tumor grade, tumor size, SEER stage, AJCC stage and surgery. Based on the selected variable combination, the CS-based nomogram successfully predicted 3-, 5-, and 10-year overall survival and 10-year CS. The model was validated through calibration curves, ROC curve analysis, and decision curve analysis, and effectively stratified the cohort into risk groups. After propensity-score matching, radiotherapy was not associated with overall survival (HR 1.06, 95% CI 0.84–1.34; P = 0.62). CS analysis provided valuable insights into the evolving survival outcomes of RPLS, offering a more accurate prognostic tool for clinicians. Radiotherapy did not show a significant survival benefit, after adjustment for measured confounders; given the likelihood of residual confounding by indication, its role in RPLS treatment warrants further prospective investigation.