Prognostic Risk Factors for Cancer-Specific Bone Metastasis: A Registry-Based Analysis of 13,742 Patients 

Background Accurate survival prediction for patients with bone metastatic cancer remains challenging. Existing prognostic models frequently show poor external validity, primarily due to small sample sizes, single-center designs, and insufficient inclusion of pathological and molecular variables. Moreover, few studies have concentrated on the prognostic heterogeneity of bone metastasis (BM) across different cancers using large, standardized datasets within a cancer-specific manner. This retrospective, multicenter, registry-based cohort study was conducted to evaluate the prognostic significance of BM across multiple cancer types and to identify cancer-specific clinical factors associated with survival. Methods Baseline demographic and clinical characteristics of 13,742 patients with AJCC stage IV or TNM stage M1 metastatic cancer diagnosis were collected across 42 clinical studies registered in the cBioPortal for Cancer Genomics database. Overall survival (OS) following metastatic diagnosis was set as the primary outcome. Univariate analyses were conducted to identify potential prognostic risk factors mainly using the Kaplan–Meier, log-rank test, and non-parametric tests. Variables with p < 0.20 were included in multivariate Cox proportional hazards models for further validation. Multiple imputation and bootstrap were applied for the missing value process and validation. Results BM was associated with favorable outcomes compared with other metastatic sites in osteotropic cancers such as breast, prostate, and thyroid cancer, whereas it indicated a worse prognosis in hepatobiliary, uterine sarcoma, and colorectal cancer with low affinity to skeletal tissue. Among prognostic variables, no single metastatic site served as a universal adverse prognostic factor across all cancers. Poorly differentiated or undifferentiated histology independently correlated with reduced survival (HR = 1.249, p < 0.001). Age above 60 years was also associated with inferior survival (univariate analysis, p < 0.001), while the primary cancer type remained the most influential prognostic determinant (HR = 1.422–1.758, p < 0.001). Conclusions BM demonstrates cancer-specific and heterogeneous influences on survival. Population for survival prediction in traditional studies could be expanded within a cancer-specific framework. Among the included prognostic variables, primary cancer type, pathological differentiation, and age stratify outcomes significantly, highlighting the demand for pathology-integrated, cancer-specific prognostic models. Incorporation of standardized treatment and molecular variables is essential for improving model precision and clinical applicability in the future.