Related Experiment Video
Updated: Sep 26, 2025
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
Predicting Bone Metastasis Risk Based on Skull Base Invasion in Locally Advanced Nasopharyngeal Carcinoma
Bo Wu1,2, Yu Guo2, Hai-Hua Yang3
1Department of Radiotherapy and Oncology, The Second Affiliated Hospital of Soochow University, Suzhou, China.
Objective:
To develop and validate a bone metastasis prediction model based on skull base invasion (SBI) in patients with locally advanced nasopharyngeal carcinoma (LA-NPC).
Methods:
This retrospective cohort study enrolled 290 patients with LA-NPC who received intensity-modulated radiation therapy in two hospitals from 2010 to 2020. Patient characteristics were grouped by SBI and hospital. Both unadjusted and multivariate-adjusted models were used to determine bone metastasis risk based on SBI status. Subgroup analysis was performed to investigate heterogeneity using a forest graph. Cox proportional hazard regression analysis was used to screen for risk factors of bone metastasis-free survival (BMFS). A nomogram of BMFS based on SBI was developed and validated using C-index, receiver operating characteristic curve, calibration curves, and decision curve analysis after Cox proportional hazard regression analysis.
Results:
The incidence of bone metastasis was 14.83% (43/290), 20.69% (24/116), and 10.92% (19/174) in the overall population, SBI-positive group, and SBI-negative group, respectively. In the unadjusted model, SBI was associated with reduced BMFS [HR 2.43 (1.32-4.47), P = 0.004], and the results remained stable after three continuous adjustments (P <0.05). No significant interaction was found in the subgroup analyses (P for interaction >0.05). According to Cox proportional hazard regression analysis and clinical value results, potential risk factors included SBI, Karnofsky performance status, TNM stage, induction chemotherapy, concurrent chemoradiotherapy, and adjuvant chemotherapy. Using a training C-index of 0.80 and a validation C-index of 0.79, the nomogram predicted BMFS and demonstrated satisfactory prognostic capability in 2, 3, and 5 years (area under curve: 83.7% vs. 79.6%, 81.7% vs. 88.2%, and 79.0% vs. 93.8%, respectively).
Conclusion:
Skull base invasion is a risk factor for bone metastasis in patients with LA-NPC. The SBI-based nomogram model can be used to predict bone metastasis and may assist in identifying LA-NPC patients at the highest risk of bone metastasis.
