AUTHOR=Dong Shengtao , Yang Hua , Tang Zhi-Ri , Ke Yuqi , Wang Haosheng , Li Wenle , Tian Kang TITLE=Development and Validation of a Predictive Model to Evaluate the Risk of Bone Metastasis in Kidney Cancer JOURNAL=Frontiers in Oncology VOLUME=11 YEAR=2021 URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2021.731905 DOI=10.3389/fonc.2021.731905 ISSN=2234-943X ABSTRACT=Background

Bone is a common target of metastasis in kidney cancer, and accurately predicting the risk of bone metastases (BMs) facilitates risk stratification and precision medicine in kidney cancer.

Methods

Patients diagnosed with kidney cancer were extracted from the Surveillance, Epidemiology, and End Results (SEER) database to comprise the training group from 2010 to 2017, and the validation group was drawn from our academic medical center. Univariate and multivariate logistic regression analyses explored the statistical relationships between the included variables and BM. Statistically significant risk factors were applied to develop a nomogram. Calibration plots, receiver operating characteristic (ROC) curves, probability density functions (PDF), and clinical utility curves (CUC) were used to verify the predictive performance. Kaplan-Meier (KM) curves demonstrated survival differences between two subgroups of kidney cancer with and without BMs. A convenient web calculator was provided for users via “shiny” package.

Results

A total of 43,503 patients were recruited in this study, of which 42,650 were training group cases and 853 validation group cases. The variables included in the nomogram were sex, pathological grade, T-stage, N-stage, sequence number, brain metastases, liver metastasis, pulmonary metastasis, histological type, primary site, and laterality. The calibration plots confirmed good agreement between the prediction model and the actual results. The area under the curve (AUC) values in the training and validation groups were 0.952 (95% CI, 0.950–0.954) and 0.836 (95% CI, 0.809–0.860), respectively. Based on CUC, we recommend a threshold probability of 5% to guide the diagnosis of BMs.

Conclusions

The comprehensive predictive tool consisting of nomogram and web calculator contributes to risk stratification which helped clinicians identify high-risk cases and provide personalized treatment options.