Prognostic significance of lab data and performance comparison by validating survival prediction models for patients with spinal metastases after radiotherapy
Hung-Kuan Yen,Ming-Hsiao Hu,Hester Zijlstra,Hsiang-Chieh Hsieh,Jiun-Jen Yang,Po-Chao Chen,Yu-Han Chen,Po-Hao Huang,Yu-Hung Chen,Fu-Ren Xiao,Jorrit-Jan Verlaan,Rong-Sen Yang,Shu-Hua Yang,Wei-Hsin Lin,Feng-Ming Hsu,Olivier Q. Groot,Aditya V. Karhade,Joseph H. Schwab
DOI: https://doi.org/10.1016/j.radonc.2022.08.029
IF: 6.901
2022-10-01
Radiotherapy and Oncology
Abstract:Background and purposeWell-performing survival prediction models (SPMs) help patients and healthcare professionals to choose treatment aligning with prognosis. This retrospective study aims to investigate the prognostic impacts of laboratory data and to compare the performances of Metastases location, Elderly, Tumor primary, Sex, Sickness/comorbidity, and Site of radiotherapy (METSSS) model, New England Spinal Metastasis Score (NESMS), and Skeletal Oncology Research Group machine learning algorithm (SORG-MLA) for spinal metastases (SM).Materials and methodsFrom 2010 to 2018, patients who received radiotherapy (RT) for SM at a tertiary center were enrolled and the data were retrospectively collected. Multivariate logistic and Cox-proportional-hazard regression analyses were used to assess the association between laboratory values and survival. The area under receiver-operating characteristics curve (AUROC), calibration analysis, Brier score, and decision curve analysis were used to evaluate the performance of SPMs.ResultsA total of 2786 patients were included for analysis. The 90-day and 1-year survival rates after RT were 70.4% and 35.7%, respectively. Higher albumin, hemoglobin, or lymphocyte count were associated with better survival, while higher alkaline phosphatase, white blood cell count, neutrophil count, neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, or international normalized ratio were associated with poor prognosis. SORG-MLA has the best discrimination (AUROC 90-day, 0.78; 1-year 0.76), best calibrations, and the lowest Brier score (90-day 0.16; 1-year 0.18). The decision curve of SORG-MLA is above the other two competing models with threshold probabilities from 0.1 to 0.8.ConclusionLaboratory data are of prognostic significance in survival prediction after RT for SM. Machine learning-based model SORG-MLA outperforms statistical regression-based model METSSS model and NESMS in survival predictions.
oncology,radiology, nuclear medicine & medical imaging