Predicting Epithelial Ovarian Cancer first recurrence with Random Survival Forest: Comparison Parametric, Semi-Parametric, and Random Survival Forest Methods
Objective: Rapid technological advances in the last century and the large amount of information have made it difficult to analyze a large number of independent variables. In such circumstances, the existence of interactions of different degrees in the model is expected, in this case, the Cox model cannot be useful and the nonparametric method of random survival forest can be a useful alternative. This study compares the prediction error of random survival forest with Cox and Weibull models in predicting the time to the first recurrence in patients with epithelial ovarian cancer.
Method: In this retrospective study, the records of patients with epithelial ovarian cancer who referred to Imam Hossein Hospital in Tehran from 2007 to 2018 were used. To investigate the factors affecting the first recurrence of these patients, RSF was fitted to the data. Finally, prediction error of Cox, Weibull and RSF were compared using C-Index and Brier score.
Results: Brier score was calculated 0.16 for RSF, and 0.24 for Cox, also C-Index was calculated 0.34 for RSF and 0.42 for Cox. Brier score was calculated 0.092 for Cox and 0.089 for Weibull, so the prediction error of RSF was lower than both Cox and Weibull models.
Conclusion: Random survival forest with a suitable fit on many variables and without the need for a special default with a prediction error less than the Weibull and Cox methods can predict the response variable when confronted with high-dimensional data.
2. Clarke CL, Kushi LH, Chubak J, Pawloski PA, Bulkley JE, Epstein MM, et al. Predictors of long-term survival among high-grade serous ovarian cancer patients. Cancer Epidemiology and Prevention Biomarkers. 2019;28(5):996-9.
3. Sun Y, Liu S, Feng Z, Cheng J, Lu L, Wang M, et al. Preoperative omental metastasis-related maximum standardized fluorine-18-fluorodeoxyglucose uptake value can predict chemosensitivity and recurrence in advanced high-grade serous ovarian cancer patients. Nuclear medicine communications. 2018;39(8):761-7.
4. Li Z, Hong N, Robertson M, Wang C, Jiang G. Preoperative red cell distribution width and neutrophil-to-lymphocyte ratio predict survival in patients with epithelial ovarian cancer. Scientific reports. 2017;7:43001.
5. Myte R. Covariate Selection for Colorectal Cancer Survival Data: A comparison case study between Random Survival Forests and the Cox Proportional-Hazards model. 2013.
6. Lipson R. Predicting Ovarian Cancer Survival Times: Performance of Parametric Methods and Random Survival Forests. 2014.
7. Liu V. Predicting ovarian cancer survival times: Feature selection and performance of parametric, semi-parametric, and random survival forest methods. 2019.
8. Weathers B. Comparision of Survival Curves Between Cox Proportional Hazards, Random Forests, and Conditional Inference Forests in Survival Analysis. 2017.
9. Dietrich S. Investigation of the machine learning method Random Survival Forest as an exploratory analysis tool for the identification of variables associated with disease risks in complex survival data. 2016.
10. Roshanaei G. Determining affected factors on survival of kidney transplant in living donor patients using a random survivalforest. Koomesh. 2018;20(3):523-17.
11. Mohebbi M. Application of random survival forest model in prediction of the first metastasis in breast cancer patients and comparison with cox regression analysis. 2015.
|Issue||Vol 6 No 4 (2020)|
|Epithelial ovarian cancer; Cox; Weibull; First recurrent; Random Survival|
|Rights and permissions|
|This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.|