Cervical cancer has a high incidence of malignant tumors and a high mortality rate, with squamous cervical carcinoma (SCC) accounting for 80% of cases. A competing-risks model is recommended as being more feasible for evaluating the prognosis and guiding clinical practice in the future compared to Cox regression. Data originating from the Surveillance, epidemiology, and end results (SEER) database during 2004 to 2013 were analyzed. Univariate analysis with the cumulative incidence function was performed to assess the potential risk of each covariate. Significant covariates (P < .05) were extracted for inclusion in a Cox regression analysis and a competing-risks model that included a cause-specific (CS) hazard function model and a sub-distribution (SD) hazard function model. A total of 5591 SCC patients met the inclusion criteria. The three methods (Cox regression analysis, CS analysis, and SD analysis) showed that age, metastasis, American Joint Committee on Cancer stage, surgery, chemotherapy, radiation sequence with surgery, lymph node dissection, tumor size, and tumor grade were prognostic factors affecting survival in patients with SCC. In contrast, race and radiation status were prognostic factors affecting survival in the Cox regression and CS analysis, but the results were different in the SD analysis. Being separated, divorced, or widowed was an independent prognostic factor in the Cox regression analysis, but the results were different in the CS and SD analyses. A competing-risks model was used as a new statistical method to more accurately identify prognostic factors than conventional Cox regression analysis leading to bias in the results. This study found that the SD model may be better suited to estimate the clinical prognosis of a patient, and that the results of an SD model analysis were close to those of a CS analysis.
Copyright © 2022 the Author(s). Published by Wolters Kluwer Health, Inc.