Modeling and analyzing predictive monthly survival in females diagnosed with gynecological cancers

Abstract

Cancer ranks as a leading cause of death worldwide; an estimated 1.7 million new diagnoses were reported in 2021. Ovarian cancer, the most lethal of gynecological malignancies, has no effective screening with over 70% of patients being diagnosed in an advanced stage. The aim of this study was to determine the most statistically significant contributing factors through a multivariate regression into the severity of female gynecological cancers. Data from the surveillance, epidemiology, and end results program (SEER) cancer database were utilized in this study. Several attempted multivariate linear regressions were implemented with further reduced models; however, a linear model could not be properly fit to the data. Because of unmet assumptions, a nonparametric moving, local regression, locally estimated scatterplot smoothing (LOESS), was performed. After smoothing factors were included to reduced-models, residual information was minimized although few conclusions can be drawn from the resulting statistics. These issues were prevalent mainly because of the massive variability in the data and inherent lack of linearity. This can be a significant issue with clinical data that does not dive deeper into cancer-dependent factors including genetic expression and cell surface receptor overexpression. General patient demographic data and diagnostic information alone does not provide enough detail to make a definite conclusion or prediction on patient survivability. Increased attention to the acquisition of tumor tissue for genomic and proteomic analysis in addition to next-generation sequencing methods can lead to significant improvements in prognostic predictions.