Achmad Riyan, Surya Nengsih

Application of the Cox Proportional Hazard Model on Survival Data of Multiple Myeloma Patients Using the R Application

Introduction

Application of the cox proportional hazard model on survival data of multiple myeloma patients using the r application. Analyze Multiple Myeloma patient survival data using Cox Proportional Hazard Model in R. Explores age, sex, and protein level impact. Valid model, but predictive power limited, needs enhancement.

Abstract

Multiple Myeloma is a type of blood cancer characterized by the proliferation of malignant plasma cells in the bone marrow and can affect the patient's survival. This study aims to analyze the influence of age, sex, and protein levels on patient survival time. Multiple Myeloma uses the Cox Proportional Hazards model. The data used came from 47 patients with variables of survival time, patient status (dead or alive), age, gender, and protein content. The analysis was carried out using R software. The model match test with the likelihood ratio test also showed insignificant results, but testing of the assumption of proportional hazards through residual Schoenfeld showed that all variables met the model's assumptions. Thus, the Cox PH model in this study is technically valid, but its predictive power is still limited, so further model development is recommended by increasing the amount of data or considering other more relevant variables.

Review

This study endeavors to apply the Cox Proportional Hazards (PH) model to analyze survival data in Multiple Myeloma (MM) patients, aiming to assess the influence of age, sex, and protein levels. Utilizing R software, the authors investigated a modest cohort of 47 patients, meticulously checking the crucial proportional hazards assumption through Schoenfeld residuals. The reported satisfaction of this assumption for all variables is a strong point, lending technical validity to the selected modeling approach and demonstrating a good understanding of Cox PH model diagnostics. However, the findings are significantly constrained by several factors. A primary concern is the stated insignificant result from the likelihood ratio test for model fit, which strongly suggests that the model with the chosen predictors (age, sex, and protein levels) may not offer a statistically significant improvement over a baseline model. This calls into question the explanatory power of the included covariates. Furthermore, the small sample size of 47 patients inherently limits the generalizability and robustness of the findings, potentially leading to underpowered analyses and difficulties in capturing the intricate multifactorial nature of MM prognosis. As the authors themselves note, this contributes to the model's limited predictive power. To advance this research, several key improvements are recommended. Most critically, expanding the sample size is imperative to achieve greater statistical power, enhance the reliability of parameter estimates, and improve the generalizability of the derived survival model. Additionally, the explicit recommendation to consider "other more relevant variables" should be pursued rigorously. Future iterations of this work would greatly benefit from incorporating established prognostic factors in MM, such as disease stage, specific cytogenetic abnormalities, response to treatment, or other relevant biochemical markers. Integrating such clinically pertinent variables is essential to develop a more robust, predictive, and clinically actionable Cox PH model for Multiple Myeloma patient survival.

Full Text

Comments

You need to be logged in to post a comment.