A narrative review of survival analysis in oncology using R
Akash Pawar, Oindrila R. Chowdhury, Omkar Salvi
Abstract
ABSTRACT Clinical trials in oncology largely include survival analysis, that is, the time to the occurrence of an event of interest as the primary outcome of the study. Survival analysis can be performed using non-parametric methods. In this review, we have provided the details of the statistical model for the probability of an individual surviving until time t . We have also discussed the use of the Cox proportional model to obtain the hazard or risk of occurrence of an event at a certain time while enrolled in the study. Hazard ratios and their confidence intervals are discussed and interpreted. We have also discussed the reporting of various survival endpoints, including the median recurrence-free and overall survival (in days and months), and the five-year recurrence-free and overall survival probability with 95% confidence intervals (CIs). We have discussed the P values obtained through the comparison of the survivals between various groups of patients and the methods of graphically representing the survival curves. To prepare this review, we searched the internet using Google for packages available for survival analysis on the Comprehensive R Archive Network (CRAN). We selected the Survival package and used the latest version 3.4-0. As an example, to demonstrate the methodology and applications, we used the rotterdam data set on primary breast cancer patients available in the Survival package, and the survminer package version 0.4.9 to visualize the survival curves. We have thus attempted to provide an easily understandable process for analyzing survival data using RStudio in the oncology setup. The results obtained by the listed procedure are accurate and highly dependable.