Applying additive hazards models for analyzing survival in patients with colorectal cancer in Fars province, Southern Iran

Background: Colorectal cancer (CRC) Considered as the most common fatal cancer and ranked as the third worldwide and the third and the fifth in Iranian women and men respectively. There are several methods for analyzing time to event data. Although Cox proportional hazards model is popular but when the absolute change risk instead risk ratio (hazard ratio), is of primary concern or when the proportionality assumption does not hold the additive hazard regression models may be more suitable. The main objective of this study was to apply additive hazards models beside Cox proportional hazards model to study of survival CRC patients.Materials and Methods: The data were used in this study includes 561 colorectal cancer patient information that is recorded in Shiraz Namazi hospital, Iran during the 2005 to 2010 years and they followed until December 2015. The nonparametric Aalen’s additive hazard model and semiparametric Lin and Ying’s additive hazard model besides Cox proportional hazards model were used for data analysis. The proportionality assumption for Cox model was evaluated by a test based on correlation between the Schoenfeld residuals and also for test goodness of fit in additive models ranked survival times beside graphical methods and Cox-Snell residual plots were used. The analysis was done with SAS 9.2 and R3.2 softwares.Results: The median follow-up time was 49 months. The five-year survival rate and the mean survival time after cancer diagnosis were 59.6% and 68.12±1.4 months respectively. The multivariate analysis using the Lin and Ying’s additive model and Cox proportional model identified that the age of diagnosis, site of tumor, stage, proportion of positive lymph nodes, lymphovascular invasion and type of treatment were observed as factors influencing survival of the CRC patients.Conclusion: When the interest is evaluating absolute change risk the additive models or the proportionality assumption is violated the additive models are a suitable alternative to Cox proportionality model.