Study Of Competing Risks Analysis And Its Application In The Follow-up Study Of The Linxian General Population Nutrition Intervention Trial

Objectives:To evaluate commonly used methods for competing risks analysis systematically,explore the methods for assessing effects of competing events on the risk of the events of interest,compare the pros and cons of these methods,propose the general statistical strategies for handling the issues of competing risks and provide a theoretical basis for their application.Materials and methods:The association between the fresh fruits intake and the long-term risk of esophageal cancer death in the Linxian General Population Nutrition Intervention Trial cohort was estimated by the cause-specific hazard regression analysis and the sub-distribution hazard regression analysis for handling the competing risks issues,respectively.Then,the effect of competing events on the long-term risk of esophageal cancer death was assessed through the relative hazard function,and the baseline hazard function of the esophageal cancer death was also calculated based on the pseudo-observation approach.The advantages and disadvantages of these methods for competing risks analysis were simultaneously compared.Finally,the general strategies for the competing risks analysis can be proposed.Results:The relative hazard function and/or relative hazard ratio,the pseudo-observation approach and the sub-distribution hazard regression analysis can be respectively used to describe the effect of competing events on the risk of the event of interest,to estimate the baseline risk of the event of interest,and to associate the research factors to the events of disease more accurately when competing risks bias exists.In the Linxian General Population Nutrition Intervention cohort,the results from cause-specific hazard regression model showed that compared with subjects who never or rarely consumed fresh fruits,participants whose fresh fruits consumption more than once per week might had a lower long-term risk of esophageal cancer death,and this protective effect did not become stronger along with the increased frequency of fresh fruits consumption(P_Trend =0.444);the results based on the sub-distribution hazard regression model showed that the fresh fruits intake could not decrease the long-term risk of esophageal cancer death,indicating that the association between fresh fruit intake and the long-term risk of esophageal cancer death was exaggerated when competing risks bias appeared.The time-varying effects of competing events on the long-term risk of esophageal cancer death were observed by the relative hazard function,especially for the cardiovascular diseases deaths,and varied in different groups of the fresh fruit consumption.In addition,the baseline hazard function of esophageal cancer death was 1.39 in the Linxian General Population Trial cohort,and increased to 1.62 when adjusted for the frequency of the fresh fruits intake,which suggested that the fresh fruits consumption could not decrease the long-term risk of esophageal cancer death.Conclusions:In the follow-up study,the sub-distribution hazard regression model is recommended to evaluate the association between the research factors and diseases,and the relative hazard function and/or relative hazard ratio are recommended to assess the effects of competing risks to achieve the aim for improving the interpretability of the results.The fresh fruits intake might not decrease the long-term risk of esophageal cancer death among the Linxian General Population Nutrition Intervention Trial cohort.