Investigating the Association Between a Continuous Variable and a Time-To-Event Outcome: Going Beyond the Cut-Off...

When describing relationships between variables and an outcome, dichotomization of continuous variables remains a widely used approach in medical research despite many drawbacks: the loss of information which reduces the statistical power to detect an association, the risk of misclassification and the problem of comparability of the results. Alternative approaches based on flexible functions are available and would allow to use all the information contained in the data and thus to model the possible non-linear relation between the continuous variable and the outcome. But these alternative approaches are rarely used probably because of a lack of clear guidance. This article aimed to illustrate the use of splines through an example based on hematological study. We showed the information provided by the plot of survival probabilities at a pre-specified time and according to the level of a continuous variable, thus displaying the trends of the studied phenomenon, as compared to a simple cut-off approach. In view of the major issues surrounding the patients' health, it is more than necessary to use the most powerful statistical approaches for greater precision in the understanding of health phenomena so that we may make more informed decisions. We hope this article will encourage the use of these approaches.