[Poisson regression use in nephrology]

Poisson regression is a powerful tool for the analysis of incidence rates from cohort survival studies and facilitates simple, straightforward analyses of temporal patterns that may be difficult to assess with other methods. The Kaplan-Meier method, the logrank test, and the Cox model each have their respective parallels in grouped data analysis: instantaneous hazards, the hazard ratio for grouped data, and Poisson regression, grouped by intervals. This approach makes it possible to present the instantaneous speed of occurrence of events that may be more significant for clinicians and to consider more easily some constraints like parameters according to time (like time dependent variables or a time-dependent effect, neither of which are included in the conventional Cox model). However the application of Poisson regression requires that data on individual subjects be organized into event-time tables stratified by time and other factors of interest. This approach therefore requires the use of a large-scale database when small time intervals or many adjustment variables are necessary.