[Bias in survival analysis: the specific case of end-stage renal disease patients]

Main systematic errors that may occur in the design or in the conduction of a clinical research are selection bias, information bias, and confusion bias. In survival analysis, one of the selection biases is the loss-to-follow-up bias that occurs in subjects lost to follow-up before occurrence of the event of interest, i.e. death. These subjects are censored in analysis. In end-stage renal disease patients, this can occur at dialysis modality switch or at renal transplantation. Modifying selection criteria (incidence/prevalence bias, sampling of the cohort) or modifying censoring management, especially for renal transplant patients, may distort results in this population. These choices in analysis plan should be taken into account when interpreting survival analysis results both in observational study and in randomised trial among these renal patients.