How should we analyze and present mortality in our patients?: A multicentre GCDP experience.

ABSTRACT

INTRODUCTION: There are different strategies to analyse mortality in peritoneal dialysis (PD) with different definitions for case, event, time at risk, and statistical tests. A common method for the different registries would enable proper comparison to better understand the actual differences in mortality of our patients. METHODS: We review and describe the analysis strategies of regional, national and international registries. We include actuarial survival, Kaplan-Meier (KM) and competitive risk (CR) analyses. We apply different approaches to the same database (GCDP), which show apparent differences with each method. RESULTS: A total of 1,890 incident patients in PD from 2003-2013 were included (55 years; men 64.2%), with initial RRF of 7ml/min; 25% had diabetes and a Charlson index of 3 [2-4]; 261 patients died, 380 changed to haemodialysis (HD) and 682 received a transplant. Annual mortality rates varied up to 20% in relative numbers (6.4 vs. 5.2%) depending on the system applied. The estimated probability of mortality measured by CR progressively differs from the KM over the years 3.6 vs. 4.0% the first year, then 9.0 vs. 11.9%, 15.6 vs. 28.3%, and 18.5 vs. 43.3% the following years. CONCLUSIONS: Although each method may be correct in themselves and express different approaches, the final impression left on the reader is a number that under/overestimates mortality. The CR model better expresses the reality of PD, where the number of patients lost to follow-up (transplant, transfer to HD) it is 4 times more than deceased patients and only a quarter remain on PD at the end of follow up.