Preservation of glomerular filtration rate on dialysis when adjusted for patient dropout

Background. Residual renal function (RRF) plays an important role in dialys is patients. Studies in patients on maintenance dialysis suggest that RRF i s better preserved in patients receiving peritoneal dialysis (PD) vis-a-vis those receiving hemodialy sis (HD). We speculated that regardless of the p atient's type of therapy, the estimate obtained for the rate of decline in glomerular filtration rate (GFR) may be biased because of informative censo ring associated with patient dropout. Informative censoring occurs when pat ients who die or transfer to another modality very early have associated wi th them a lower starting GFR or a higher rate of decline of GFR than patien ts who either complete the study or who die or transfer much later. If pati ent dropout is indeed related to the rate of decline in GFR and if this rel ationship is ignored in the analysis, then the estimate obtained of the rat e of decline in GFR may be biased. Methods. In an attempt to determine if there is a relationship between pati ent dropout and the decline in GFR, we reanalyzed the CANUSA data by modeli ng GFR as a nonlinear function of time with the rate of decline being expon ential. Results. This article highlights the significance of "informative censoring " when studying the decline of RRF on dialysis. The results show that for t he CANUSA cohort, the mean initial GFR was significantly lower, and the rat e of decline was significantly higher for patients who died or transferred to HD than for patients who were randomly censored or received a transplant . It is important to emphasize that the impact of informative censoring on previous analyses of the decline of RRF between PD versus HD is presently u nclear. If bias caused by informative censoring is the same regardless of w hat therapy a patient is on, then conclusions from previous studies compari ng the decline in GFR between PD and HD would still be valid. However, if t he magnitude of the bias differs according to therapy, then additional adju stments would be needed to fairly compare the decline in GFR between PD and HD. Because this analysis is restricted to patients on PD, it would be sci entifically incorrect to interpret previous studies solely on the basis of the results from this analysis. Conclusion. In any longitudinal study designed to estimate trends in an out come measured over time, it is important that the analysis of the data take s into account any effect patient dropout may have on the estimated trend. This analysis demonstrates that among PD patients, both the starting GFR an d the rate of decline in GFR are associated with patient dropout. Consequen tly, future studies aimed at estimating the rate of decline in GFR among PD patients should also account for any dependencies between dropout and GFR. Similarly, data analyzing for apparent differences in the rate of decline of GFR between PD and HD should also adjust for possible informative censor ing.