MO703: Nonlinear Association of Fluid Overload to Technique Failure in Peritoneal Dialysis? Application of a Cubic Spline Model

Abstract

Abstract BACKGROUND AND AIMS Severe volume overload in peritoneal dialysis (PD) patients is associated with an elevated risk of technique failure [including death and transfer to hemodialysis (HD)] compared to moderate volume overload, euvolemia and volume depletion (Vrtovsnik et al., 2021 Clin Kidney J). Severe volume overload is often defined by a single cut-off value. However, such dichotomization of continuous variables in analytical models has shortcomings, especially when a nonlinear association is likely to be present. The aim of the present analysis was to investigate whether the association of volume overload to technique failure was nonlinear in nature. METHOD The IPOD-PD study enrolled incident patients, on either continuous ambulatory PD or automated PD. Besides regular documentation of demographic, medical and laboratory data, body composition data from measurements employing the Body Composition Monitor (BCM, Fresenius Medical Care, Bad Homburg, Germany) were recorded. In contrast to the previously published analysis (Vrtovsnik et al., 2021 Clin Kidney J), a competing risk analysis (kidney transplantation as competing risk) with cubic spline analysis was applied to take the continuous nature of the parameter fluid overload into account. Of 1054 incident PD patients from the IPOD-PD patient cohort (Van Biesen et al., 2019 CJASN), after excluding those who terminated the study before month 6, 719 patients were included in the present analysis. The ‘starting point’ of the follow-up was month 6 on PD, and technique failure (composite endpoint of change to HD and death) in the next 12 months was recorded as the outcome. The association of relative fluid overload with technique failure was estimated using penalized cubic splines regression, with kidney transplantation as competing risk. The model was adjusted for same variables as in previous study, including age, percentage of PD patients in the center, and number of incident PD patients in the last 2 years in the center, which were derived from backward selection based on Akaike's information criteria. RESULTS The patients included in the present analysis (n = 719) were in average 58.4 ± 14.9 years old, and 57% were male. Mean relative fluid overload was 9.7% at baseline, and 8.0% at month 6. With the means of cubic spline analysis, the course of subdistributional hazard (SDH) over the range of relative fluid overload as measured in our patient cohort was illustrated. Whereas SDH was rather constant in the range of fluid overload representing euvolemia, it increased strongly at more elevated relative fluid overload levels, which could accordingly be considered as overhydration (Fig. 1). Of covariates included in the model, increasing age was significantly associated with an elevated risk of technique failure and being treated in a center with a high versus low proportion of PD patients (≥ versus < 56.5% of the patients on PD) was significantly associated with a lower risk of technique failure. CONCLUSION An analytical approach considering the continuous nature of the parameter fluid overload allows to visualize a differentiated nonlinear pattern of associations to technique failure.