Monitoring Daily Ultrafiltration in Automated Peritoneal Dialysis

Efficacy of peritoneal dialysis (PD) is often time-limited by accompanying chronic damage to the peritoneal membrane (1). High-quality, goal-directed dialysis is characterized by adequacy of peritoneal ultrafiltration (UF) and small solute clearance while maintaining fluid and salt homeostasis (2). Occasional measurements of clinical parameters of peritoneal membrane dysfunction in individual PD patients are highly variable. Inspired by time-series analyses such as Holter electrocardiography (3), we studied automated PD (APD) cycler readouts on daily inflows and outflows of PD fluid dwells to investigate the value of continuous UF monitoring under clinically relevant conditions.

We analyzed data from 77,138 patient days representing 129 APD patients (79 male, 50 female [61%:39%], aged mean 55±13 SD years). Patients were observed for 598±517 days with incident (first APD cycler UF measurement within first 90 days of PD):prevalent cases being n=99:30 (77%:23%), of PD vintage 154±334 days, diabetic:nondiabetic cases being n=32:97 (25%:75%), extending from initial APD treatment until kidney transplantation (n=55, 43%), transfer to hemodialysis (HD) being n=29 (23%), death being n=35 (27%), kidney function recovery being n=1 (0.8%), and loss to follow-up being n=9 (7%; n=7 transferred to other centers, n=2 data unavailable >6 months before PD termination). Ethics Committee approval was obtained (EK 2035/2015).

Available cycler measurement data were retrospectively extracted from PD link cycler card software of APD patients using Baxter Homechoice Pro cyclers (Baxter, Deerfield, IL, USA) at our single center from January 2000 to July 2019. Data for each treatment day and patient were analyzed using R software (R Core Team 2020, https://www.R-project.org/). Time-dependent changes in UF and glucose-corrected UF (=daily UF [ml]/daily glucose load [g]) were modeled utilizing mixed models for repeated measurements with time as covariate and patient as random factor, allowing random slopes and intercepts as functions of time on PD, with an autoregressive covariance structure. Individual 30-day mean UFs were calculated for the 6 months pre-endpoints and before and after peritonitis episodes. Thirty-day UF data encompassing peritonitis episodes were expressed as percentage of values before the occurring event. Differences were analyzed by Wilcoxon signed rank test. Differences in distributions of patients destined for kidney transplantation and for transfer to HD were calculated below and above the median decrement using Fisher's Exact test.

Patients were treated with Dianeal (low pH lactate-buffered high-glucose degradation products [GDP] solution, Baxter; n=76, 59%) or Physioneal (neutral pH bicarbonate/lactate-buffered low-GDP solution, Baxter; n=68, 53%) based on availability, and experienced 0.3 (±0.7 SD) peritonitis episodes per patient-year. A total of 17% of patients (n=22) received more than one glucose-containing PD solution during the observation period. Cycler ultrafiltration was 621±447 ml/d achieved by glucose loads of 236±80 g/day (1.7±0.4%, total cycler fill volume 13,912±2852 ml excluding last bag). Mean time on cycler was 9.2±0.8 h/d with 11±3 cycles/nightly treatment, 1213±322 ml/cycle tidal volumes and 6.0±0.8 h/d dwell times (excluding inflow/outflow times), yielding glucose-corrected UF of 2.5±1.5 ml/g per day.

These data allowed monitoring of time-dependent changes in peritoneal membrane function in mixed models for repeated measurements, considering intrapatient and interpatient variability, missing data, different APD starting points (not all patients initiated PD with APD), dropouts by endpoint achievement, and temporal intrapatient autocorrelation (Figure 1). Slopes of glucose-corrected UF over time increased in the low-GDP dialysate group but remained steady in the high-GDP group over time on PD, with significant difference between dialysates, consistent with previous selective measurements (4). Upon adjustment for residual kidney function, UF and glucose-corrected UF increments over time lost significance and steepness. Our study presents real-world data for peritonitis-associated membrane dysfunction, as mean glucose-corrected UF significantly declined in the 30 days post-peritonitis, with more severe changes in patients later transferred to HD. Moreover, glucose-corrected UF on APD during 6 months pre-endpoint decreased to a greater extent in patients later transferred to HD than in patients ultimately transplanted, suggesting higher prevalence of peritoneal membrane damage among the former.

Figure 1.

Daily ultrafiltration in automated peritoneal dialysis. (A) Reconstruction of automated peritoneal dialysis cycler readouts with accompanying peritoneal dialysis prescriptions for one representative patient. The green line displays daily cycler ultrafiltration (UF) measurements (ml/day), the blue line displays daily cycler glucose loads (g/day), and the red vertical line marks a peritonitis episode. (B) Ultrafiltration trends (ml/day) were modeled over time on peritoneal dialysis (PD) with mixed models for repeated measurements for all patients represented by the black line and density of measurement points by color ranging from purple (low) to yellow (high density). UF (ml/day) and glucose-corrected UF (ml/g/day) significantly increased over time, with respective slopes of 0.25 (95% confidence interval [95% CI], 0.07 to 0.43) and 0.0007 (95% CI, 0.00005 to 0.0014). Upon adjustment for residual kidney function, UF and glucose-corrected UF increments over time decreased and were no longer significant (0.09 [95% CI, −0.07 to 0.26], −0.0003 [95% CI, −0.0009 to 0.0003]). (C) Glucose-corrected UF (ml/g/day) is represented by the purple dashed line for low–glucose degradation products (GDP) dialysate and by the solid orange line for high-GDP dialysate (regression lines truncated at 3 years). In the low-GDP group, UF (ml/day) and glucose-corrected UF (ml/g/day) significantly increased over time with respective slopes of 0.45 (95% CI, 0.17 to 0.73) and 0.002 (0.001 to 0.003). In contrast, in the high-GDP group, UF (ml/day) and glucose-corrected UF (ml/g/day) did not significantly change over time, with respective slopes of 0.14 (95% CI, −0.08 to 0.37) and 0.00008 (−0.0009 to 0.001). The interaction estimates of the joint model, implicating a difference between the two groups, was significant for glucose-corrected UF with an estimate of 0.002 (95% CI, 0.0005 to 0.004) in contrast to UF with an estimate of 0.31 (95% CI, −0.05 to 0.66). (D) Glucose-corrected UF before (median [interquartile range] 2.03 [95% CI, 0.93 to 3.3]) and after each peritonitis episode (median [interquartile range] 1.95 [95% CI, 0.82 to 3.06]). Median 30-day UF and glucose-corrected UF significantly decreased after peritonitis (black dashed line) by −20% (95% CI, −51% to +3%). Differences in distributions of patients destined for kidney transplantation (green, left panel) and those destined for transfer to hemodialysis (red, right panel) below and above the median decrement (= black dashed line) were significant (P<0.05). (E) Changes of glucose-corrected UF during the final 6 months of automated peritoneal dialysis. Glucose-corrected UF significantly decreased in patients later transferred to hemodialysis (red lines and symbols, right panel, slope −0.007 [95% CI −0.01 to −0.002]) in contrast to glucose-corrected UF in patients eventually transplanted (green lines and symbols, left panel, slope −0.003 [95% CI, −0.008 to 0.001]).

Daily peritoneal UF monitoring by continuous APD data differs from traditional clinical data in quantity, noise, and missing normative values. Such analysis must consider unique patient trajectories, intra-individual variability, data noise and temporal autocorrelation. Our study demonstrated that continuously collected data allow analysis of unexpected, otherwise undetected PD-related events and their effects on UF. Our work is limited by absence of icodextrin-use analysis, which might mitigate falling UF, restricting interpretation of individual patient UF changes, and by inability to define causes of reduced UF.

In conclusion, this retrospective analysis of continuous APD cycler readouts for the first time correlates daily UF changes with evolving peritoneal membrane function during long-term PD, in relation to clinical risk factors and outcomes. Increased use of APD cyclers with remote patient monitoring will generate large datasets for real-time treatment analysis through cloud-based services, allowing timely treatment modifications to improve patient outcomes. Therefore, continuous UF measurement with automated analysis could monitor peritoneal membrane function and alert clinical teams upon detection of relevant changes, without need for serial “screening” membrane tests (5). Future trials are needed to validate clinical potential of this novel technology.

Disclosures

S.L. Alper reports consultancy agreements with Medical University of Vienna and Swiss National Science Foundation, receiving research funding from Quest Diagnostics, receiving honoraria from Swiss National Science Foundation, and serving as a scientific advisor or member of Swiss National Science Foundation. C. Aufricht is cofounder of Zytoprotec GmbH, a spin-off of the Medical University Vienna that holds the patent “Carbohydrate-based peritoneal dialysis fluid comprising glutamine residue” (International Publication Number: WO 2008/106702 A1). C. Aufricht reports consultancy agreements with Delta 4 GmbH and Zytoprotec GmbH, ownership interest in Delta 4 GmbH and Zytoprotec GmbH, receiving research funding from Zytoprotec GmbH, receiving honoraria from Zytoprotec GmbH, serving as a scientific advisor or member of Delta 4 GmbH and Zytoprotec GmbH, and other interests/relationships with Scientific Advisory Committee of Medicines from AGES (Austrian Agency for Health and Food Safety). R. Herzog is a former employee and consultant of Zytoprotec GmbH. R. Herzog reports consultancy agreements with, research funding from, and honoraria from Zytoprotec GmbH. K. Kratochwill is a former employee and consultant of Zytoprotec GmbH, reports consultancy agreements with Delta 4 GmbH and Zytoprotec GmbH, ownership interest in Delta 4 GmbH, and research funding from Zytoprotec GmbH. K. Oviedo Flores is an employee of Baxter GmbH, Austria, Medical Affairs and is funded as an early stage researcher by the Research and Innovation Program IMPROVE-PD (812699)/Horizon 2020 (Marie Curie Grant Agreement 812699). A. Vychytil reports consultancy agreements with Baxter and receiving honoraria from Baxter, Fresenius Kabi, Fresenius Medical Care, and Zytoprotec GmbH. All remaining authors have nothing to disclose.

Funding

This work is part of the IMPROVE-PD project that has received funding from the European Union's Horizon 2020 Research and Innovation Program under Marie Sklodowska-Curie grant number 812699.