Abstract BACKGROUND AND AIMS Due to chronic damage of the peritoneal membrane on long-term treatment, peritoneal dialysis (PD) is a time-limited therapy. Effective PD is achieved by adequate peritoneal ultrafiltration (UF) and small solute clearance while maintaining an appropriate fluid and salt homeostasis. Availability of remote patient monitoring on automated PD (APD) allows clinicians to remotely retrieve data for treatment analysis, proposed to improve clinical outcomes. However, machine-derived data are vastly different from traditionally observed data (e.g. quantity, noise, missing normative values). Recently correlation of daily UF changes on APD with changing peritoneal membrane function in association with clinical risk factors and outcomes has been observed. High intra- and interpatient variability of UF remains not fully explained. We aim at a better understanding of UF variability by submitting APD machine readouts to spectral analysis in association with clinical risk factors of membrane dysfunction. METHOD This was a secondary analysis of the Medical University of Vienna APD cycler treatment data from patients using Homechoice Pro cyclers (Baxter, IL, USA), extracted from the cycler management software (PD link, Baxter). Analysis was conducted using R software (R Core Team 2020). Daily UF was processed using the Lomb-Scargle periodogram (LSP), a method commonly used in astrophysics and neurophysiology to detect rhythms in (unevenly sampled) time series. Peaks of power spectrum, frequencies, and periodicities were compared between patients with and without peritonitis episodes during their PD treatment utilizing a Mann–Whitney U-test and associated with the total number of peritonitis episodes using Kendall rank correlation. A P-value < 0.05 was considered significant; all tests were two-sided. RESULTS LSP analysis was performed on daily UF data from n = 129 APD patients with a mean observation time of 598 days (SD ± 517), male: female 61:39% (n = 79:50), incident: prevalent patients 77:23% (n = 99:30), and diabetic: non-diabetic patients 25:75% (n = 32:97). Observation time extended from the introduction of APD until kidney transplantation (43%, n = 55), transfer to hemodialysis (23%, n = 29), death (27%, n = 35), kidney function recovery (8%, n = 1), or loss to follow-up (7%, n = 9). After inspection of LSPs n = 9 patients were excluded from further analysis as their periodicity matched the time of their clinical endpoint. A total of 45 of the remaining 120 patients (38%) never experienced a peritonitis episode, and a total of 0.3 (SD ± 0.7) peritonitis episodes/patient-year were observed. LSP computation revealed cyclic periodicity in UF in all patients with an overall median peak period length of 329 days (IQR 62–882 days), median peak power of 21 (IQR 7–91), and median peak frequency of 0.42 (IQR 0.36–0.46). Differences between patients with and without peritonitis episodes during PD treatment are displayed in Table 1. Correlation analysis revealed significant association between the number of peritonitis episodes and peak periodicity of UF (tau=0.14, P = 0.04), and peak power (tau = 0.21, P = 0.003). Peak frequency of UF was not significantly correlated. CONCLUSION Patients on APD cycler treatment seem to display cyclic periodicity of UF, being significantly associated with the number of peritonitis episodes and significantly different between patients with and without peritonitis episodes. Smaller-scale periodicity (median 216 days) in patients without peritonitis seems to be disrupted by peritonitis resulting in significantly higher periodicity (median 670 days). The causes and implications of these results remain unclear and need further investigation. However, as peritonitis is a known risk factor for peritoneal membrane dysfunction, prolongation of UF periodicity might reflect disruption of peritoneal membrane function.