Relationship between initial peritoneal dialysis modality and risk of peritonitis

Peritonitis is a critical complication of peritoneal dialysis (PD). Investigators have reported the risk of peritonitis in patients on continuous ambulatory peritoneal dialysis (CAPD) versus automated peritoneal dialysis (APD), but the available evidence is predominantly based on observational studies which failed to report on the connection type. Our understanding of the relationship between peritonitis risk and PD modality thus remained insufficient. We studied 285 participants who began PD treatment between 1997 and 2014 at three hospitals in Nara Prefecture in Japan. We matched 106 APD patients with 106 CAPD patients based on their propensity scores. The primary outcome was time to first episode of peritonitis within 3 years after PD commencement. In total, PD peritonitis occurred in 64 patients during the study period. Patients initiated on APD had a lower risk of peritonitis than did those initiated on CAPD in both the unadjusted and adjusted models. The hazard ratio (HR) and 95% confidence interval (CI) for the primary endpoint were 0.30 (0.17–0.53) in the fully adjusted model including connection type. In the matched cohort, APD patients had a significantly lower risk of peritonitis than did CAPD patients (log-rank: p < 0.001, HR 0.32, 95% CI 0.16–0.59). The weighting-adjusted analysis of the inverse probability of treatment yielded a similar result (HR 0.35, 95% CI 0.18–0.67). In conclusion, patients initiated on APD at PD commencement had a reduced risk of peritonitis compared with those initiated on CAPD, suggesting APD may be preferable for prevention of peritonitis among PD patients.


Results
Baseline characteristics. Study flowchart was shown in Fig. 1. Two hundred eight-five PD patients (median age 62 years, interquartile range 60-63 years; 192 men) were analyzed in the present study. Table 1 lists their baseline characteristics at PD commencement. One hundred thirty-three patients were on APD; 152 were on CAPD. Diabetes prevalence and connecting device use were significantly higher in patients on APD than in those on CAPD. In our cohort of 205 patients with available solution data, use of 2.5% dextrose peritoneal dialysis solution bag was similar between APD (4%) and CAPD (8%) patients (p = 0.25) but CAPD (27%) patients have significantly high prevalence with use of icodextrin solution compared to APD (6%) patients (p < 0.001).
The groups did not significantly differ after PS matching (Table 1).
Outcome and PD modality. During  www.nature.com/scientificreports/ tis than was CAPD (log-rank p < 0.001; Fig. 2). Both the unadjusted and adjusted models showed that patients who started on APD had a lower risk of peritonitis. The hazard ratio (HR) and 95% confidence interval (CI) for the primary endpoint were 0. After PS matching, APD continued to present a significantly lower risk of peritonitis than did CAPD (log-rank p < 0.001, HR 0.32, 95% CI 0.16-0.59; Fig. 1). We then used inverse probability of treatment weighting (IPTW) using PS to minimize the differences in patient characteristics. The IPTW-adjusted HR (95% CI) for the primary outcome was 0.35 (0.18-0.67) for APD patients compared with CAPD patients.
The analyses stratified by age, sex, diabetes, overweight patients, connection device use and year at PD commencement, showed similar associations among participants, excluding female, no use of connection device and earlier years at PD commencement (Fig. 3).

Discussion
Our study is the first to show that selecting APD rather than CAPD at PD commencement was significantly associated with a lower peritonitis incidence in Japanese PD patients. Adjusting for confounding covariates did not attenuate the HR of APD. Similar results were obtained in the PS-matching and IPTW analyses, suggesting that starting PD patients on APD rather than CAPD can reduce the risk of PD-associated peritonitis within 3 years after PD commencement. Peritonitis is a major cause of PD failure via structural and functional alterations of the peritoneal membrane 16 , which can lead to life-threatening events 9,17 . Previous reports have shown a relationship between PD modality and peritonitis, but these studies are observational 9-13 rather than randomized control trials 14,15 . Results of many studies have shown that patients on APD have a lower or similar incidence of peritonitis than do those on CAPD. A randomized control study in the Netherlands revealed that PD peritonitis occurred significantly less often in APD versus CAPD patients (0.51 versus 0.94 episodes per patient-year, respectively) 14 . A retrospective observational study in the United Kingdom 9 showed that the peritonitis rate was 1:36.7 patient-months for APD-treated patients and 1:28.8 patient-months for CAPD-treated patients, representing an odds ratio of 0.78 favoring APD. Conversely, three recent observational studies found that PD modality was unassociated with a higher likelihood of developing peritonitis [11][12][13] . Differing study populations, follow-up times, improvements in CAPD and APD connection systems, and advances in nursing care and dialysis treatment may have led to the differing results in these studies. The analysis of these studies fails to report on the connection device in the cycler used. These data should also be interpreted with caution because of differences in the analytical methods. Our multicenter study in Japan showed that patients who started on APD at PD commencement experienced significantly less peritonitis than did those starting on CAPD in both the unmatched and matched patients and subgroup analyses, confirming the robustness of our results.
Two hypotheses may explain why APD was associated with a lower peritonitis incidence than was CAPD. First, the number of connections and disconnections required to perform PD may be the most important determinant of peritonitis rates; APD requires fewer connections and disconnections than does CAPD. Second, APD, especially nocturnal intermittent PD, involves shorter dialysate dwelling times than does CAPD, which requires at least 4 to 5 bag exchanges. Fewer connections and shorter dwelling times may reduce the peritonitis incidence by reducing touch contamination in APD-treated patients. However, we found increased number and rate of peritonitis causing gram-positive cocci in CAPD versus APD patients, but without statistically significant. There may be other mechanisms that contribute to risk reduction of peritonitis in APD.
Several study limitations are noted. First, this was an observational study; therefore, the cause-effect relationship between APD and peritonitis incidence was uncertain. Second, the PD modality was selected by the clinicians and their patients, possibly introducing selection bias into the survey findings, although some clinical data were adjusted. Third, only the PD modality at PD commencement was analyzed; changes in PD modality were not followed. Fourth, the enrollment period was relatively long; thus, advances in dialysis treatment, including the dialysate and devices, might have affected the results. Based on results of the subgroup analysis, APD may contribute to risk reduction of peritonitis in recent years. Fifth, some variables associated with the choice of modality and peritonitis were not evaluated. Socioeconomic data, caregiver status and education level are associated with the choice of modality and incident peritonitis in PD patients 18,19 .
In conclusion, selection of APD rather than CAPD at PD commencement was associated with a lower peritonitis incidence, suggesting that APD may be preferable for prevention of peritonitis in PD patients. Larger prospective randomized studies are needed to ensure the robustness of our results.

Methods
Patients. In total, we screened 337 consecutive PD patients who were treated at three centers in Nara Prefecture in Japan between 1 April 1997 and 31 December 2014. Fifty-two patients were excluded because of the inclusion and exclusion criteria, leaving 285 patients for analysis. The inclusion criterion was that the patients must have received maintenance PD for at least one month. Exclusion criteria were missing clinical data, transition from hemodialysis, PD commencement at other clinics, transplantation within 3 years, and age less than Clinical definitions. Baseline demographics and blood sample results were obtained within one month before starting PD through patient interviews and medical records. Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg, or current treatment with oral antihypertensive drugs. Diabetes was defined as a fasting glucose level ≥ 126 mg/dl or current treatment with oral hypoglycemic medications or insulin. Dyslipidemia was defined as low-density lipoprotein cholesterol ≥ 140 mg/dl or current treatment with lipid-lowering medications.
Peritoneal dialysis. In the study PD system of two companies including Baxter Healthcare Corporation and Terumo Corporation was used; 253 (89%) patients were on the system of Baxter Healthcare Corporation and the rest were on that of Terumo Corporation. Connection devices were UV Flash (Baxter Healthcare Corporation, Deerfield, Illinois, USA) and the TSCD (Terumo Sterile Connector Device, Terumo Corporation, Tokyo, Japan). The clinicians and patients determined whether to use CAPD or APD with or without a connection device.
Study outcomes. The primary outcome was time to first episode of peritonitis within 3 years after PD commencement. PD-related peritonitis was defined as an effluent leukocyte count of > 100 cells/mm 3 , with at least 50% being polymorphonuclear leukocytes according to ISPD guidelines 8 . In the entire cohort 2 patients