Predictors of outcomes in patients on peritoneal dialysis: A 2-year nationwide cohort study

There has been no nationwide study of prognostic factors and outcomes in patients on peritoneal dialysis (PD) in Japan. We conducted a cohort study using data from the nationwide registry of the Japanese Society for Dialysis Therapy. We followed 8,954 prevalent PD patients for 2 years, 2014–2015. Cox proportional hazards regression analysis was used to determine factors that were independently associated with patient survival. Survival rates were compared between patients with and without diabetes after adjusting for potential confounders. During the 2-year study period, 893 (10.0%) of 8,954 patients died, 148 (1.6%) underwent kidney transplantation, and 2,637 (29.4%) were switched to hemodialysis; 5,276 (58.9%) patients were alive at the end of the study period. After multivariate adjustment, older age, longer duration of dialysis, presence of diabetes, cardiovascular comorbidity, use of 2.5% glucose dialysate, higher C-reactive protein and phosphate levels, and a lower serum albumin level were independently associated with increased hazard ratios for all-cause mortality. A combination of PD and hemodialysis was associated with a lower mortality rate. The new-onset cardiovascular event rate was significantly higher in the diabetes group than in the non-diabetes group (P < 0.0001). After adjusting for all variables, the hazard ratio was 1.509 (95% confidence interval 1.029–2.189, P = 0.036) in the diabetes group. Diabetes, older age, longer duration of dialysis, cardiovascular comorbidity, and inflammation were predictors of mortality in patients on PD.

In Japan, the limited availability of kidney transplantation as a form of renal replacement therapy means that most patients with end-stage kidney disease (ESKD) require lifelong dialysis. Therefore, the amount of time spent on hemodialysis (HD) or peritoneal dialysis (PD) is longer in Japan than in the US or European countries. The number of Japanese patients on long-term dialysis has been steadily increasing and reached 314,180 at the end of 2013, with 4.2% of these patients having been on dialysis for over 25 years 1 . Many studies have shown that patients on HD in Japan have lower rates of hospitalization and mortality than their counterparts in other countries [2][3][4] . However, there is limited information on hospitalization and mortality rates in patients on PD. Furthermore, the number of patients with diabetes-related ESKD continues to increase 5,6 , which means that the number of patients with diabetes on PD is also increasing 7 .
In view of the lack of data on outcomes in patients on PD in Japan, we undertook a cohort study using a nationwide registry of such patients to clarify the outcomes, including mortality and cardiovascular events, in patients with and without diabetes on PD.
Outcome analysis by demographic factors, dialysis dose, and nutritional factors. We divided the patients into six a priori categories based on duration of PD (<2, 2 to <4, 4 to <6, 6 to <8, 8 to <10, and ≥10 years) to examine the relationship between category of dialysis duration and risk of death. The patients were also categorized by dialysis-related factors, including residual renal function, i.e., anuric or non-anuric state, use of 2.5% glucose or icodextrin dialysate, and use of an automated PD or connecting system. We also divided the patients into five a priori categories based on total Kt/V, i.e., <1.1 and ≥2.0, with intervening increments of 0.3, to examine the dose-response association between the categories of Kt/V and risk of death. BMI, hemoglobin, albumin, non-HDL cholesterol, HDL cholesterol, calcium, phosphate, and CRP levels were included as nutrition-related and inflammation-related factors. Age, hemoglobin, CRP, BMI, and serum albumin, non-HDL cholesterol, HDL cholesterol, calcium, and phosphate levels were analyzed as continuous variables. We estimated the association between demographic factors at enrollment and survival using multivariate Cox proportional hazards regression and analyzed models that were both unadjusted and adjusted for relevant covariates measured at study enrollment, including age, sex, duration of PD, presence or absence of diabetes, comorbid CVD, use of antihypertensive agents, residual renal function, total Kt/V, type of dialysate, hemoglobin, CRP, BMI, albumin, non-HDL cholesterol, HDL cholesterol, calcium, and phosphate. Patients who were switched to HD or underwent kidney transplantation during follow-up were censored from the analysis of outcomes.
Outcome analysis by diabetes and non-diabetes groups. We divided the patients into two groups based on the presence or absence of diabetes. Survival analyses with Cox proportional hazards regression were used to examine whether the presence or absence of diabetes predicted survival for up to 2 years of follow-up. Kaplan-Meier survival analysis was used to compare all-cause, cardiovascular, and infection-related mortality rates. The log-rank statistic was used to test for differences between the two groups. New-onset cardiovascular events were also analyzed. Finally, we examined associations between the two groups and all-cause mortality. Models were analyzed when adjusted for the above-mentioned demographic, dialysis-related, nutrition-related, and inflammation-related factors measured at enrollment. Statistical analysis. Data were summarized using proportions, with mean ± standard deviation or median (interquartile range) as appropriate. Categorical variables were analyzed using the chi-square test and continuous variables using the Student's t-test. Categorical data were compared between groups using repeated-measures analysis of variance and Tukey's honestly significant difference test or Kruskal-Wallis test as appropriate. Missing covariate data were imputed by the conventional method for multivariate regression as appropriate. Proportional www.nature.com/scientificreports www.nature.com/scientificreports/ hazards assumption was tested by graphical and formal testing. All analyses were performed using JMP ® version 13.0 (SAS Institute, Cary, NC). A P-value < 0.05 was considered statistically significant.

Patient demographics and clinical characteristics.
In total, 315,631 patients on maintenance dialysis were registered at the end of 2013. After exclusions, 8,954 patients on PD were enrolled in the study (Fig. 1). Demographic and clinical characteristics of the 8,954 patients at enrollment are shown in Table 1. Mean patient age was 62.7 ± 13.1 years, 63.8% of the patients were male, and 15.1% had comorbid CVD, comprising coronary artery disease, stroke, or limb amputation. Mean duration of PD was 33 (15-60) months, mean BMI was 23.4 ± 3.8 kg/m 2 , and mean albumin and hemoglobin levels were 3.3 ± 0.5 g/dL and 10.8 ± 1.4 g/dL, respectively. The most common cause of ESKD was glomerulonephritis (39.2%), followed by diabetic nephropathy (30.3%) and hypertension (12.9%). Renal Kt/V and PD Kt/V values at enrollment were 0.3 (0-0.8) and 1.4 ± 0.6, respectively. An automated PD system was used in 43.2% of patients and a connecting device in 68.2%. During the 2-year study period, 893 patients (10.0%) died, 148 (1.6%) underwent kidney transplantation, 2,637 (29.4%) were switched to HD; 5,276 (58.9%) patients were still alive at the end of the study period.
All-cause mortality according to demographic, dialysis-related, nutrition-related, and inflammation-related factors at enrollment. Hazard ratios (HRs) in the unadjusted analysis of variables evaluated as potential predictors of mortality in all patients are presented in Table 2. There was no significant effect of sex on patient survival (P = 0.114). Older age, longer duration of PD, presence of diabetes, and comorbid CVD were significant predictors of mortality. Lower dialysis dose, assessed by total Kt/V, was associated with higher risk of mortality. Use of 2.5% glucose or icodextrin dialysate and an anuric state at enrollment were also significant predictors of mortality. Poor nutritional status and more severe inflammatory status, indicated by lower hemoglobin, higher CRP, lower BMI, lower serum albumin, lower calcium, and lower phosphate, were also associated with higher mortality in patients on PD. PD + HD combination therapy was associated with a lower mortality risk. After multivariate adjustment, older age, longer duration of PD, presence of diabetes, comorbid CVD, use of 2.5% glucose dialysate, higher CRP, lower BMI, and lower albumin and phosphate levels continued to be independently associated with increased HR for all-cause mortality ( Table 2). Use of antihypertensive agents, higher HDL levels, and PD + HD combination therapy continued to be associated with lower mortality risk.
In total, 1,240 episodes of PD-related peritonitis occurred in 859 patients during the study period. The mean peritonitis rate was 0.18 per patient-year. Multivariate analysis was performed to identify predictors of peritonitis (Supplementary Table 1). A previous history of peritonitis, use of a connecting device, comorbid CVD, and higher BMI were independent predictors of peritonitis.

Demographics and clinical characteristics in the diabetes and non-diabetes groups. Patients
were divided into two groups based on the presence or absence of diabetes. Table 3 shows the demographics and clinical characteristics of each group. Patients with diabetes were older and more likely to be male, with shorter duration of PD, higher BMI, higher rate of CVD comorbidity, lower HDL cholesterol, and higher CRP levels. There was no significant difference in hemoglobin level, total Kt/V value, or frequency of use of an automated PD system. More patients in the diabetes group used 2.5% glucose and/or icodextrin dialysate than those in the non-diabetes group.

Comparison of outcomes between the diabetes and non-diabetes groups. Outcome varia-
bles are shown in Table 4. In total, 1,020 cardiovascular events (401 fatal, 619 non-fatal) were recorded during the study period. The cardiovascular event rate was significantly higher in the diabetes group than in the non-diabetes group (15.9% vs. 8.7%; P < 0.0001). The leading cause of death in the diabetes group was CVD (42.1%), followed by infection (27.2%). The leading cause of death in the non-diabetes group was also CVD (35.7%), followed by infection (21.7%). EPS was the cause of death in 1 patient in the diabetes group and in 5 patients in the non-diabetes group; the difference was statistically significant (P = 0.0013). Kaplan-Meier survival analysis (Fig. 2) showed a 2-year survival rate of 82.2% in the diabetes group and 87.4% in the non-diabetes group (P < 0.0001). Kaplan-Meier survival analyses for cardiovascular and infection-related deaths revealed www.nature.com/scientificreports www.nature.com/scientificreports/ significant differences in the patient survival rates between the diabetes and non-diabetes groups (both P < 0.0001; Supplementary Fig. 2a,b).
HRs for all-cause mortality between the diabetes and non-diabetes groups. As shown in Fig. 3, unadjusted HR for all-cause mortality in the diabetes group when compared with that in the non-diabetes (reference) group was 1.471 (95% CI 1.303-1.661; P < 0.0001). After adjusting for sex, age, duration of PD, smoking status, and comorbid CVD, HR for all-cause mortality in the diabetes group when compared with the non-diabetes (reference) group was 1.435 (95% CI 1.263-1.631; P < 0.0001). After adjusting for modification of dialysis, including total Kt/V value, anuric state, use of 2.5% glucose dialysate or icodextrin, and PD + HD combination therapy, as well as for baseline factors, HR for the diabetes group compared with the non-diabetes (reference) group was 1.544 (95% CI 1.195-1.996; P = 0.005). Finally, after adjusting for nutrition-related and inflammation-related factors in addition to the demographic factors and dialysis-related factors, the HR for the diabetes group when compared with the non-diabetes (reference) group was 1.509 (95% CI 1.029-2.189; P = 0.036).

Discussion
This study is the first to identify predictors of 2-year mortality in patients on PD in Japan. Duration of PD and the serum albumin level were the most common factors associated with patient survival. Interestingly, PD + HD combination therapy was associated with lower mortality risk. The patients with diabetes who were receiving PD had lower survival rates than those without diabetes. The main strength of this study is that it is a nationwide cohort analysis of outcomes in patients on PD in Japan. Moreover, this is the first Japanese study to address prognostic risk factors in these patients. www.nature.com/scientificreports www.nature.com/scientificreports/ Data from the US Renal Data System for 2010 revealed that the overall 1-year, 2-year, and 3-year survival rates of patients on PD were 83%, 67.2%, and 53.8%, respectively, whereas those in a diabetes subgroup were 80.3%, 61.7%, and 46.6% 12 . A further study from Canada reported 1-year, 2-year, and 3-year survival rates of 91%, 76%, and 66%, respectively, for patients with diabetes and respective rates of 94%, 89%, and 84% for patients without diabetes 13 . Data from single centers in Asia and Europe also showed that survival rates were significantly lower in patients with diabetes on PD than in their counterparts without diabetes 14 . In the present study, the survival rate was significantly lower in the diabetes group than in the non-diabetes group. However, the 1-year and 2-year survival rates in all patients in our study were 91.4% and 85.5%, respectively, and higher than in previous reports; furthermore, the respective 1-year and 2-year survival rates were 91.3% and 82.2% in the diabetes group and 93.8% and 87.4% in the non-diabetes group. These findings suggest that patients with diabetes on PD have lower survival rates than their counterparts without diabetes. BMI and renal Kt/V values at enrollment were significantly higher in the diabetes group than in the non-diabetes group; however, the older age, higher rate of CVD comorbidity, higher CRP level, and lower HDL cholesterol level at enrollment might be attributed to lower survival rate in the diabetes group.
The diabetes group in our cohort consisted of patients with ESKD caused by diabetic nephropathy and those with diabetes as a comorbid condition. Previous studies of patients on HD have compared those with and without diabetic nephropathy as the primary kidney disease and showed poorer survival in those with diabetic nephropathy 15,16 . Furthermore, The Netherlands Cooperative Study on the Adequacy of Dialysis (NECOSAD) reported that mortality in patients with diabetes as the primary kidney disease was similar to that in patients with diabetes as a comorbid condition 17 . Therefore, diabetes has a strong impact on survival even if it is not the primary cause of ESKD; in our cohort, cardiovascular mortality and new-onset CVD rates were higher in the diabetes  Table 3. Demographic, clinical, and laboratory data in diabetes and non-diabetes patients on peritoneal dialysis. Data are shown as the number, percentage, mean ± standard deviation, or median [interquartile range]. APD, automated peritoneal dialysis; Cr, creatinine; CVD, cardiovascular disease; D/P, dialysate/plasma; HD, hemodialysis; HDL, high-density lipoprotein; PD, peritoneal dialysis; PET, peritoneal equilibration test; PTH, parathyroid hormone.
www.nature.com/scientificreports www.nature.com/scientificreports/ group. Patients on PD are exposed to a large amount of glucose absorbed from the dialysate; therefore, continuous exposure to dialysate might worsen glycemic control and cause peritoneal damage in patients with diabetes 18 . Peritoneal absorption of glucose degradation products might enhance formation of advanced glycosylation end products (AGEs), a nonenzymatic reaction between reducing sugars and proteins 19,20 . In the PD population, increased peritoneal membrane transport and protein permeability is more common in patients with diabetes than in those without diabetes, and these characteristics has been shown to be associated with a number of risk factors   Comparison of hazard ratios for all-cause mortality between diabetes and non-diabetes patients on peritoneal dialysis using standard Cox proportional hazards regression. Model 1 is adjusted for age, sex, duration of PD, and presence or absence of comorbid cardiovascular disease. Model 2 is adjusted for basic demographic and dialysis-related factors, including total Kt/V, anuric or non-anuric state, type of dialysate, and PD + HD combination therapy. Model 3 is adjusted for demographic, dialysis-related, nutrition-related, and inflammation-related factors, including body mass index and C-reactive protein, hemoglobin, serum albumin, non-HDL cholesterol, HDL cholesterol, calcium, and phosphate levels. CI, confidence interval; HD, hemodialysis; HDL, high-density lipoprotein; HR, hazard ratio; PD, peritoneal dialysis.
www.nature.com/scientificreports www.nature.com/scientificreports/ for mortality and technique failure, including age, racial origin, BMI, burden of comorbid illness, hypoalbuminemia, and elevated inflammatory markers [21][22][23] . Patients with diabetes on PD may have higher levels of peritoneal AGEs associated with endothelial dysfunction and atherosclerotic CVD 24,25 . However, it was reported that overall survival of patients on PD was significantly higher than that of those on HD with diabetes on incident dialysis, and that the better patient survival on PD was influenced by the degree of glycemic control, that is, an HbA1c level below 8.0% 26 . The present study revealed that 1.5% glucose dialysate and icodextrin dialysate are frequency used in Japan. However, given that the PD dialysate prescription patterns vary from country to country, further studies are needed to clarify whether use of dialysate with a low glucose concentration might affect patient survival.
PD + HD combination therapy has been approved in Japan since 2010 and was used by 18.8% of the patients in our study cohort. Combined therapy is indicated for patients in whom adequate solute clearance (total Kt/V urea >1.7) cannot be maintained with a standard PD prescription and for patients with symptoms of uremia despite a total Kt/V urea >1.7 or a state of persistent fluid overload. It was reported that 50% of patients on combined therapy had a PD duration of more than 8 years 27 . In the present cohort, the duration of PD was significantly longer in the patients receiving combination therapy than in those receiving PD alone (72 months vs. 38 months, P < 0.0001). On the day of HD and on the following day, patients on combination therapy have a PD holiday, that is, they are released from bag exchange for PD 28,29 , which might lead to a decrease in the D/P Cr ratio obtained from the peritoneal equilibration test, suggesting improved functioning of the peritoneal membrane. PD + HD combination therapy is widely used in Japanese patients on PD in view of its known effectiveness for solute clearance and fluid management 30,31 . In one report, there was a decrease in body weight, blood pressure, and serum creatinine, as well as an increase in hemoglobin level, after switching from PD alone to combined therapy, suggesting significant improvement in dialysis and decreased fluid overload 32 . In the present study, all-cause mortality was significantly lower in patients who were receiving combination therapy than in those on PD alone. However, patients selected for PD + HD combination therapy might be different from those in the general PD population, which would influence all-cause mortality and other clinical outcomes. Furthermore, we could not compare the characteristics, including duration of PD and HD, of patients on PD + HD combination therapy in this study. Therefore, further prospective cohort studies are needed to determine whether this combination therapy has a survival benefit, since the patients receiving PD + HD combination therapy and those receiving only PD therapy may constitute different cohorts.
This study has several limitations. First, as in any annual survey or observational cohort study, facility bias could not be excluded. Therefore, the mortality rate and incidence of peritonitis may have varied between centers because of differences in practice and in patient populations. Second, the follow-up duration of 2 years was relatively short, and prevalent patients rather than incident patients were included. Therefore, there are several factors that could vary according to duration of PD, including residual renal function, total Kt/V, and type of dialysate used. Furthermore, early technique failure (within the first 12 months of commencement of PD) is common, and peritonitis is one of the causes of early technique failure. Therefore, we could not detect patients with early technique failure or the exact rate of development of peritonitis. It is not known whether the rates of peritonitis were affected by differences in patient education on how to prevent these complications. Similarly, the effects of physical activity levels, employment status, and socioeconomic factors are unknown. Therefore, prospective studies on whether the outcomes for patients on PD are affected by differences in these factors are required in the future. Finally, glycemic control in patients with diabetes was not assessed because of the small number of patients with data on glycemic indices. Further studies would be needed to assess the important issue of whether the outcomes in diabetes patients on PD would be affected by glycemic control.
In conclusion, older age, longer duration of PD, use of 2.5% glucose dialysate, lower serum albumin level, more severe inflammatory status, comorbid CVD, and higher phosphate level were significant predictors of mortality in Japanese patients on PD. Higher BMI and HDL cholesterol levels were associated with lower mortality risk in these patients. This study identified that all-cause mortality rate is higher in diabetes patients on PD than in their counterparts without diabetes. Further studies are needed to clarify whether PD + HD combination therapy has a survival benefit. The findings of this study underscore the need for further research on the factors that influence patient outcomes to identify alternative interventions that would improve the outlook for patients receiving PD.