Variation in Peritoneal Dialysis Time on Therapy by Country

Visual Abstract

Abstract

Background and objectives

Quantifying contemporary peritoneal dialysis time on therapy is important for patients and providers. We describe time on peritoneal dialysis in the context of outcomes of hemodialysis transfer, death, and kidney transplantation on the basis of the multinational, observational Peritoneal Dialysis Outcomes and Practice Patterns Study (PDOPPS) from 2014 to 2017.

Design, setting, participants, & measurements

Among 218 randomly selected peritoneal dialysis facilities (7121 patients) in the PDOPPS from Australia/New Zealand, Canada, Japan, Thailand, the United Kingdom, and the United States, we calculated the cumulative incidence from peritoneal dialysis start to hemodialysis transfer, death, or kidney transplantation over 5 years and adjusted hazard ratios for patient and facility factors associated with death and hemodialysis transfer.

Results

Median time on peritoneal dialysis ranged from 1.7 (interquartile range, 0.8–2.9; the United Kingdom) to 3.2 (interquartile range, 1.5–6.0; Japan) years and was longer with lower kidney transplantation rates (range: 32% [the United Kingdom] to 2% [Japan and Thailand] over 3 years). Adjusted hemodialysis transfer risk was lowest in Thailand, but death risk was higher in Thailand and the United States compared with most countries. Infection was the leading cause of hemodialysis transfer, with higher hemodialysis transfer risks seen in patients having psychiatric disorder history or elevated body mass index. The proportion of patients with total weekly Kt/V ≥1.7 at a facility was not associated with death or hemodialysis transfer.

Conclusions

Countries in the PDOPPS with higher rates of kidney transplantation tended to have shorter median times on peritoneal dialysis. Identification of infection as a leading cause of hemodialysis transfer and patient and facility factors associated with the risk of hemodialysis transfer can facilitate interventions to reduce these events.

Podcast

This article contains a podcast at https://www.asn-online.org/media/podcast/CJASN/2022_05_31_CJN16341221.mp3

Introduction

Worldwide, among those receiving dialysis, only 11% receive treatment with peritoneal dialysis (PD), with the majority receiving facility-based hemodialysis (HD) (1,2). PD is associated with similar survival compared with facility-based hemodialysis (HD) (3–6). Compared with patients on HD, patients receiving PD experience superior treatment satisfaction, longer preservation of residual kidney function, and, often, lower annualized treatment costs (7–11). As a result, PD increase in low- and middle-income countries may contain growing kidney failure–associated costs, with countries, such as Thailand, using a PD-first policy (12). Dialysis reimbursement changes have led to recent PD utilization increases in the United States (13).

Increasing PD utilization is limited by the shortened treatment time on PD as compared with facility-based HD. A transition from PD to HD is costly, and it is a period of high morbidity and impaired quality of life (14,15). The Standardized Outcomes in Nephrology-Peritoneal Dialysis study established PD treatment failure as a core outcome of importance to stakeholders (16). In collaboration with the International Society for Peritoneal Dialysis (ISPD), the Peritoneal Dialysis Outcomes and Practice Patterns Study (PDOPPS) is a multicenter, international, observational cohort study designed to identify modifiable practices associated with patient and technique survival among patients on PD (17). The primary objective of this investigation was to quantify PD time on therapy across countries using a standard definition of transition to HD. Secondary objectives were to explore differences by country in causes of HD transfer, identifying patient and facility factors associated with lower risks of death and HD transfer.

Materials and Methods

Study Design and Data Source

Patients on maintenance PD ≥18 years old were enrolled randomly from national samples of randomly selected PD facilities treating at least 20 patients on PD at selection and stratified by geographic region and center size within countries. We included 2014–2017 data from Australia/New Zealand, Canada, Japan, Thailand, the United Kingdom, and the United States. Study details are provided at https://www.dopps.org/OurStudies/PeritonealDialysisPDOPPS.aspx.

Patient demographics, comorbidities, and laboratory data were captured at study enrollment (Table 1). Data from US patients receiving care at large dialysis organization facilities were imported from electronic health records. Other data were obtained by abstraction from medical charts into a web-based data collection tool. PDOPPS obtained institutional review board study approval, and patient consent was obtained to meet national and local ethics regulations.

Table 1.

Patient characteristics

Characteristics	Australia/ New Zealand	Canada	Japan	Thailand	United Kingdom	United States
No. of patients	549	907	795	830	354	3686
Patient age, yr	63 (14)	61 (15)	64 (13)	56 (14)	60 (15)	58 (15)
Men, n (%)	365 (67)	561 (62)	531 (67)	420 (51)	226 (64)	2060 (56)
Time on PD, yr	1.3 (1.7)	1.2 (2.0)	1.9 (2.3)	1.5 (1.9)	1.3 (2.1)	1.5 (1.8)
Body mass index, kg/m2, n (%)
<20	16 (3)	48 (6)	139 (19)	197 (26)	16 (6)	107 (4)
20–29	339 (68)	516 (65)	573 (77)	508 (68)	198 (69)	1519 (56)
30+	141 (28)	231 (29)	33 (4)	40 (5)	74 (26)	1084 (40)
Race, n (%)
Asian	95 (17)	122 (13)	792 (100)	821 (99)	29 (8)	149 (4)
Black	0 (0)	41 (5)	0 (0)	0 (0)	14 (4)	896 (26)
White	382 (70)	634 (70)	2 (0)	2 (0)	303 (86)	2398 (68)
Other	71 (13)	110 (12)	1 (0)	7 (1)	8 (2)	75 (2)
Heart disease, n (%)	276 (50)	443 (49)	335 (42)	202 (24)	156 (44)	1357 (37)
Diabetes, n (%)	237 (44)	435 (48)	313 (40)	400 (49)	95 (27)	1902 (52)
Psychiatric disorder,a n (%)	56 (10)	127 (14)	24 (3)	5 (1)	26 (7)	692 (19)
Prior HD,b n (%)	76 (20)	154 (124)	84 (17)	278 (39)	29 (15)	193 (36)
Urine volume, L/24 h	0.98 (0.76)	1.06 (0.74)	0.90 (0.62)	0.50 (0.58)	1.27 (0.83)	0.83 (0.77)
Caregiver(s) involved in PD exchanges,b n (%)	91 (17)	162 (19)	104 (13)	496 (60)	76 (23)	145 (16)
Albumin, g/dl	3.3 (0.5)	3.4 (0.5)	3.3 (0.5)	3.3 (0.6)	3.4 (0.6)	3.5 (0.5)
Transplant waiting list referred,b n (%)	257 (48)	459 (51)	114 (15)	198 (24)	219 (62)	505 (58)

Results are shown as prevalence or mean (SD). PD, peritoneal dialysis; HD, hemodialysis.

^aPsychiatric disorders include depression, bipolar disorder, schizophrenia/psychotic disorder, alcohol abuse within the past 12 months, or other substance abuse within the past 12 months.

^bPrior HD, caregiver(s) involved in PD exchanges, transplant waiting list referred, and PD solution dextrose concentration were missing in US large dialysis organizations.

Clinical Outcomes

Follow-up started at study enrollment and ended with the first of: death, 7 days after permanent change in dialysis modality, loss to follow-up, kidney transplantation, end of study phase, or a recent date of available data. Permanent HD transfer was defined as either a modality switch to HD identified as permanent or temporary transfers from PD to HD lasting at least 12 weeks (84 days). In primary analyses, hybrid therapy switches (the addition of any HD to continued PD therapy) were counted as an HD transfer event. Patients died within 7 days of HD transfer were counted as death events, not HD transfer.

The Fine and Gray (18) competing risks model was used to generate cumulative incidence curves for death, HD transfer, and kidney transplantation separately for each country and determine median time on PD by country and outcome proportions at 3 years.

Cox regressions were used to investigate the association between country and HD transfer, all-cause mortality, the composite of both, and kidney transplantation. Models were left truncated to account for PD vintage at study enrollment. All models accounted for patient clustering within facilities using robust sandwich-type covariance estimators and, where appropriate, stratified by country. Analyses were conducted on death, HD transfer, composite of death and HD transfer, and transplant using cause-specific models (19).

Crude event rates within each PD vintage time period were estimated as the number of events per 100 patient-years at risk within that time period. In this analysis, each patient who had follow-up in PDOPPS during the corresponding period (e.g., <6 months on dialysis) contributed time at risk to that period and, potentially, an event during that period.

Association of Patient and Facility Factors with Clinical Outcomes

Patient factors evaluated included age, sex, cardiovascular disease, diabetes, prior HD experience, caregiver involvement, psychiatric disorder, body mass index (BMI), Black race, urine volume, serum albumin, and transplant waiting list status.

Facility factors evaluated included facility size, PD facility age (years since facility first started caring for patients on PD), patient-nurse ratio, routine multidisciplinary review, and whether PD nurses were routinely retrained, as well as facility proportion of patients using 4.25% PD solution (Table 2) or with a Kt/V of above 1.7/wk. Given variability in median facility size across countries, quartiles of facility size within a country were used to define small facilities (lower quartile) and large facilities (upper quartile); the remainder was defined as medium. Facility size was also evaluated as a continuous variable.

Table 2.

Facility characteristics

Characteristics	Australia/ New Zealand	Canada	Japan	Thailand	United Kingdom	United States
No. of facilities	20	20	31	22	19	106
Facility size	53 [36–82]	51 [40–90]	29 [20–36]	102 [48–208]	50 [27–65]	32 [24–43]
Patient-nurse ratio	11 [7–13]	14 [11–17]	5.78 [2.58–8.00]	38 [19–49]	8.13 [6.63–9.25]	11 [8–14]
Facility yr of experience treating patients with PD, n (%)
<5	0 (0)	0 (0)	0 (0)	1 (5)	0 (0)	4 (7)
5–9	1 (9)	3 (15)	0 (0)	15 (68)	0 (0)	16 (28)
10+	10 (91)	17 (85)	23 (100)	6 (27)	16 (100)	37 (65)
Facility % of patients with total Kt/V urea <1.7, n (%)
<10	5 (29)	4 (24)	2 (17)	3 (25)	4 (40)	69 (69)
10–19	4 (24)	6 (35)	1 (8)	2 (17)	5 (50)	26 (26)
20+	8 (47)	7 (41)	9 (75)	7 (58)	1 (10)	5 (5)
Facility % of patients who use 4.25% solution, n (%)
0	9 (47)	6 (32)	27 (96)	5 (24)	11 (79)	3 (12)
1–19	7 (37)	10 (53)	1 (4)	11 (52)	3 (21)	2 (8)
20+	3 (16)	3 (16)	0 (0)	5 (24)	0 (0)	20 (80)
Routine multidisciplinary review, n (%)	9 (50)	8 (40)	18 (69)	10 (46)	15 (79)	67 (81)

Results are shown as prevalence or median [interquartile range]. PD, peritoneal dialysis.

Cox models were used to calculate the adjusted hazard ratios (HRs) for death, HD transfers, and the combined outcome and stratified by country and large dialysis organization in the United States. Models on transplantation were adjusted for all factors except kidney transplant waiting list status. Each facility factor was assessed one at a time in models that controlled for the patient characteristics.

Sensitivity Analyses

We conducted two sensitivity analyses: (1) hybrid transfers (defined as the addition of HD but continuation of PD; events were censored and not counted as HD transfer) and (2) testing the effect of our choice of threshold for temporary versus permanent transfers (temporary HD and hybrid transfers were counted as HD transfer).

Treatment of Missing Data

Missing values were multiply imputed using the Sequential Regression Multiple Imputation Method by IVEware (20). Twenty imputed datasets were combined for the final analysis using the Rubin formula (21). Missing data were <25% for all imputed covariates, except PD facility age (27% missing), caregiver involvement (41% missing), urine volume (36% missing), transplant waiting list (40% missing), and prior HD (59% missing). Analyses used SAS software, version 9.4 (SAS institute Inc., Cary, NC).

Results

Patient and Facility Characteristics by Country

Table 1 presents patient characteristics. Mean patient age ranged from 56 (SD 14) years in Thailand to 64 (13) years in Japan. Patients in Japan (15%) and Thailand (24%) were less likely to be on a kidney transplant waiting list (versus 48%–62% elsewhere).

The median numbers of patients treated in Japanese and US PD facilities were 29 (interquartile range [IQR], 20–36) and 32 (IQR, 24–43), respectively. Other countries had median facility sizes of 50–53 patients on PD except Thailand, with a median number of 102 patients. US facilities had a higher proportion of facilities with a Kt/V of 1.7 or greater (69%), and Japan and Thailand had the highest proportions with Kt/V of 1.7 or lower (58%–75%). Hypertonic (4.25%) PD solution was rarely used in Japan and the United Kingdom, but it was used in at least 20% of patients in 80% of US facilities (Table 2).

Absolute Time on Therapy and Relative Hazard by Country

Overall, median time on PD was 2.3 (IQR, 1.1–4.4) years (Supplemental Table 1), the longest time of 3.2 years being in Japan followed by 2.8 years in Thailand; 2.1–2.3 years in Australia/New Zealand, Canada, and the United States; and 1.7 years in the United Kingdom (Figure 1).

Figure 1.

Unadjusted outcomes of peritoneal dialysis (PD) therapy vary by country and by time period after start of PD. Cumulative incidence curve of death, hemodialysis (HD) transfer, and transplant in (A) Australia/New Zealand; (B) Canada; (C) Japan; (D) Thailand; (E) the United Kingdom; and (F) the United States.

Overall, 19% (n=1379) of the patients on PD transferred to HD/hybrid, 13% (n=907) died, and 7% (n=514) received a transplant during follow-up (median of 1.1 years; IQR, 0.6–1.7); however, outcomes varied by country. Less than 1% (n=8) died within 7 days of HD transfer. By 3 years on PD, 36% of patients in Thailand died compared with 11%–18% of patients in other countries. HD transfer was common, with 24%–35% of patients having HD transfer by 3 years on PD except in Thailand, where only 16% of patients transferred to HD. In Japan, 11% of patients switched to hybrid therapy by 3 years on PD, whereas this was <1% in other countries. Only 2% had received a kidney transplant by 3 years on PD in Japan and Thailand, whereas in other countries, 10%–32% had a transplant by 3 years on PD. Similar results were seen including any HD/hybrid transfer (including any temporary transfer) as an event (Supplemental Figure 1).

When compared with the United States, the adjusted HR for the composite outcome, HD transfer or death, was lowest for patients in Japan (HR, 0.75; 95% confidence interval [95% CI], 0.61 to 0.91), followed by Thailand (HR, 0.77; 95% CI, 0.61 to 0.97) and Australia/New Zealand (HR, 0.85; 95% CI, 0.69 to 1.04). Adjusted HRs in Canada and the United Kingdom were similar to that of US patients (Figure 2). Adjusted HRs of HD transfer were lowest in Thailand and similar elsewhere (Figure 2). Adjusted HRs for death were lowest for patients in Australia/New Zealand and Japan (Figure 2). When not counting patients transferred to hybrid therapy as HD transfer, Japan had HRs (compared with the United States) of 0.84 (95% CI, 0.67 to 1.05) for HD transfer and 0.56 (95% CI, 0.46 to 0.68) for the combined outcome of death or HD transfer (Supplemental Figure 2). Japan and Thailand had a lower incidence of transplantation, whereas Australia/New Zealand, Canada, and the United Kingdom had higher incidence compared with the United States (Figure 2).

Figure 2.

Adjusted and unadjusted outcomes of PD are associated with country. Hazard ratios of PD discontinuation due to death/hemodialysis transfer (HDT), death, HDT, or transplant by country compared with the United States. Hazard ratios were estimated separately for each outcome using Cox models left truncated on the basis of PD vintage. Models were adjusted for patient age, sex, body mass index, Black race, heart disease, diabetes, psychiatric disorder, prior HD experience, urine volume, albumin, caregiver involvement, transplant waiting list referred, and accounting for facility clustering. Transplant waiting list referred was excluded as an adjustment from the model for transplant outcome. A/NZ, Australia/New Zealand; 95% CI, 95% confidence interval; ref, reference; UK, United Kingdom; US, United States.

In the model on mortality adjusted only for demographics and comorbidities, Thailand had an HR of 1.79 (95% CI, 1.35 to 2.36) compared with the United States, but this dropped to 0.99 (95% CI, 0.73 to 1.33) after adjustment for serum albumin, urine volume, and caregiver involvement (Supplemental Table 2).

Event Rate by Time on Therapy and by Country

In general, rates of the combined outcome of HD transfer or death were higher with higher PD vintage. Yet Thailand, and, to some degree, the United States were exceptions, with rates largely constant over PD vintage (Supplemental Figure 3A). With longer PD vintage, the rate of HD transfer was slightly higher in Australia/New Zealand, Japan, Thailand, and the United Kingdom but not in Canada and the United States (Supplemental Figure 3B). Patients in Thailand had the lowest rate of HD transfer. Thailand had high death rates across all PD vintages (Supplemental Figure 3C).

Reason for Hemodialysis Transfer

Infection was the leading cause of patients on PD permanently switching to HD (Figure 3A). Issues relating to solute clearance were common, especially in Canada (13%) and Japan (16%). Low Kt/V or low creatinine clearance represented 60% of the causes of inadequate solute clearance leading to HD transfer (Supplemental Table 3). Water removal problems were common in Japan (29%) but not elsewhere (4%–10%). Switching for psychosocial/medical reasons did not occur in Thailand but was at 8%–20% in other countries. Catheter-related problems accounted for 17% of HD transfers in Thailand and <6% elsewhere. As PD vintage increased, there were more infections, solute/water clearance problems, and psychosocial/medical problems but fewer peritoneal leaks/hernia and catheter-related problems (Figure 3B).

Figure 3.

Primary and secondary reasons patients switch to HD vary by country and by time period after start of PD. (A) By country and (B) by PD vintage at the time of modality switch, excluding EHR data where no reason was reported, with 287 events. Another 196 events were excluded due to missing reason. EPS, encapsulating peritoneal sclerosis; Jpn, Japan; Thai, Thailand; EHR, electronic health record.

Association of Patient Characteristics with Clinical Outcomes

Men, diabetes, or psychiatric disorders were associated with higher risk of death or HD transfer (Figure 4A, Supplemental Table 4A). Those needing caregiver help with PD exchanges had a higher death risk (HR, 1.57; 95% CI, 1.30 to 1.91) but not a higher HD transfer risk (HR, 0.86; 95% CI, 0.70 to 1.06). Patients on a kidney transplant waiting list had a lower death risk (HR, 0.62; 95% CI, 0.49 to 0.78) while not being associated with HD transfer. Patients with BMI values <20 kg/m² had a higher mortality risk (HR, 1.38; 95% CI, 1.06 to 1.78). Patients with BMI values >30 kg/m² had a higher risk of HD transfer (HR, 1.28; 95% CI, 1.11 to 1.14). Black race was associated with a lower risk of death (HR, 0.75; 95% CI, 0.59 to 0.94).

Figure 4.

Patient and facility characteristics are associated with PD outcomes. (A) Hazard ratios for patient characteristics. Hazard ratios were estimated using the left truncated Cox model on the basis of PD vintage. The model was adjusted for patient age, sex, body mass index (BMI), Black race, heart disease, diabetes, psychiatric disorder, prior HD experience, urine volume, albumin, caregiver involvement, transplant waiting list referred, and accounting for facility clustering. There are separate models for each outcome. (B) Adjusted hazard ratios of facility factors. Hazard ratios were estimated using the left truncated Cox model on the basis of PD vintage. The model was adjusted for patient age, sex, BMI, Black race, heart disease, diabetes, psychiatric disorder, prior HD experience, urine volume, albumin, caregiver involvement, transplant waiting list referred, and accounting for facility clustering. There are separate models for each outcome. For facility size and patient-nurse ratio, small ≤quartile 1, large ≥quartile 3, and medium =quartile 1–quartile 3 within the country. Median interquartile ranges are listed in Table 2.

Association of Facility Factors with Clinical Outcomes

Compared with medium-sized facilities, patients in smaller facilities had lower risk of death (overall HR, 0.81; 95% CI, 0.68 to 0.97) (Figure 4B, Supplemental Table 4B), largely driven by the low mortality risk in small facilities in Canada (Supplemental Table 5C). There was little association between facility size and HD transfer and within countries (Supplemental Table 5B).

Patients in facilities with routine multidisciplinary review had an HR for the combined outcome of death or HD transfer of 0.92 (95% CI, 0.83 to 1.03) and an HR for HD transfer alone of 0.88 (95% CI, 0.77 to 1.01) (Figure 4B, Supplemental Table 4B). Facilities using any 3.86% PD solution in >20% of their patients had an HR of mortality of 1.24 (95% CI, 0.93 to 1.65). We found no relationship between the proportion of patients at a facility achieving a total Kt/V of 1.7 or higher and the risk of death or HD transfer. We did not observe strong associations between other facility factors and clinical outcomes.

Discussion

In this large-scale, multinational comparison of key clinical outcomes for patients on PD, we quantified an overall median time on PD of 2.3 years with wide observed variability across countries, mostly driven by differences in kidney transplantation (Figure 1). Modest differences in mortality were evident after adjusting for patient case mix, with smaller differences in HD transfer.

The cumulative incidence curves of time on PD demonstrate difficulties in making meaningful international comparisons regarding what time on PD means. For example, the observed time on PD was longer in Japan (3.5 years) than in the United Kingdom (1.7 years). However, this ignores that a larger fraction of UK patients on PD received a kidney transplant in the first 3.5 years than in Japan (34% versus 2%, respectively), which is important from a patient outcomes perspective.

In Thailand, the median time on PD of 2.8 years was longer than that seen for the United Kingdom, but most Thai patients no longer receiving PD 3.5 years after PD start had died. One can expect that the same issues in making meaningful international comparisons regarding time on PD can also exist across facilities within a country. Consideration of a time on PD quality metric(s) for making comparisons across countries or facilities should account for both transplantation and death. Patients considering PD as a dialysis modality may appreciate a metric considering the likelihood of either remaining on PD or receiving a kidney transplant as a positive outcome measure.

Adjusting for patient case mix highlights apparent differences in mortality between countries. Effect estimates for all countries relative to the United States become smaller, with Thailand moving from a higher to an equivalent death rate and other countries moving from an equivalent to a lower death rate. Transplantation may cause informative censoring due to the removal of healthier patients from the PD population, such that the effect estimate for countries with a high transplant rate could be biased upward. As the United States had a lower adjusted transplant incidence compared with Australia/New Zealand, Canada, and the United Kingdom, the apparent differences in survival could be larger, although the change for Thailand and Japan could be reduced (Figure 2).

Studies from 30 years ago demonstrated higher mortality in the United States compared with Canadian patients on PD (22). This discrepancy was not explained then and is not explained now by measurable differences in patient case mix; however, there have been improvements in death rates in the US PD population (23,24). An analysis of time on therapy in the United States using 2008–2011 United States Renal Data System data demonstrated a median survival on PD of 1.9 years (25) compared with our finding of 2.3 years. The difference may be reconciled by our restriction to facilities with >20 patients, inclusion of all insurance types, and inclusion of a more contemporary cohort.

Despite younger age and lower comorbidities, individuals in Thailand had higher caregiver involvement and lower serum albumin and residual kidney function compared with US patients (Table 1), all of which may have been indications of patient frailty and suboptimal predialysis CKD care, potentially explaining the worse unadjusted mortality. Unique socioeconomic, health care, and educational factors in Thailand may also explain this mortality difference (26), including possible reluctance for HD transfer in moribund individuals. Adjusting for measured patient case mix differences, Thailand did not have a higher adjusted death rate compared with the United States (Supplemental Table 2), providing encouraging results to a country with a “PD-first” policy (27).

Adjusting for patient case mix had little effect on the overall rate of HD transfer, explaining a small difference in HD transfer risks in Australia/New Zealand. Age and comorbidities, including diabetes and cardiovascular disease, were more strongly associated with death than with HD transfer.

Similar to a previous report, we found an association between the reason for HD transfer and duration of therapy. Catheter problems, leaks, and hernia were important causes for HD transfer early in PD therapy and became proportionally less common. Poor water or solute clearance and psychosocial/medical reasons became more common with increasing PD vintage (28). The absence of psychosocial causes of HD transfer in Thailand may speak to resistance by patients to HD transfer due to logistical or transportation challenges or consequences of a “PD-first” policy resulting in many long-term patients persisting with PD despite struggling with burnout. External pressures could result in HD transfer for when absolutely medically indicated.

Those with a history of psychiatric disorders faced higher risks of transfer to HD. Previous studies found that depression in patients on PD carried higher peritonitis risks and may have been relatively underdiagnosed across patients on HD and patients on PD, particularly in Japan (29,30). Whether interventions targeting mental health may extend time on PD for these individuals requires evaluation.

Higher BMI was associated with a higher risk of HD transfer. Previous studies have yielded inconsistent findings (31–33), but differences across studies may reflect the different BMI cutoffs used, populations studied, and definitions of HD transfer used. It is also unclear if the risks associated with elevated BMI were driven by increased muscle, fat, or fluid and whether already established mechanisms, such as the known association with peritonitis, or concerns regarding adequate small solute clearance or metabolic complications exacerbated by glucose absorption could explain this. The United States had a greater proportion of high-BMI individuals and the greatest use of hypertonic (4.25%) PD solutions. Whether glucose minimization strategies may mitigate these risks requires further study.

We found that routine multidisciplinary review had a trend to lower risk of HD transfer. Patient review may facilitate identification of at-risk patients and earlier interventions, which could mitigate HD transfer events but may be a proxy for other practices, such as the routine availability of allied health professionals like pharmacists, dietitians, and social workers.

We found no consistent relationship between PD facility size and the risk of HD transfer. In countries such as the United States, the majority of facilities treat <20 patients (24), but to ascribe practices at a facility to outcomes, the PDOPPS sample required a minimum number of 20 patients on PD per facility for study participation. These results should be exercised with caution because it is possible that all PDOPPS clinic sizes may have been above a critical threshold, whereby no relationship exists between size and outcomes. This is supported by a systematic review demonstrating that small facility size adversely affects outcomes predominantly with clinic sizes of <20 (34).

The 2006 ISPD PD prescribing guidelines emphasized the need for a total (peritoneal and kidney) Kt/V of 1.7 on the basis of limited evidence (35), and many national quality metrics now focus on this (36). Recent ISPD guidelines and a Kidney Disease Outcomes Quality Initiative commentary have de-emphasized the role of small solute clearance in delivering high-quality PD care, recommending that many other factors be considered (37,38). We found no relationship between the proportion of patients at a facility at PDOPPS enrollment achieving a total Kt/V of 1.7 or higher and the risk of death or HD transfer. This needs to be interpreted in the context of the restricted number of facilities for which this was available. However, this finding reinforces the need to re-evaluate facility-based Kt/V as a quality measure and a sole marker of the quality of dialysis delivery. Just over 50% of problems with solute clearance causing HD transfer were due to concerns over Kt/V (Supplemental Table 3). It is unclear if this was the sole indication to transfer to HD, but if so, it suggests that a significant minority of patients may experience HD transfer unnecessarily.

The strengths of this study include the large, diverse, multinational patient sample, providing a unique opportunity to quantify and explore time on PD differences and mortality risks across countries. Furthermore, interpretation of the patient factors shown in Figure 4 should be understood to represent their independent contribution within the models used, outside of the treatment factors listed, and not an attempt to describe their overall causal effect as per the “table 2 fallacy” described by Westreich and Greenland (39). The PDOPPS sample relies on voluntary participation, and despite attempts to obtain nationally representative samples, differences may result from the restriction of clinics with >20 patients, and other differences may exist in facility and patient characteristics compared with nonparticipating clinics and patients (40,41). We explored a select set of patient and facility characteristics, but our findings may have been affected by unmeasured confounding (i.e., health care system differences between countries) or informative censoring. Lastly, whether the differences we observed in mortality between countries parallel similar differences seen in the general population or are unique to patients receiving PD was out of the scope of this study.

Notwithstanding these limitations, we have demonstrated differences in median time on PD across countries in PDOPPS and the important role that death and kidney transplantation play in their interpretation. Infection remains a leading cause of HD transfer across all countries. In all countries, opportunities exist to improve time on PD. PDOPPS will continue to focus on identifying important facility practices that will shed light on reducing premature HD transfer and improve survival and quality of life for patients receiving PD.

Disclosures

S.J. Davies reports consultancy agreements with Baxter Healthcare, Ellen Medical, and Zytoprotec; reports research funding from Baxter Healthcare; reports honoraria from Baxter Healthcare and Fresenius Medical Care; has received payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Fresenius Medical Care; reports serving as president of EuroPD and as a member of the publications committee for ISPD; and reports serving in an advisory or leadership role for the International Society of Nephrology (Kidney Failure Strategy). D.W. Johnson reports consultancy agreements with AstraZeneca, AWAK, Bayer, and Lilly; research funding from Baxter and Fresenius; honoraria from Baxter, Fresenius, and Ono; consultancy fees, research grants, speaker’s honoraria, and travel sponsorships from Baxter Healthcare and Fresenius Medical Care; speaker’s honoraria from BI & Lilly and Ono; travel sponsorships from Amgen and Ono; serving in an advisory or leadership role for American Journal of Kidney Disease, CJASN, the Cochrane Kidney and Transplant Group, the National Health & Medical Research Council Academy, and Peritoneal Dialysis International; serving as an Australian and New Zealand Society of Nephrology councilor, a past International Society of Nephrology councilor, and the ISPD Immediate Past President; speakers bureau for Baxter Healthcare and Fresenius Medical Care; and other interests or relationships with Amgen (accommodation sponsorship), Australian and New Zealand Society of Nephrology President, Kidney Health Australia (advisor), past International Society of Nephrology councilor, and ISPD immediate past president. He is a current recipient of an Australian National Health and Medical Research Council Leadership Investigator Grant. T. Kanjanabuch reports research funding from George Clinical PTY Ltd., Australia and VISTERRA Inc.; reports honoraria from AstraZeneca and Baxter Healthcare; has received consultancy fees from VISTERRA as a country investigator; is a current recipient from the National Research Council of Thailand; reports serving as an executive member and representative of developing nations of Cross-Regional Education and Exchange in Dialysis; and other interests or relationships with AstraZeneca and Baxter Healthcare. H. Kawanishi reports honoraria from Kyowa-Kirin Co. Ltd; serving on the PDOPPS steering committee; serving on the editorial boards of Blood Purification, Journal of Vascular Access, and Peritoneal Dialysis International; and serving as the President of the International Society of Blood Purification. Y.-L. Kim reports honoraria from FibroGen; serving on the editorial boards of Life, Peritoneal Dialysis International, and Therapeutic Apheresis and Dialysis; and other interests or relationships with the American Society of Nephrology, European Renal Association - European Dialysis and Transplant Association (ERA-EDTA), International Society of Nephrology (ISN), International Society for Peritoneal Dialysis (ISPD), and the Korean Society of Nephrology. M. Lambie has received speakers honoraria from Baxter Healthcare and Fresenius Medical Care. He received a research grant from Baxter Healthcare in 2013. J. Perl reports consultancy agreements with AstraZeneca, Baxter Health Care Canada, Bayer, DaVita Healthcare Partners, Fresenius Medical Care, LiberDi, and Otsuka; research funding from AHRQ and Arbor Research Collaborative for Health; grants from AHRQ during the conduct of the study; honoraria from AstraZeneca, Baxter Healthcare USA/Canada, DaVita Healthcare partners, DCI, Fresenius Medical Care, and US Renal Care; speakers bureau for Baxter Healthcare and Fresenius Medical Care; salary support from AHRQ and Arbor Research Collaborative for Health; and being on the advisory board for Liberdi. K. McCullough, R.L. Pisoni, B.M. Robinson (principal investigator), and J. Zhao report employment with the Arbor Research Collaborative for Health. The Dialysis Outcomes and Practice Patterns Study Program is supported by Amgen, Baxter Healthcare, and Kyowa Hakko Kirin. Additional commercial support is provided by AstraZeneca, FMC Asia-Pacific Ltd., FMC Canada Ltd., Janssen, Keryx, MEDICE Arzneimittel Pütter GmbH & Co KG, Proteon, and Vifor Fresenius Medical Care Renal Pharma. Optimizing Prevention of PD-Associated Peritonitis in the US (OPPUS) is supported by AHRQ and the National Institute of Diabetes and Digestive and Kidney Diseases. All support is provided without restrictions on publications. All grants are made to Arbor Research Collaborative for Health and not to K. McCullough, R.L. Pisoni, B.M. Robinson, or J. Zhao directly. R.L. Pisoni reports employment with Arbor Research Collaborative for Health, serves on the editorial board for Kidney360, and has served on an advisory board meeting for Vifor regarding uremic pruritus. B.M. Robinson reports employment with Arbor Research Collaborative for Health and serving on the editorial board of American Journal of Kidney Diseases. B.M. Robinson has received consultancy fees or travel reimbursement in the last 3 years from AstraZeneca, GlaxoSmithKline, Kyowa Kirin Co., and Monogram Health, all paid directly to his institution of employment. M. Sanabria is an employee of Baxter Renal Care Services, a company of Baxter Health Care Corp. J.A. Sloand was employed by Baxter Healthcare Corporation at the time of conception/planning of work, was employed by AstraZeneca at the time of development, reports consultancy agreements with GlaxoSmithKline and Sequana, and is an owner of AstraZeneca and Baxter stock/options. J.I. Shen reports serving as an associate editor of Kidney Medicine and an editorial board member of Peritoneal Dialysis International; reports serving in an advisory or leadership role for the North American Council of the ISPD and the PDOPPS Steering Committee; reports serving as a member of the American Society of Nephrology and the National Kidney Foundation; has received funding from the Canadian Institutes of Health Research; and is supported by National Institutes of Health–National Institute of Diabetes and Digestive and Kidney Diseases grant K23DK103972.

Funding

This manuscript was directly supported by Baxter International Inc. (USA). The Dialysis Outcomes and Practice Patterns Study (DOPPS) Program is funded by a consortium of private industry, public funders, and professional societies. More information on DOPPS funding can be found at https://www.dopps.org/AboutUs/Support.aspx. Funding for PDOPPS has been provided by National Health and Medical Research Council (Australia); National Institute for Health Research (UK); National Institute of Diabetes and Digestive and Kidney Diseases (USA); Patient-Centered Outcomes Research Institute (USA); Japanese Society of Peritoneal Dialysis; Canadian Institute for Health Research (Canada); National Research Council of Thailand grant 2558-113; Rachadaphiseksompot Endorcement Fund grant GCURS_59_12_30_03; Chulalongkorn University, Thailand; and the National Science and Technology Development Agency (NSTDA), Thailand.

Author Contributions

S.J. Davies, D.W. Johnson, T. Kanjanabuch, Y.-L. Kim, M. Lambie, K. McCullough, J. Perl, R.L. Pisoni, B.M. Robinson, J.A. Sloand, and J. Zhao conceptualized the study; S.J. Davies, D.W. Johnson, T. Kanjanabuch, Y.-L. Kim, M. Lambie, K. McCullough, J. Perl, R.L. Pisoni, B.M. Robinson, J.A. Sloand, and J. Zhao were responsible for data curation; S.J. Davies, D.W. Johnson, T. Kanjanabuch, Y.-L. Kim, M. Lambie, K. McCullough, J. Perl, R.L. Pisoni, B.M. Robinson, J.A. Sloand, and J. Zhao were responsible for investigation; S.J. Davies, D.W. Johnson, T. Kanjanabuch, Y.-L. Kim, M. Lambie, K. McCullough, J. Perl, R.L. Pisoni, B.M. Robinson, J.A. Sloand, and J. Zhao were responsible for formal analysis; S.J. Davies, D.W. Johnson, T. Kanjanabuch, Y.-L. Kim, M. Lambie, K. McCullough, J. Perl, R.L. Pisoni, B.M. Robinson, J.A. Sloand, and J. Zhao were responsible for methodology; S.J. Davies, D.W. Johnson, T. Kanjanabuch, H. Kawanishi, Y.-L. Kim, M. Lambie, K. McCullough, J. Perl, R.L. Pisoni, B.M. Robinson, J.I. Shen, J.A. Sloand, and J. Zhao wrote the original draft; and S.J. Davies, D.W. Johnson, T. Kanjanabuch, H. Kawanishi, Y.-L. Kim, M. Lambie, K. McCullough, J. Perl, R.L. Pisoni, B.M. Robinson, J.I. Shen, J.A. Sloand, and J. Zhao reviewed and edited the manuscript.

Data Sharing Statement

Data may be shared upon reasonable request to Arbor Research Collaborative for Health.

Supplemental Material

This article contains the following supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.16341221/-/DCSupplemental.

Supplemental Figure 1. Cumulative incidence curve of death, HD transfer, and transplant in Australia/New Zealand, Canada, Japan, Thailand, the United Kingdom, and the United States, including all temporary transfer as the HD transfer event.

Supplemental Figure 2. Hazard ratio of PD discontinuation by country compared with the United States, not counting hybrid transfers as transfer to HD events.

Supplemental Figure 3. Crude event rates by country and by PD vintage: permanent transfer to HD or death, permanent transfer to HD, or death.

Supplemental Table 1. Distribution of time on therapy in years by country on the basis of the cumulative incidence curve in Figure 1.

Supplemental Table 2. Hazard ratios for mortality with 95% confidence intervals in comparison with the United States, showing effects of the sequential levels of adjustment.

Supplemental Table 3. Detailed reasons among 111 patients with solute clearance as the reason for HD transfer.

Supplemental Table 4. Hazard ratio for patient characteristics and adjusted hazard ratio of facility factors.

Supplemental Table 5. Adjusted hazard ratio for other facility factors overall and by country for outcomes of death or HD transfer, HD transfer, or death.