Association between causes of peritoneal dialysis technique failure and all-cause mortality

Technique failure is a frequent complication of peritoneal dialysis (PD), but the association between causes of death-censored technique failure and mortality remains unclear. Using Australian and New Zealand Dialysis and Transplant (ANZDATA) registry data, we examined the associations between technique failure causes and mortality in all incident PD patients who experienced technique failure between 1989–2014. Of 4663 patients, 2415 experienced technique failure attributed to infection, 883 to inadequate dialysis, 836 to mechanical failure and 529 to social reasons. Compared to infection, the adjusted hazard ratios (HR) for all-cause mortality in the first 2 years were 0.83 (95%CI 0.70–0.98) for inadequate dialysis, 0.78 (95%CI 0.66–0.93) for mechanical failure and 1.46 (95%CI 1.24–1.72) for social reasons. The estimates from the competing risk models were similar. There was an interaction between age and causes of technique failure (pinteraction < 0.001), such that the greatest premature mortality was observed in patients aged >60 years post social-related technique failure. There was no association between causes of technique failure and mortality beyond 2 years. In conclusion, infection and social-related technique failure are associated with premature mortality within 2 years post technique failure. Future studies examining the associations may help to improve outcomes in these patients.

Association between causes of PD technique failure and all-cause mortality. Between 0 and 2 years post-PD technique failure, compared to patients who had experienced infection-related PD technique failure, the adjusted HR of all-cause mortality were 0.83 (95% CI 0.70-0.98), 0.78 (95% CI 0.66-0.93) and 1.46 (1.24-1.72), respectively for patients who had experienced inadequate dialysis, mechanical failure and social-related technique failure. Other covariates associated with mortality are shown in Table 3. There was no association between causes of PD technique failure and mortality beyond 2 years post-PD technique failure in the Cox regression models (Fig. 2).
In the competing risk analysis, compared to patients who had experienced infection-related PD technique failure, the adjusted subdistribution HR for mortality between 0 and 2 years post-PD technique failure in patients who had experienced inadequate dialysis, mechanical failure and social reasons were 0.83 (95% CI 0.70-0.98), 0.78 (95% CI 0.66-0.93) and 1.47 (95%CI 1.24-1.73), respectively (Fig. 3a). Other covariates in the competing risk models that were associated with mortality are shown in Table 3.

Sensitivity analyses.
In sensitivity analyses, similar association between causes of PD technique failure and all-cause morality was observed within the first six months and one year post technique failure, but not observed beyond the first 2 years. Between 2 and 3 years post-PD technique failure, compared to patients who had experienced infection-related PD technique failure, the adjusted HR of all-cause mortality were 1.04 (95% CI Cause-specific mortality. Cardiac death was the most common cause of death post-PD technique failure (n = 959, 36%), followed by dialysis withdrawal (n = 712, 27%) and infection (n = 297, 11%) ( Supplementary  Fig. S1). Between 0 and 2 years post-PD technique failure, the most common causes of death in patients who had experienced PD technique failure attributed to infection and social reasons were cardiovascular disease and withdrawal from dialysis, respectively. Beyond 2 years, cause-specific mortality rates were similar among all groups ( Table 2).
Compared to patients who had experienced infection-related PD technique failure, the adjusted HR for cardiac mortality between 0 and 2 years post-PD technique failure were 0.92 (95% CI 0.70-1.20), 0.82 (95% CI 0.61-1.10), and 1.31 (95%CI 0.98-1.75) for patients who had inadequate dialysis, mechanical failure, and social-related failure, respectively. The adjusted HR for infection and withdrawal-related mortality are shown in Fig. 4. There was no association between causes of PD technique failure and cause-specific mortality beyond 2 years post-PD technique failure in the Cox regression models. In the competing risk model, compared to patients who had experienced infection-related PD technique failure, the adjusted subdistribution HR for cardiac mortality within the first 2 years post-technique failure were 0.95 (95% CI 0.30-1.25), 0.75 (95%CI 0.56-1.00), 1.11 (95% CI 0.83-1.49) for patients who experienced inadequate dialysis, mechanical and social-related failure, respectively. The competing risk regression results for infection and withdrawal mortality are shown in Fig. 3b.
Interaction between causes of PD technique failure and age for mortality. Patient age modified the association between causes of PD technique failure and all-cause mortality between 0 and 2 years post-PD technique failure (p-value for interaction <0.001), and therefore, age was categorized into quartiles (i.e. <= 50, 51-60, 61-70 and >70 years) for subsequent analysis. For patients aged >70 years (n = 1221), compared to infection-related PD technique failure, the adjusted HR for mortality within the first 2 years in patients who had experienced inadequate dialysis or mechanical failure (considered as a single group) was 0.78 (95% CI 0.62-0.97, p = 0.03), and 1.73 (95%CI 1.36-2.20, p < 0.001) for patients with social-related technique failure. Similar findings were observed in patients aged 61-70 years. For patients aged ≤50 years (n = 1126), compared to patients who had experienced infection-related technique failure, the adjusted HR for mortality within the first 2 years in patients who had experienced inadequate dialysis or mechanical failure was 0.79 (95% CI 0.54-1.16, p = 0.23), and 0.81 (95%CI 0.46-1.41, p = 0.45) for patients who had experienced social-related technique failure ( Supplementary Fig. S2). Causes of death stratified by age groups between 0 and 2 years post-PD technique failure are shown in Fig. 5. Era did not modify the association between causes of PD technique failure and mortality at any time points.

Discussion
This longitudinal cohort study over a period of two decades showed differential and time-dependent associations between causes of PD technique failure and all-cause mortality. Infection remained the most common cause of PD technique failure over successive eras, although the incidence was reduced by over 10% between 1989-1998 and 2007-2014. The greatest risk of mortality was observed in patients who had experienced social-related and infection-related PD technique failure, with 35% and 30% dying within the first 2 years post-PD technique failure, respectively. Cardiac mortality was the most frequent cause of death in patients who had experienced infection-related PD technique failure, whereas withdrawal mortality was the most frequent cause of death for social-related technique failure. Age modified the association between causes of PD technique failure and mortality such that, in older patients, only those who had experienced social-related technique failure were more likely to die within 2 years of PD technique failure, with the majority of the deaths attributed to psychosocial reasons. There were no associations between the other causes of PD technique failure and death in older patients.
Infection remained the most frequent cause of PD technique failure, with similar findings being observed in other countries across the United States, Canada, Europe and Asia 6,7,10-13 . Although a substantial reduction in the proportion of infection-related PD technique failure in Australia and New Zealand was observed over the last two decades, this was in direct contrast to observations in other countries. Data from the Canadian registry showed that infection-related technique failure had remained constant between 1995 and 2009, but PD technique failure attributed to inadequate PD had decreased over time 14   characteristics and systematic differences in the uptake/management of PD patients between countries as well as a lack of standardized definitions of PD technique failure attributed to infections or other causes. Population cohort studies have consistently shown that PD patients who had experienced peritonitis were 2-6 times more likely to experience premature mortality within the first 1-2 months post-occurrence of peritonitis, with cardiac mortality being the predominant cause of mortality 8,9,15,16 . However, the association between different causes of PD technique failure and mortality remains unclear. In a single-centre study involving 286 incident PD patients between 2004-2010, risk of mortality doubled in patients who had experienced infection-related  Table 3. Association between causes of peritoneal dialysis technique failure and all-cause mortality between 0-2 years post PD technique failure. Data expressed as hazard ratio (95% confidence intervals, p-value). SHRsubdistribution hazard ratios. CAPD -continuous ambulatory peritoneal dialysis. APD -automated peritoneal dialysis. #There was a significant interaction between age and causes of technique failure and the outcome of all cause mortality in the first 2 years post PD technique failure (p < 0.001).
PD technique failure in the first 30 days post peritonitis, compared to other causes of PD technique failure 15 . In contrast, a cohort study from the United Kingdom showed that PD patients who had experienced social-related technique failure had greater mortality, compared to those who had experienced recurrent peritonitis or ultrafiltration-related PD technique failure (2-year survival 50%, 70%, 80%, respectively) 11 . The present study suggested that there was a time-dependent association between causes of PD technique failure and mortality, with infection and social-related PD technique failures having the highest risks of mortality within the first 2 years post-technique failure. However, there were no associations between causes of PD technique failure and mortality beyond 2 years post-technique failure. In patients with PD technique failure attributed to social reasons, 31% died from cardiac mortality and 35% died from withdrawal mortality, with 44% reported to have withdrawn for comorbid conditions such as cardiovascular disease and malignancy. The greater comorbid burden is likely to explain the higher mortality rates within the first few years post-PD technique failure in patients who had experienced social-related technique failure, as suggested by previous studies 17,18 . The finding of a higher risk of mortality in those who had experienced infection-related PD technique failure was predictable and consistent with previously published studies that have shown an independent association between infection and premature mortality. Even though patients who had experienced infection and social-related PD technique failures were more likely to die from infection and social-related mortality, respectively, cardiac death remained a common cause of death in these patients suggesting that causes of mortality were often multifactorial and intertwined. The lack of association between causes of PD technique failure and mortality beyond 2 years post-technique failure suggested a lack of persistent long-term adverse effects of PD failure attributed to infections or social causes. A direct relationship between age and mortality risk has been consistently shown in incident PD patie nts 1,5,7,10,19 . In this study, we have shown that the association between PD technique failure and mortality was modified by age. Older patients with social-related PD technique failure were more likely to die prematurely post-technique failure compared to those who had experienced other causes of technique failures, likely related to the greater prevalence of comorbid burden in older patients. In contrast, younger patients aged between 51-60 years whom had experienced infection-or social-related PD technique failure appeared to have a higher risk of premature mortality (majority being cardiac death) compared to other causes of PD technique failure, suggesting that clinicians will need to be more vigilant in monitoring complications post-PD technique failure in this group of patients. Given the small number of events, we are unable to generate reliable estimates with any certainty.
There are several limitations that must be considered when interpreting the findings of this study. Even though there were multiple confounders adjusted for in the analysis, there are likely to be unmeasured and residual confounders. Data on the severity of comorbid conditions, medications, urine volume and volume status, socioeconomic status, care giver status, selection of ESKD patients for PD, management of PD and related complications, residual renal function at the time of PD technique failure, and other factors that could have potentially modified the association between causes of PD technique failures and mortality were not collected by the ANZDATA registry. In addition, the ANZDATA registry does not verify the reported causes of technique failure by the treating nephrologist and, as such, misclassification bias for the causes of PD technique failure is possible. The association between causes of PD technique failure and mortality was time-dependent, with technique failure attributed to infection or social causes being associated with greater risks of mortality within the first 2-years following PD technique failure. Cardiac and infectious mortality were the commonest causes of  premature mortality, particularly in younger patients who had experienced infection-related technique failure. Therefore, clinicians should be cognizant of the need for close follow-up with appropriate management of these patients. Future studies aiming to explore and understand these associations in greater detail may provide clinicians the opportunity to help better inform the development of individualized management plans in PD patients, which may help to improve their outcomes following transition to haemodialysis.

Study Population. All adult ESKD patients (aged ≥18 years) who had commenced PD between 1989
and 2014 and had experienced PD technique failure in Australia and New Zealand were included. PD technique failure was defined as transfer to haemodialysis for ≥30 days, and excluded technique failure attributed to death 12 . Patients who were maintained on PD for less than 90 days, those who were re-initiated on PD following technique failure and patients who had died whilst on maintenance PD were excluded. Data were censored as of 31 st December 2014. Approval of study by research ethics committee and informed consents were not required because only de-identified information was utilized for analysis. However, consents for inclusion in the ANZDATA registry were sought from all patients with end stage kidney disease in Australia and New Zealand. Data Collection. Baseline characteristics included gender, ethnicity, causes of primary ESKD, comorbidities at time of PD initiation (i.e. presence of diabetes, coronary artery disease, chronic lung disease and cerebrovascular disease), smoking history, body mass index (BMI), PD-related factors (such as PD duration and modality) and era (categorized as 1989-1998, 1999-2006 and 2007-2014).

Exposure factors.
Causes of PD technique failure were categorized into four groups (infection, inadequate dialysis, mechanical failure and social reasons) 12 . PD technique failures attributed to unknown or other causes were excluded from the study (n = 580; n = 125 unknown cause, n = 21 encapsulating peritoneal sclerosis, n = 19 transfer outside Australia or New Zealand, n = 415 other causes). Definitions of each causes of PD technique failure are shown in Supplementary Table S1.
Clinical Outcomes. The primary outcome was all cause mortality. Secondary outcomes included cause-specific mortality (cardiac, infectious and dialysis withdrawal). Cardiac death included cardiac arrest, myocardial ischaemia and cardiac failure. Infection included septicaemia, peritonitis (bacterial, fungal, other) and infections at any sites. Dialysis withdrawal included psychosocial reasons, comorbidities and dialysis access difficulties (please see http://www.anzdata.org.au/forms/ANZDATA/anzdata_A3_2013.pdf).
Statistical analysis. Data were expressed as numbers (percentages) for categorical data, means and standard deviations (SD) for normally distributed continuous data, and median and interquartile range [IQR] for continuous data that were not normally distributed, with comparisons between causes of PD technique failure using chi-square test, analysis of variance (ANOVA) and Kruskall-Wallis test, respectively where appropriate. The temporal trends in the causes of PD technique failure were examined across three time-periods of 1989-1998, 1999-2006 and 2007-2014. Mortality rates, expressed as the number of deaths per 100 patients, were calculated for each of the four causes of technique failure. The association between causes of PD technique failure and mortality was examined using adjusted Cox proportional hazard regression, censoring for kidney transplantation. The proportional hazard assumptions of the models were checked graphically by plotting the Schoenfeld residuals. As the models for mortality were found to violate the proportional hazards assumption, the time period between PD technique failure and mortality was separated into 0-2 years, >2-5 years and >5 years for analysis. Potential interaction between the exposure factor and age or era was examined using two-way interaction terms in the