Are Peritoneal Dialysis Center Characteristics a Modifiable Risk Factor to Improve Peritoneal Dialysis Outcomes?

Survival of patients commencing with peritoneal dialysis (PD) has steadily improved over many years and is equivalent to hemodialysis, perhaps better early on in selected cohorts (1). The same cannot be said for technique survival, which is less competitive, remaining an important challenge for the modality. If we are to improve outcomes in PD, then technique failure must be tackled, and this will require both the identification of the best practices and solutions that ameliorate cause-specific technique failure and the uniform implementation of these practices, so that each patient on PD benefits from the best possible management. There is already evidence that dialysis unit practice differences result in significant variation in access to PD treatment, and this likely extends to variability in technique survival that is independent of patient-level characteristics (2,3). Technique failure can be defined as leaving the modality due to either death, including death within a month of transfer or transition to hemodialysis (referred to here as combined technique failure), or a modality switch that is not associated with death (death-censored technique failure). To fully appreciate the effect of predictors on a composite outcome, such as in the case of combined technique failure, it is necessary to look at the individual outcomes separately. Therefore, to understand drivers of combined technique failure, both predictors of death-censored technique failure and death should be examined, whereas for death-censored technique failure, specific causes are likely to be more important. In particular for death-censored technique failure, the definition should also specify the period off PD (for example, >30 or >180 days). The relevance of this time distinction is that the chances of returning to PD are different according to the specific causes and periods off PD (for example, technique failure due to catheter malfunction is more common early in treatment and more likely to lead to a return to PD) (4). The picture is further complicated by transplantation, an important but positive reason for discontinuing PD but not considered a technique failure. In the most recent analysis of technique failure by the Australian and New Zealand Registry (ANZDATA), these approaches to defining technique failure have been used to establish the relative contributions of patient characteristics and dialysis unit practices (5). In this comprehensive registry-based study including 51 Australian centers and 10,571 patients over a 10-year period (2004–2014), the annual combined technique failure rate was 0.35 episodes per patient-year, of which just under two thirds were death-censored technique failure events. Importantly, the between-center variabilities were sevenfold for combined technique failure and closer to tenfold for death-censored technique failure rates. Of the center-level effects, by far the most apparent was that of center size. For death-censored technique failure but not death, this was a clinically relevant result, with rates highest in centers taking on <16 patients per year and lowest in those with >48. Undoubtedly, patient mix accounts for some of the between-center variation (as expected, this was higher for combined technique failure [28%] than death-censored technique failure [15%]), but adjustment for several additional center-level characteristics explained much more of the variation—an additional 53% for combined and 37% for death-censored technique failure. This is not the first time that outcomes on PD have been associated with dialysis center size. As the ANZDATA authors point out, there has been a signal for some time that small PD programs have higher technique failure rates, such as those found by several registries, including the Dutch, Brazilian, French (6), and Canadian registries and the US New England Network (7), well summarized in the systematic review by Pieper et al. (8). Many of these studies do not clearly distinguish between combined and death-censored technique failure; however, by implication, the greater effect was on the latter, and the cutoff for worse outcomes is seen with PD programs between 15 and 25 patients, most commonly <20. It should, of course, be remembered that death-associated and -censored technique failures are competing events, and this is evident from the ANZDATA analysis, in which greater age, as would be expected, is a risk factor for combined technique failure (hazard ratio [HR], 1.08 per decade; 95% confidence interval [95% CI], 1.05 to 1.1) due to a high risk for death (HR, 1.58 per decade; 95% CI, 1.51 to 1.65) but reduced risk for death-censored technique failure (HR, 0.93; 95% CI, 0.91 to 0.96) (5). This finding is similar to that observed by the French Peritoneal Dialysis Registry, which found reduced death-censored technique failure for patients on assisted PD due to their relatively high mortality (9). Case mix is, therefore, important in interpreting these findings, and ideally, data should be analyzed using a competing risk model that also includes transplantation. The ANZDATA group did take transplantation into account but only by including transplant status as a center-level characteristic rather than as center-specific transplant rates (which might also include preemptive transplantation). Despite these caveats regarding case mix, cumulative data that small PD dialysis centers are a risk factor for worse technique failure are now very strong. It tends to suggest that experience or confidence in the modality is particularly important. It is also likely that this goes hand in hand with clinician’s propensity to use this modality in their patients. This is supported by the observation that not just larger PD units but also, centers where a greater proportion of patients started on this modality in the ANZDATA cohort had better outcomes. Data from the United Kingdom suggest that, here again, both patient- and center-level characteristics are important (2). Independent of patient mix, United Kingdom clinicians who favored home-based therapies had proportionally larger PD programs in their centers. There are, therefore, at least two modifiable center-level risk factors that could be addressed to improve the chances of a patient benefiting from a good outcome on PD: critical mass and experience of the PD center as well as enthusiasm for the use of PD by the clinical team. Likely, these are, to some extent, interdependent. The ANZDATA group also examined cause-specific technique failure divided into four categories: infection, social, inadequate dialysis, and mechanical. The overall highly significant and welcome improvements over time seen in both combined and death-censored technique failure observed comparing 2010–2014 with 2004–2009 showed an interesting pattern, with a marked reduction in the infection-associated failures (HR, 0.64; 95% CI, 0.57 to 0.72) but an increase in social failures (HR, 1.22; 95% CI, 1.00 to 1.47). The dramatic improvement in infection as a cause of poor outcomes may well reflect the concerted effort by the ANZDATA group to reduce overall infection rates and in particular, the between-center variation in infection rates that has been well documented. Again, this is strong evidence that reducing center variation is indeed a modifiable risk factor for PD technique survival. The more modest but significant increase in social failures may well reflect a change in the patient mix over time, because dialysis populations become older and frailer, or again, this could reflect a competing risk; because patients remain on PD longer due to infection prevention, they may be more subject to treatment burnout and social failure as time passes. What are the implications of these findings for the international Peritoneal Dialysis Outcomes and Practice Patterns Study (PDOPPS)? The PDOPPS is a global initiative by the International Society of Peritoneal Dialysis and the Arbor Research Collaborative for Health that is primarily focused on the problem of technique failure. Involving several countries, including the United States, Canada, the United Kingdom, Japan, Thailand, Australia, and New Zealand, it seeks to establish the most important practice patterns associated with avoidance of combined and death-censored technique failure (10). It is measuring both between- and within-country variability in patient outcomes, including cause-specific technique failure, and establishing to what extent these are determined by patient- and center-level characteristics. The high amount of center-level variation observed by the ANZDATA is likely to reflect measurable differences in center practices, providing further rationale if it were needed to undertake this study. Whereas the ANZDATA study relied on a relatively small number of center-level practices, generally limited to measures that could be derived from registry-level data, the PDOPPS has developed comprehensive questionnaires that will interrogate center practices as well as enable both between- and within-country comparisons that are hypothesis driven. This should also allow the PDOPPS to identify confounding by correlated practice patterns that are potentially unmeasured in this study, thus pinpointing the key performance indicators that will translate into reduced technique failure. However, the PDOPPS may not add significantly to the now strong evidence regarding center size. Centers with low patient numbers (i.e., <20) have largely been excluded from the PDOPPS on the basis of that they would be unable to recruit sufficient numbers to achieve statistical power and also, the likelihood that these centers would be least responsible for driving best practice (10). The exception to this is Japan, which typically has smaller-sized units, and almost 50% of those selected for the PDOPPS have between 20 and 30 patients. It will be interesting to see, given the low technique failure rates reported for Japan, if these smaller centers also have less favorable outcomes when compared internationally. This decision to exclude smaller centers also has implications for some countries (e.g., many in Europe), where PD center size is generally too small to join the PDOPPS. In summary, it is increasingly clear that variation in center performance is more important than patient characteristics in determining PD technique failure. This plus the concern around small center size are now sufficiently strong that each country should be addressing this issue by measuring and reporting technique outcomes and where problems are apparent, putting targeted initiatives in place. The ANZDATA group has already shown that this is achievable. Disclosures None.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.