Taiwan has a high burden of chronic kidney disease (CKD) (1). In a comparison of international data from the United States Renal Data System in 2016, Taiwan was reported to have the highest incidence [0.455/1,000 person-years (PY)] and prevalence (0.32%) of end-stage renal disease (ESRD) requiring renal replacement therapy among surveyed countries, and it maintained high levels of incidence and prevalence in subsequent years (2, 3). The trend of choosing peritoneal dialysis (PD) as treatment for ESRD has gradually decreased in Taiwan, accounting for only 6–8% of dialysis treatments from 2000 to 2018. This may be due to a decrease in new-onset PD from 14.2% in 2007 to 9.4% in 2018. Because of this decrease, hemodialysis (HD) has gained prevalence (4).

Compared with HD, PD offers several clinical benefits. First, patients undergoing PD have been reported to have a better 3-year survival rate, although the age-adjusted life span for both treatments was similar (2). Second, the degree of residual kidney function preservation and quality of life has been reported to be more favorable in PD than in HD (5, 6). Third, PD is more cost-effective than HD (7, 8). Accordingly, PD has been widely recognized and employed as the first therapy of choice (9).

However, several limitations and risks have been reported for PD that lead to patients requiring a transition to HD. The most prevalent risk is PD-related peritonitis, which is the reason for 28–35% of PD dropouts (10). In addition, gradual loss of residual kidney function and progressive reduction of peritoneal function resulting from PD may cause insufficient clearance and ultrafiltration (11). These limitations often negate the initial benefits of PD, and approximately 35% of patients undergoing PD drop out of the therapy and require transition to HD (12).

Because transition from PD to HD is inevitable for patients with ESRD, no study has yet analyzed outcomes by using a real-world dataset. In this study, we compared the outcomes of patients who transitioned from PD to HD with those of patients undergoing HD-only using a nationwide population-based dataset.

Through analysis of data from a nationwide database, we demonstrated the outcomes of patients with ESRD who transitioned from PD to HD due to PD failure. The results indicate no significant differences in hospitalization and MACE in the transition from PD to HD group compared with the HD-only group. However, a higher rate of all-cause mortality was observed in the transition from PD to HD group, with the significance of this association remaining consistent in most subgroups. Moreover, the risk of mortality was noted to be the highest in the 5 months following dialysis transition. The results of our study not only expand on current understanding of dialysis but also indicate that greater care should be given to patients undergoing HD who have transitioned from PD.

Whether PD or HD is a more favorable treatment option for patients with ESRD has long been a topic of debate (16). Generally, PD offers more favorable preservation of residual renal function (17), quality of life (18), and stability of hemodynamic changes than HD (19). In addition, the survival outcomes did not seem to differ between PD and HD (20, 21). Accordingly, PD has been identified as a favorable first treatment choice. However, such favorability was not indicated in a small randomized controlled trial conducted by Korevaar JC et al. (22). Therefore, clinicians should view both PD and HD as viable options for treating ESRD.

Transitioning from PD to HD is not uncommon during the disease course among patients with dialysis-dependent ESRD. The average mortality-censored medial technique survival for this group is only 3.7 years (11). In a report that used data from the NECOSAD Study Group, the 2-year PD technique survival was 64%, indicating that one-third of the patients had dropped out of the treatment (23). Transition from PD to HD is a common and gradual process in patients with ESRD who undergo PD as their first dialysis modality (11). Jaar et al. (24) conducted a prospective study of 262 patients and reported no difference in survival between a PD group that transitioned to HD and a PD group that did not (HR, 0.89; 95% CI: 0.41–1.93). Chen et al. by using Australian and New Zealand Dialysis and Transplant (ANZDATA) registry data, reported infection and social-related technique failure to be associated with deaths within 2 years of technique failure in patients undergoing PD (25). To our knowledge, no study has compared the outcomes of transitioning from PD to HD with those of HD-only.

In our study, transitioning from PD to HD yielded the same clinical outcomes as (all-cause hospitalization, infection-hospitalization, and MACE) but poorer survival outcomes than HD-only. Furthermore, the piecewise HR showed that such worse outcome occurred especially in the first 5 months after transitioning to HD. This is possibly because the early period after transitioning to HD are key with respect to premature mortality in such a group, which is consistent with the results of Chen et al. (25). Therefore, a personalized care program for patients undergoing HD who have transitioned from PD may be warranted to prevent mortality.

In the subgroup analysis, we found that older adults had the poorest mortality outcomes (HR: 2.14; 95% CI, 1.79–2.56). Dialysis in older patients presents several challenges; the choice of dialysis modality for older patients is mostly dependent on caregiver (such as family) preferences. In the literature, whether mortality is higher in PD than in HD is highly debated. Several reports have indicated no difference in mortality between PD and HD in older patients (26, 27). However, some studies have reported that the mortality rate of patients undergoing PD was higher than that of those undergoing HD (28, 29). In a study conducted by Sakai K et al. older adults undergoing PD had similar rates of peritonitis or catheter-related complications but lower rates of transition to HD compared with adult patients undergoing PD (30). This may explain our findings. The conditions and complications of transitioning to HD may have been severe for the older adults included in our study, which may have led to poorer outcomes in the older adults in this group than in those in the HD control group. Moreover, in our study, the association was non-significant in patients with severe comorbidities, such as CAD, CHF, PVD, and hyperlipidemia. These comorbidities may, therefore, be hypothesized to play a major role in mortality in older adult populations transitioning from PD to HD. Atherosclerosis and stenosis of the large vessels could result in reduced blood flow to the smaller arteries and arterioles of the peritoneum leading to reduced ultrafiltration and clearance, which resultantly to PD technique failure (31). In addition to the above hypothesis, the other possible explanation for these insignificant results is the small sample size of these subgroups in our study.

Notably, we also found that patients with prior stroke had higher mortality rates when they had transitioned from PD to HD (HR: 1.80; 95% CI: 1.36–2.38). According to a retrospective study by Wu et al. patients who underwent continual ambulatory PD with prior stroke had a higher rate of all-cause mortality and death-censored technique failure compared with those without prior stroke. The study further reported older age and nutritional status to be independent risk factors (32). In this cohort study, patients with prior stroke undergoing PD had higher rates of peritonitis, leading to a need for transition to HD. Malnutrition has been well documented as a risk factor for mortality in PD (33, 34). Patients who have experienced previous stroke are prone to malnutrition (35). This may be due to underlying chronic inflammation, infection, and an inability to achieve optimal intake. In our cohort, patients with stroke transitioning from PD to HD had higher mortality rates, suggesting that these two factors have additive effects, which may be caused by malnutrition.

Our study had several strengths. First, this was a nationwide population-based cohort study (real-world evidence), which ensured a large sample size and the generalizable results. Second, our follow-up duration was sufficiently long to obtain primary outcomes. Finally, we used propensity scores to remove baseline imbalances and adjusted for potential confounders, which improved the reliability of our inferences. However, our study also has some limitations. First, data on potential confounding factors, such as body mass index, blood pressure, laboratory data, lifestyle, and dialysis quality, were not available in the NHI database. However, the consistency in the majority of the subgroup analyses indicates that our study results are robust. Second, the reasons for PD failure were also unavailable in the NHI database, which prevented us from conducting further risk analyses for deaths in patients who had undergone PD. Chen et al. reported that infection and social-related PD technique failure are associated with premature death within 2 years in patients who transition from PD to HD (25). Therefore, further study of risk assessment in such a population is required. Third, the medication use between groups was not balanced. The more prescription of RASB, beta-blockers, CCB and anticoagulants in the group transitioned from PD to HD might indicate volume overload, which lead to higher mortality. Nevertheless, further multiple variables adjusted in the data analysis can reduce this bias. Finally, the results derived from this Taiwanese population of patients undergoing HD may require further verification to be applied to other ethnic populations.

Patients transitioning from PD to HD may experience poorer mortality outcomes compared with patients who are HD-only, especially within 5 months of the transition. Although this finding does not evidence the superiority of either PD or HD, it indicates that clinicians transitioning patients from PD to HD should be well trained and well prepared.

The dataset used in this study is held by the Taiwan Ministry of Health and Welfare (MOHW). Due to the General Data Protection Regulation, the dataset is not available on request from the corresponding author. Any researcher interested in accessing this dataset can apply for access. The MOHW will evaluate the application for accessing this dataset. Please visit the website of the National Health Informatics Project of the MOHW (https://dep.mohw.gov.tw/dos/np-2497-113.html).

The studies involving human participants were reviewed and approved by Institutional review board of Shin-Kong Wu Ho-Su Memorial Hospital. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

M-HT, T-NJ, and Y-WF: conceived and designed the experiments. Y-YC and J-TW: performed the experiments. Y-YC: analyzed the data. M-HT and Y-WF: wrote the manuscript. All authors contributed reagents, materials, and analysis tools and approved the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.