Observational, descriptive study, in which we analyzed all incident patients older than 16 years of the Madrid Registry of Renal Patients (REMER) starting RRT during the period between January 1, 2008 and December 31, 2018 without excluding any patient starting RRT during this study period. According to the definitions of the GRER (Group of Registries of renal patients), an incident patient is considered to be anyone who starts any of the dialysis or transplant techniques for the first time, and any subsequent change in technique will be considered a transition. Thus, for example, the KTof a patient on HD would not be an incidence in KTbut transfer or change of technique, the same would occur in a patient who started dialysis after KTdysfunction. The REMER cross-references its information with the database of the National Transplant Organization (ONT) and the National Institute of Statistics (INE in spanish), to refine mortality and follow-up inconsistencies.

Hospitals in the Community of Madrid are classified into three groups according to their portfolio of services and degree of complexity of clinical attention according to the criteria of the General Directorate of Specialized Care.18 At the level 3 or high complexity are the centers with a complete service portfolio that perform some type of transplant. Hospitals with a broad portfolio of services, but without transplants, belong to level 2 or intermediate complexity, and centers with a reduced portfolio of services belong to level 1 or low complexity.

The main variable analyzed is the state in which the patient has been in throughout the follow-up time. These states represent the different types of RRT: HD, PD, and the KT.

The transitions are defined as the progressions between the different available treatments until the final event of death of the patient, and reflect the evolution of advanced kidney disease.

All the variables included in the description of the REMER are included: demographic variables, kidney disease, type of RRT and outcome.19 It is performed a descriptive analysis of the integrated evolution of the different groups of patients either always staying in one technique or changing between them. We analyzed the initial RRT modality relative to the degree of complexity of the hospital and dialysis centers.

As a secondary objective of interest, we described the causes of mortality for each of the groups.

Statistical analysis is initially performed by intention to treat (ITT) and it categorizes the patients according to the baseline technique; in this way, transfers between techniques are prevented from influencing the results. Next, an analysis is performed by the RRT received up to the time of the event, as-treated (AT), to study whether the results vary depending on the changes in modality of RRT. In this way, we defined four subgroups based on the different dialysis transition situations until KT or the death event: only HD, only PD, start HD and go to PD or start PD and go to HD.

The results are expressed as percentages, mean and standard deviation (SD) or median and interquartile range (IQR) according to whether data is following a normal distribution and the nature of the variables. Comparisons between categorical variables are made with chi-square and between quantitative variables with Student’s t or Mann–Whitney U depending on the distribution of the sample. For the analysis of time-dependent events (survival, maintenance in technique), the results have been estimated using the Kaplan–Meier (KM) method and the log-rank test to assess statistical significance. Survival estimates are made by ITT, considering the first RRT performed. To analyze the impact of transitions on survival, only the first transition between the different RRTs is considered.

Cox regression was used to define the multivariate risk models, including those significant covariates in the univariate model or those that modified the model with a pIN/pOUT 0.05/0.1.

When an event modifies the probability of obtaining the event of interest, proportional hazards models are not suitable for estimating the probability of survival (Kaplan Meier (KM) model, Cox). In this case, a competing risks model should be used, which, unlike the KM model, does not overestimate the real rate of the event20.

Regression analysis to assess competing risks has been used to determine the effects of different dialysis modalities and their transitions on mortality (main event) and KT (competing event). We have represented the curves of cumulative events (CIF, cumulative incidence failure), for the events of interest (mortality and KT). All analyzes were performed with Stata 14 (StataCorp2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP.X).

This project has been approved by the CEIm of the Hospital Universitario Puerta de Hierro-Majadahonda (NREF 132/16), by the REMER Technical Committee and authorized by the Ministry of Health of the Community of Madrid under the supervision of the Office of Security of Systems of Health Information (OSSI) as described in previous publications6.

A total of 8971 patients started RRT in the Community of Madrid during the 11 years of follow-up. The majority, 80.3% (7207), were started on HD while 15.6% (1401) started on PD and 4.1% (363) received pre-dialysis KT, with 33.3% of living donor. The percentage of patients who started home hemodialysis (HHD) was 0.3%.

Table 1 shows the characteristics of the patients classified by ITT, with a mean follow-up of 4.38 years (SD 3.08). Patients who started HD were older and had diabetic comorbidity and had a higher mortality. Patients on PD and those receiving KT in pre-dialysis have a higher proportion of patients with polycystic Kidney Disease or glomerular pathology; furthermore, those who received predialysis KTwere the youngest and had the lowest mortality.

A 40% of patients start RRT in high-complexity hospitals, which are reference centers for performing KT and have broader PD programs.

Of the 8608 incident dialysis patients, 30.4% (2190) of HD patients and 51.6% (723) of PD patients received KT during a follow-up of 4.09 years (SD 3.04) on HD and 4.98 years (SD 3.14) on PD.

If we analyze the technique in which the patients were at the end of their follow-up, most of those who were transplanted in pre-dialysis (93.4%) and those who started HD (71.6%) continued with the same technique. However, almost half of the patients who chose PD were transplanted at the end of follow-up, with only approximately one third remaining on PD (Table 1).

More than a third of patients died during an overall follow-up time in RRT of 4.28 years (SD 3.08), with the most frequent cause of death being cardiovascular (37.1%). In our series, patients who received KT initially died almost 5 times less than those who started on HD and 2.4 times less than those who started on PD (mortality on HD of 40.9% during a follow-up of 4.09 years, SD 3.04; mortality in PD of 22.7% during a follow-up of 4.98 years, SD 3.14 and mortality in KT of 8.3% in a follow-up of 5.52 years, SD 3.04).

Table 2 presents the characteristics of the patients according to the type of initial RRT and their first transition. It shows that the patients who only receive HD are the oldest in the group, with lowest rate of transplants and with the highest mortality; at the end of the analysis period, 41.6% of these patients had died. By contrast the patients who only underwent PD were the youngest, with the highest incidence of glomerular pathology and received most transplants, in addition to having the best survival in the series.

Regarding the possibility of changing modality, it is almost 7 times higher in patients who start PD (25.0%) than those who start HD (3.6%).

It is noteworthy that patients who start RRT in one modality and then switch to another resemble patients in the destination technique. Patients who switch from HD to PD, do it soon, in an average time of 0.7 (SD 1.1) years; furthermore, most transitions occur before the end of the first year. They are younger, have lower mortality and receive more KT than patients who are maintained in HD, and they tend to share more characteristics with patients who are only on PD.

Transference of patients from PD to HD occurs late after a mean of 1.5 (SD 1.4) years on PD. They have similar baseline characteristics to patients who only receive HD, although with a higher incidence of glomerular pathology.

Patients who only receive PD are the ones with the highest probability of survival. Thus, in those who start HD the mortality is twice HR 2.5 95% CI [2.2–2.9]; those who start PD and switch to HD have a mortality risk of 1.6 [1.3–2.0], and those who start HD and switch to PD the risk is 1.2 [0.9–1.6]. PD patients also receive more transplants, double that of those maintained in HD; 1.8 times more than those who go from PD to HD and 1.2 times more than those who start in HD and go to PD.

Regarding the number of transitions recorded per patient, 63% (5652) remained in the initial treatment throughout the analysis period and 29% (2612) changed treatment once. At the end of follow-up, 36.8% had died, 58.2% continued RRT, and 5% had a lost follow-up.

The most common technique used to initiate RRT is HD. As time goes by, KT becomes the most prevalent modality in active patients at the end of the third year of follow-up (Fig. 1A).

Fig. 1.

(A) Distribution of the RRT techniques of active patients at the end of each year. (B) Cumulative incidence curve until technique change for those patients who changed from PD to HD and from HD to PD.

PD: peritoneal dialysis; HD: hemodialysis; KT: kidney transplant. PD > HD: patients who start peritoneal dialysis and go on to hemodialysis. HD > PD: patients who start hemodialysis and switch to peritoneal dialysis.

(0.16MB).

Among the patients who change technique (Fig. 1B), those who start HD and switch to PD occurs early (66% of transfers occur before the first 6 months) while the switch from HD to PD is more gradual and lasts longer.

During a mean follow-up of 4.28 (SD 3.04) years, a total of 3303 deaths occurred. In the ITT analysis, after correcting for age, sex and diabetes, survival is higher in patients who initially receive KT , followed by patients on PD and lastly those on HD (Fig. 2A).

A comparison of all-cause mortality between the different forms of RRT and their transitions at the time of the as-treated event, with KT being the reference technique (Fig. 2B), it is observed that patients who receive only HD have higher mortality risk than the rest of the groups during the period of analysis.

Fig. 2.

Cumulative incidence curve for all-cause mortality; (A) according to the Kaplan–Meier model by intention to treat and according to baseline RRT; (B) according to treatment carried out in relation to the RRT sequence. The HR is adjusted for age, sex and diabetic nephropathy, with KT being the reference technique (HR = 1).

PD: peritoneal dialysis; HD: hemodialysis; KT: kidney transplant. PD > HD: patients who start peritoneal dialysis and switch to hemodialysis. HD > PD: patients who start hemodialysis and switch to peritoneal dialysis. CI: confidence interval; HR: hazard ratio.

(0.23MB).

In our series, a high percentage of patients underwent transplantation, especially in the group with PD-only (58.2%) and those switched from HD to PD early (49.4%), which modifies the probability of reaching the mortality event which is the focus of the analysis. For this reason, it is most appropriate to analyze mortality considering KT as a competing risk and compare the subgroups of evolution in predefined RRT: only PD, only HD, PD to HD and HD to PD (Fig. 3). These graphical representations show that the mortality of patients who start RRT in one modality and then change tend to be similar to the patients of the destination technique rather than to those of the original technique.

Fig. 3.

Cumulative incidence curve (CIF) according to the starting technique and the first change in treatment (as-treated).

PD: peritoneal dialysis; HD: hemodialysis; KT: kidney transplant. PD > HD: patients who start peritoneal dialysis and switch to hemodialysis. HD > PD: patients who start hemodialysis and switch to peritoneal dialysis.

(0.21MB).

It is known that patients who start RRT without a previous schedule and often without having a follow-up in the low clearance outpatient clinic, usually initiate HD through a temporary catheter.37 Both, HD and temporary catheters have been associated with higher mortality38–40 and complications such as bacteremia, prolonged hospitalization,41 thrombosis, or central venous stenosis. The REMER registry, the basis of this work, does not include comorbidities or analytical parameters, so we cannot correct our results by these data. On the other hand, our objective is not to analyze the superiority of one form of dialysis over another, but to analyze RRT as an integrated treatment consisting of a sequence of different therapies, in which the initial technique, how and when the modality of treatment is changed could influence the evolution of the patient receiving RRT.

The characteristics of the population of the Community of Madrid may not be the same as in other communities, which limits the extrapolation of our results to other populations. However, one of the advantages of this work is that analyzes 100% of incident patients on RRT in our setting. The analyzed data comes from an official database, that is mandatory to complete and maintained in the historical series, in which internal validity is guaranteed.

Finally, our study is the first to take into account competitive risks when analyzing the evolution according to the initially modality of RRT together with the changes between the different techniques. This is especially relevant since, in the presence of these, KM survival estimates may not be correct and Cox models may lead to imprecise interpretations.

The decision-making process that leads to the choice of a RRT is complex and may be influenced by social circumstances and the health education received by the patient.42 It is known that the lack of preparation of the patient with advanced chronic kidney disease (ACKD) and an urgent initiation of dialysis is associated with lower survival and higher mortality.43,44 With adequate support and pre-dialysis training, it has been estimated that up to 50% of ACKD patients are able to perform self-care dialysis with PD or HDD45. We do not have the reasons why the technique is changed, but we have been able to observe, in our sample, that 66% of the patients who go from HD to PD do so in the first 6 months of treatment. These patients with an early change in their treatment could represent a late choice of technique in patients with urgent initiation of RRT or lack of prior preparation. The timely referral of patients with advanced kidney disease to Nephrology services, and the continued development of ACKD units, could improve these results.