No increase in small-solute transport in peritoneal dialysis patients treated without hypertonic glucose for fifty-four months

PD was first introduced in our center at Calmette Hospital in Lille in 1981. In 1988, following the publication of images illustrating glucose-damaged peritonea [6], we decided to tackle the challenge of minimizing the use of hypertonic glucose solutions in PD patients. Since then, all our patients have indeed started PD using only low-strength (1.36%) glucose solutions. A progressive restriction in dietary salt was advised. A dry night was used when nighttime bags were heavily reabsorbed. Furosemide, up to 500 mg BID, was used routinely to sustain residual renal function. Nephrotoxic drugs, such as aminoglycosides for peritoneal infections or nonsteroidal anti-inflammatory drugs, were systematically avoided. Hypertonic (3.86%) glucose solutions were initiated as late as possible, and systematically stopped at the first sign of peritoneal infection. More than one hypertonic bag per day was never used, and during the first four years of the present study, from 1992 to 1996, patients were transferred to hemodialysis if overhydration persisted. Since its availability in 1997, icodextrin has been started only when necessary to keep fluid balance, hypertonic glucose solutions being kept as a second-line defense against overhydration, particularly in anuric patients. The use of intermediate (2.27%) glucose solutions was rare.

To assess peritoneal permeability, we initially used an equivalent of the glucose PET test, using 3.86% glucose solution, which was described in 1986 as peritoneal absorption curves [7]. It was not concerned with creatinine, and no emphasis was made on ultrafiltration measurements. We later started measuring creatinine as in the PET test, but kept using 3.86 glucose solution. D/P creatinine results were thus not available in the first years of our study. We then found that the interference of glucose with creatinine measurement [8] varied considerably during the study period, probably due to changes of technique at our laboratory, which made use of these data difficult. We also found it difficult to measure ultrafiltration volume reproductively in the thousand of tests performed on the whole cohort during the study period. For all these reasons, we decided to focus on glucose.

The ratio of peritoneal glucose concentrations at 4-h dwell time to peritoneal glucose concentrations at 0-h dwell time (D₄/D₀ ratio) was used as an index of SST. The higher the D₄/D₀ ratio, the lower was the SST, illustrating a hypopermeable peritoneal membrane. Peritoneal permeability was investigated 6 months after starting PD, and then on a yearly basis. In case of peritoneal infection or clinical acute problem, the test was delayed for 2 months.

The Calmette Hospital cohort has been followed prospectively since January 1st, 1992. Demographic characteristics, primary renal diagnosis, comorbidity including diabetic status, and the occurrence of intercurrent events, such as peritoneal infections or PD drop-outs and their causes, were recorded and transmitted to the French Peritoneal Dialysis Registry [9]. Peritoneal infections were defined as the presence of white blood cells (≥100 cells/mL) in the dialysate with clinical symptoms or a positive culture. The occurrence of encapsulating peritoneal sclerosis was also recorded. Additionally, data were collected every 6 months, including plasma albumin and C-reactive protein concentrations, peritoneal and renal clearances of urea and creatinine.

For this current paper, we focused on a follow-up of the Calmette Hospital Cohort until July 1st, 2008. At that time, patients started to use biocompatible solutions (Physioneal®), which created a confounding factor. All incident PD patients attending our center were eligible for this follow-up. Were included all PD patients with at least a 6-month PD treatment (n = 304). Exclusion criteria were PD treatment for congestive heart failure without terminal renal insufficiency (n = 5), and refusal of regular follow-up (n = 6). Thus the Calmette Hospital Cohort included 293 PD patients on regular follow-up within the study period. Follow-up ended in the event of death, transfer to HD or renal transplantation, transfer to another center, recovery of renal function, or at termination of the study.

The aim of the present study was to describe the long-term evolution of peritoneal permeability in our cohort.

We screened all PD patients of the Calmette Hospital Cohort treated for at least 54 months. Such patients should have had one initial peritoneal permeability test at 6 months and then 4 yearly tests. The inclusion criteria for this study were: a minimum of 3 tests during the first 54 months of follow-up, the first one being performed between 6 and 12 months after initiating PD treatment and the last one between 48 and 60 months. Additionally, PD patients treated for more than 7 years were entered into a very long-term follow-up of peritoneal permeability.

The peritoneal glucose exposure was calculated for each patient included in the study. One daily exchange with a hypertonic (3.86%) glucose solution, an intermediate (2.27%) glucose solution, or an icodextrin solution at least 15 days per month accounted for a 1-month exposure for the specified solution.

The study was approved by the Ethics Committee for Medical Research of the University Hospital of Lille (Commission nationale de l’informatique et des libertés: CNIL-number 701012-GD).

Quantitative variables were described by mean and standard deviation (SD) in the case of a normal distribution and by median and interquartile ranges (Q1, Q3) otherwise. The normality of distribution was checked by using descriptive analyses and the Shapiro-Wilk’s test. Qualitative variables were described by frequencies and percentages. Unpaired Student’s t-test was used to compare baseline characteristics of patients treated for at least 54 months with those treated less than 54 months.

Longitudinal changes of SST were investigated over the first 54 months of PD treatment using a linear regression of D₄/D₀ ratio over time in two steps. First, the slope of the linear regression for the D₄/D₀ ratio over time was computed for each patient. The slope coefficient was used as an index of the SST evolution for each patient. The mean of the slopes was compared to zero using a one sample Student’s t-test as the Shapiro-Wilk test didn’t reject the normal distribution hypothesis. Second, an analysis of longitudinal data was performed using a linear mixed model with random coefficients [10]. Linear mixed model is adapted in situations where the number of repeated measurements differs between the subjects, and allows handling of correlations between the repeated measurements. We used random coefficients because the relationship with time was of great interest. In this model, we considered a fixed linear time effect and included a random intercept and a random linear time effect.

The very long-term evolution of SST in patients who were treated with PD for more than 7 years was analyzed in two ways. In the first way, the analysis was performed separately according two periods: before and after 54 months. In each period the distribution of the slope coefficients was compared to a symmetric distribution with a median equal to zero, using an unpaired Wilcoxon test because the Shapiro-Wilk test could not be performed because of a small sample. A linear mixed model was also performed with a linear time effect, a random intercept and a random linear time effect. In the second way, we fitted a linear mixed model on the whole data from inclusion to end of follow-up. In this model, we considered a linear and quadratic time effect as well as a random intercept, a random linear and quadratic time effect.

All statistical calculations were carried out by means of R using the nlme package.

Among the 50 patients treated for at least 54 months, 44 patients fulfilled the criteria for inclusion in our study. None was treated with automated peritoneal dialysis. The patient flow is presented in Fig. 1. Characteristics of included patients are displayed in Table 1 and compared with those of the 243 patients treated less than 54 months. The two groups were similar regarding age, diabetes mellitus, C-reactive protein, diuresis and renal clearances, but serum albumin was significantly higher in included patients (p < 0.001). No differences between the two groups were observed for peritoneal characteristics at 6 months (peritoneal clearances and D₄/D₀). The frequency of peritoneal infection during the first five years of follow-up was similar in included (1 episode every 18 patient-months) and non included patients (1 episode every 16.2 patient-months). Only one case of encapsulating peritoneal sclerosis occurred in the whole cohort during the study period, and none among included patients. Of the 44 included patients, two (4.5%) had a kidney transplantation, 18 (40.9%) were transferred to hemodialysis, 19 (43.2%) died during the study, and five (10.4%) were still alive on PD as of July 2008, end of the study.

During the 54-month study period, the patients were treated for a total of 2376 months, 2058 months without hypertonic (3.86%) glucose solutions, 2247 months without intermediate (2.27%) glucose solutions, and 2000 months without icodextrin solutions. The association of one hypertonic (3.86%) glucose solution and one icodextrin solution was used in only 3 patients, 2 from the start and 1 from month 42. All in all, these patients were treated only with low-strength (1.36%) glucose solutions for 1703 of the 2376 months. Distribution of patients according to dialysis solutions used during the 54-month study period is shown in Fig. 2. All patients used only conventional glucose solutions (Dianeal ®) during the study period.

A total of 202 tests were realized in the 44 patients, i.e. an average of 4.6 tests per patient. These data were used to compute the linear regression of D₄/D₀ ratio over time for each individual, as described above. The mean of slope coefficients was 0.0136/year 95% CI [0.0069; 0.0202] and was significantly positive (p < 0.0001). The coefficient obtained while fitting a linear mixed model with a random intercept and a random time effect was 0.0131/year (95% CI [0.0065; 0.0197]; p < 0.0001). These results are displayed in Fig. 3. The D₄/D₀ ratio had increased progressively and significantly over time on PD, reflecting a significant decrease of SST during the first 54 months of PD treatment.

A total of 12 patients were treated with PD for more than 7 years. In these patients, the mean of slope coefficients for the late period was significantly negative (−0.020; 95% CI [−0.038; −0.003]; p < 0.05). The coefficient obtained while fitting a linear mixed model with a random intercept and a random time effect was also negative (−0.018; 95% CI [−0.030; −0.005]; p < 0.01). This suggested an increase of SST in this late period for these patients. This result is displayed in Fig. 4.

This result was confirmed by the quadratic regression using a linear mixed model applied on all D₄/D₀ measures from PD initiation. The quadratic coefficient was significantly negative (−0.004; 95% CI [−0.007; −0.002]; p < 0.0001) which confirmed that peritoneal permeability decreased first, and then increased during the late period. Results are presented in Fig. 5 and covariance parameters derived from the model are presented in supplementary Additional file 1: Table S1.