Effects of Increased Peritoneal Clearances on Mortality Rates in Peritoneal Dialysis: ADEMEX, a Prospective, Randomized, Controlled Trial

Study Design

We conducted a prospective, randomized, controlled, clinical trial called ADEMEX (ADEquacy of PD in MEXico), which examined the effects of increased PD small-solute clearances on mortality rates among patients with end-stage renal disease (ESRD) who were being treated with continuous ambulatory PD (CAPD). The study protocol was approved by the local clinical research committees of all participating centers, and all study subjects gave written informed consent. Patients were recruited from 24 dialysis centers in 14 Mexican cities. Twenty-one of the dialysis centers were part of the Instituto Mexicano del Seguro Social, and two were part of the Instituto de Seguridad y Servicios Sociales de los Trabajadores del Estado. The remaining center was the Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán in Mexico City. Between June 1998 and May 1999, the study enrolled 965 patients undergoing CAPD, who were randomized into the control group or the intervention group. By design, the study was terminated in May 2001, when the last enrolled patient had completed 2 yr of follow-up monitoring. With an intent-to-treat (ITT) analysis, the total number of patient-months at risk was 10,464.7 patient-mo for the control group (mean follow-up period, 21.6 mo) and 10,629.0 patient-mo for the intervention group (mean follow-up period, 22.1 mo). As-treated follow-up periods averaged 18.9 mo for the control group and 18.8 mo for the intervention group. The study was performed in Mexico because the only prescription available at the time of study initiation was CAPD with four exchanges of 2 L daily.

Power Calculations

With a target sample of 800 patients (400 patients/group), the study was originally powered at 80% to detect a 5% absolute difference in 1-yr survival or, equivalently, a 30% reduction in the mortality rate [relative risk (RR) = 0.70]. This was based on the assumptions that the mortality rate for the control group would be 23 deaths/100 patient-yr, that the accrual period would be 4 mo, and that there would be 2 yr of follow-up monitoring from the date on which the last patient was enrolled. The final sample size of 965 patients exceeded the original study plan and, with the aforementioned assumptions, actually yields a power of 90% (power = 0.895) to detect a 30% reduction in the mortality rate or, alternatively, a power of nearly 75% (power = 0.742) to detect a 25% reduction in the mortality rate (RR = 0.75). The mortality rate of 23 deaths/100 patient-yr was based on the assumptions that new patients would average 17 deaths/100 patient-yr, prevalent dialysis patients would average 25 deaths/100 patient-yr, and 25% of the patients would be new patients. In actuality, the observed mortality rate for the control group was 18 deaths/100 patient-yr (157 deaths in 10,464.7 patient-mo) and accrual required 1 yr, thus extending the total length of the study to nearly 3 yr. Under these actual conditions, the final sample of 965 patients (484 control, 481 test) provides an observed power of 85% to detect a 30% reduction in the mortality rate (RR = 0.70) and nearly 70% power (power = 0.69) to detect a 25% reduction in the mortality rate (RR = 0.75). Therefore, with the extended accrual and observation periods and the increased number of patients, the study achieved an “observed power” sufficient to detect a reduction in the mortality rate of 25 to 30%, which is similar to that targeted in the hemodialysis study (29).

Selection of Patients

Study subjects were recruited primarily via screening of patients with ESRD who were treated with PD at the participating centers. Chronically treated and new (within 3 mo after initiation of PD) patients were eligible for participation if they met the inclusion and exclusion criteria. All patients who were between 18 and 70 yr of age were eligible if they were undergoing CAPD (with a prescription of four daily exchanges of 2 L) and exhibited measured peritoneal creatinine clearance (pCrCl) values of <60 L/wk per 1.73 m², irrespective of their residual renal function. Patients who were unable to give informed consent, were seropositive for hepatitis B or HIV, were receiving immunosuppressive drugs, had active malignancies, abdominal hernias, or cardiac failure, or had experienced a peritonitis episode ≤1 mo before being screened for the study were excluded.

Randomization

A total of 965 subjects in 24 centers were randomly assigned to the intervention or control group (in a 1:1 ratio) through a central randomization center. Subjects in the control group continued with their existing PD prescriptions, which consisted of four daily exchanges of 2 L of standard PD solution (the only prescription available in Mexico). Subjects randomized to the intervention group were prescribed a modified PD regimen to achieve a pCrCl value of 60 L/wk per 1.73 m². Only two prescription modifications to achieve that target were allowed for each patient. The first new prescription was based on body size; patients with a body surface area (BSA) of ≤1.78 m² received a prescription of four exchanges of 2.5 L in 24 h (4 × 2.5 L) and patients with a BSA of >1.78 m² received a prescription of four exchanges of 3.0 L in 24 h (4 × 3.0 L). A BSA of 1.78 m² corresponds to the value separating the highest tertile of the BSA distribution for an unselected population of patients undergoing PD in Mexico from the middle tertile. After 2 mo of this therapy, pCrCl was measured again. If the patients had reached the pCrCl target with the first prescription change, they continued with the same regimen for the remainder of the study, provided that they tolerated the increased fill volume. Patients who failed to reach the clearance goal and who manifested no intolerance for the increased fill volumes were issued a second modified prescription, which was also based on body size. Patients with a BSA of ≤1.78 m² received a second prescription of five exchanges of 2.5 L in 24 h (5 × 2.5 L), with the aid of an automated nighttime exchange device (Quantum; Baxter Healthcare Corp., Deerfield, IL). Similarly, patients with a BSA of >1.78 m² who did not reach the clearance target but tolerated the 3.0-L dwell volumes received five exchanges of 3.0 L in 24 h (5 × 3.0 L), with the aid of an automated nighttime exchange device (Quantum).

Patients with a BSA of >1.78 m² who achieved a pCrCl above the clearance target but were intolerant of the increased fill volumes underwent prescription modification as follows: if the achieved pCrCl was >70 L/wk per 1.73 m², the prescription was modified to four exchanges of 2.5 L; if the achieved pCrCl was between 60 and 70 L/wk per 1.73 m², the prescription was modified to two exchanges of 2.5 L (morning and afternoon) and two exchanges of 3.0 L (evening and night). Patients with a BSA of >1.78 m² who did not reach the clearance target and did not tolerate the 3.0-L dwell volumes received three exchanges of 2.5 L (morning, afternoon, and evening) and two exchanges of 3.0 L with Quantum (night).

Study Endpoints

Death was the primary endpoint for the study. The secondary endpoints were hospitalizations, therapy-related complications, correction of anemia, and effects on nutritional status (as determined on the basis of normalized protein equivalent of total nitrogen appearance [nPNA] measurements and serum albumin, prealbumin, and transferrin concentrations). Primary and secondary outcome events (deaths, hospitalizations, and clinical events) were counted from the time of randomization, whereas clinical and laboratory features were assessed at scheduled intervals after randomization for the control group and after stabilization of the dialysis prescription for the intervention group.

Clinical and Biochemical Assessments

Follow-up visits were scheduled at 2-mo intervals, beginning immediately after randomization for the control group and after stabilization with a final prescription for the intervention group. At baseline and at each of these visits, a clinical history assessment and a physical examination were performed. Laboratory assessments were completed at every other follow-up visit (every 4 mo). At those times, the patients brought in 24-h dialysate and urine collections, for clearance measurements, and blood samples were obtained. Hematologic features, serum electrolyte, calcium, and phosphate levels, aminotransferase levels, and bilirubin concentrations were measured in the local laboratories of the participating centers. Serum samples, as well as aliquots of dialysate and urine, were sent to the coordinating center for measurement of glucose levels, blood urea nitrogen levels, and creatinine levels [for Kt/V and creatinine clearance (CrCl) calculations], as well as cholesterol, triglyceride, HDL, and LDL levels, with conventional techniques (Synchron CX-5 analyzer; Beckman, Brea, CA). Albumin, prealbumin, and transferrin levels were measured by nephelometry (Array; Beckman). Peritoneal transport characteristics were determined at baseline for all patients (while they were receiving a standard 4 × 2 L CAPD regimen) with the dialysis adequacy and transport test (DATT), the results of which were demonstrated in the same population to be closely correlated with the classification of peritoneal transport with the peritoneal equilibration test (PET) (30).

Monitoring

Patient compliance was ascertained via dialysate bag counts during unscheduled home visits, taking into account the number of bags delivered and the expected utilization (based on the prescription). Patient noncompliance was also assessed at each office/clinic visit, in an interview with the attending nurse or physician. Patients were asked how many exchanges they had missed in the week before the current visit, and the results were recorded for analysis. Finally, serum creatinine levels were recorded as a more objective measure of compliance. Adverse events and deaths were monitored by a safety committee and an interim analysis was performed at a scheduled time, to ensure patient safety.

Statistical Analyses

Overall patient survival analysis was performed by using life-table techniques, with comparisons made on the basis of the log-rank test (31,32). Adjustments for baseline differences in demographic characteristics and comorbid conditions and for differences in nutritional parameters were performed by using the Cox proportional-hazards model (31–34). The model makes use of both time-independent and time-dependent covariates. To account for possible nonproportional hazards, a piecewise exponential survival model (or interval Poisson regression) with time-dependent covariates was also used to compare the RR between the intervention and control groups at 6-mo intervals (32). All primary analyses were performed using an ITT approach. In the ITT analysis, all patients randomized to the intervention and control groups were analyzed and compared, regardless of the actual dialysis prescription achieved by each patient (i.e., regardless of whether patients in the intervention group achieved a target pCrCl of 60 L/wk per 1.73 m²). To safeguard against informative censoring, patient follow-up data with respect to death were maintained until the study end date for all patients who switched to alternative forms of dialysis (either hemodialysis or PD). Therefore, in the ITT analysis, patient survival times were censored only when patients received a transplant, experienced a return of renal function, were lost to follow-up monitoring, or completed the study. An as-treated analysis was also performed, in which patients were censored if they withdrew from the study alive on or before May 6, 2001, regardless of the reason for withdrawal. In this analysis, patient data were analyzed according to the treatment group to which the patients were randomized but follow-up monitoring ceased upon death, withdrawal from the study, or May 6, 2001, whichever came first.

In addition to providing an overall comparison of patient survival data, ITT analyses were performed for select subgroups of patients. Two subgroups were formed, on the basis of patient baseline serum albumin concentrations and baseline nPNA values. For serum albumin levels, the two subgroups corresponded to patients with baseline serum albumin levels of <3.0 and ≥3.0 g/dl. With respect to nPNA, the two subgroups corresponded to patients with nPNA values of <0.80 g/kg per d and ≥0.80 g/kg per d. Another subgroup analysis examined whether cases were prevalent or incident at the time of study initiation. Incident cases included all patients for whom dialysis was initiated ≤3 mo before the time of randomization. Analyses were also performed with stratification according to diabetic status (diabetes mellitus or no diabetes mellitus) and age (<50 or ≥50 yr). Separate analyses were performed with adjustments for possible center-to-center differences. Lastly, a separate analysis was performed for patients who were functionally anephric (GFR of <1 ml/min) at baseline.

Hospitalization and infection rates (e.g., hospital admissions per patient-years and peritonitis episodes per patient-years) were analyzed by using Poisson regression techniques for count data. Corrections for overdispersed Poisson counts were incorporated as necessary, including the use of γ-Poisson regression (or negative binomial regression) for peritonitis rates (35,36).

Linear models for repeated measures, including ANOVA and linear mixed-effects models, were used to analyze continuous measurements recorded repeatedly with time (37,38). Both normal theory likelihood methods and semiparametric generalized estimating equations were applied, using regression procedures included in the SAS software system (39). Specifically, a repeated-measures ANOVA, incorporating the effects of time, treatment, and their interaction, was performed with generalized estimating equations. With this model, least-squares means at each time point were computed and compared for the intervention and control groups, and a test was performed to determine whether there was any significant treatment-time interaction. In the absence of such an interaction, overall least-squares means across time were computed and compared for the two treatment groups, using a main-effects ANOVA model. As a precaution against potential bias resulting from nonignorable missing data attributable to patient withdrawal, a conditional mixed-effects regression analysis was performed in which the outcome variable was modeled as a quadratic function of time, conditional on patient withdrawal times (37,38). This model was used to assess whether there were any linear or quadratic trends with time and whether such trends were the same for the intervention and control groups when data were adjusted for patient withdrawal times. Using this model, time-averaged values for the intervention and control groups were also computed and compared on the basis of least-squares means evaluated at the average withdrawal times for the two groups. For both analyses, an exchangeable correlation structure was assumed to account for correlation across observations recorded for the same patient. However, to safeguard against possible misspecification of this assumed correlation structure, all comparisons were performed by using a robust estimate of the standard errors.

For discrete variables, Pearson’s χ² test and Fisher’s exact test were used to compare baseline characteristics (e.g., gender and diabetic status) for patients in the intervention and control groups. The t test and Wilcoxon’s rank-sum test were used to compare baseline differences between the intervention and control groups for continuous measurements (e.g., baseline serum albumin levels and nPNA values).

Finally, a single interim ITT analysis was performed approximately 1 yr after the enrollment phase of the study. The method of O’Brien and Fleming (40) was used to compute the interim P value necessary to terminate the study in favor of the intervention group, demonstrating significantly better patient survival.

Baseline Characteristics of the Patients

Between June 1998 and May 1999, 965 patients undergoing CAPD were enrolled in the study and were randomized into the control or intervention group. Baseline clinical and laboratory parameters are summarized in Tables 1 and 2. The study groups were similar with respect to demographic characteristics, causes of renal disease, prevalence of coexisting conditions, residual renal function, peritoneal clearances before intervention, hematocrit values, and multiple indicators of nutritional status. Total (renal plus peritoneal) CrCl values and total Kt/V values were similar for the control and intervention groups (total CrCl, 61.8 ± 26.3 versus 59.8 ± 20.2 L/wk per 1.73 m²; total Kt/V, 1.95 ± 0.67 versus 1.93 ± 0.57; mean ± SD; P = NS). Less than 40% of the patients in either group reached or exceeded a total CrCl of 60 L/wk per 1.73 m² or a total Kt/V of 2.0. The incidences of preexisting ischemic heart disease (control, 4.3%; intervention, 3.1%) and stroke (control, 1.7%; intervention, 1.5%) were similar for the two groups. These findings illustrate the success of the randomization procedure.

Assessment of Intervention Effects

Patients in the control group continued to receive four daily exchanges of 2 L for the duration of the study. In the intervention group, 64% of the patients were assigned four daily exchanges of 2.5 L and 36% of the patients received four daily exchanges of 3 L at the time of the first prescription. Additional changes were made with the second prescription, with 85 patients (22%) in the intervention group being assigned a fifth daily exchange (with the aid of a nighttime exchange device). In the intervention group, the total prescribed daily dialysate volume was 10 L for 37% of the patients, 11 L for 20%, 12 L for 21%, 12.5 L for 8%, and 15 L for 14%.

Among the patients in the control group, pCrCl and peritoneal urea clearance (pKt/V) values remained constant, at near-baseline levels, for the duration of the study (Figures 1 and 2). In the intervention group, pCrCl and pKt/V values predictably increased and remained separated from the measurements for the control group for the entire duration of the study (P < 0.01). In the intervention group, 59% of the patients achieved a pCrCl of ≥60 L/wk per 1.73 m², and 78% reached a total CrCl at or above this level. A slightly higher percentage of patients in the intervention group (83%) reached or exceeded a total Kt/V of 2.0.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Time courses of peritoneal creatinine clearance (pCrCl) values for the control and intervention groups during the study. The groups became significantly separated soon after randomization and remained distinct during the follow-up period. Values shown are means and 95% confidence limits for the means.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Time courses of peritoneal Kt/V (pKt/V) values for the control and intervention groups during the study. The groups became significantly separated soon after randomization and remained distinct during the follow-up period. Values shown are means and 95% confidence limits for the means.

To further examine the separation in achieved peritoneal clearances between the two groups, we compared the tertile-defining values for pCrCl and pKt/V in the two groups. We computed the average postrandomization pKt/V and pCrCl values for each patient in the intervention and control groups. Using these values, we then computed the corresponding tertiles. The 33rd and 67th percentile pCrCl values were as follows: control, 42.5 and 49.1 L/wk per 1.73 m²; intervention, 53.4 and 60.5 L/wk per 1.73 m², respectively. For pKt/V, the 33rd and 67th percentile values were as follows: control, 1.45 and 1.74; intervention, 1.94 and 2.24, respectively. The clear separation between the two groups is illustrated by the fact that 67% of the patients in the control group exhibited an average pKt/V value of <1.74, whereas 67% of the patients in the intervention group exhibited an average pKt/V value of >1.94.

Although it is unlikely that a systematic bias would occur uniformly and consistently in a randomized trial of this size, measures of compliance were assessed to validate the sustained separation of the two groups in terms of peritoneal small-solute clearances. The numbers of missed exchanges per year per patient were similar for the two groups (control, 15.1 exchanges/yr per patient; intervention, 18.6 exchanges/yr per patient; P = NS), as indicated by the consumption of dialysis solutions. We could not ascertain compliance with the regimen (timing of exchanges and length of dwell), but the intended increases in clearance were consistently achieved during the study. Serum creatinine and blood urea nitrogen levels were consistently lower for the intervention group throughout the duration of the study, whereas the urea generation rates did not differ between the groups (see below).

The profiles of residual renal function during the study were also similar for the two groups. The use of larger fill volumes and particularly the use of the nighttime exchange device for 85 patients in the intervention group resulted in enhanced peritoneal ultrafiltration, compared with the control group.

Primary Outcomes

Despite differences in small-solute clearances, patient survival was similar for the control and intervention groups, as indicated by ITT analysis (Figure 3), with a RR of death (intervention/control) of 1.00 [95% confidence interval (CI), 0.80 to 1.24]. This was also reflected in the similarity of the 6-mo-interval mortality rates computed by Poisson regression analysis. The time-dependent RR (intervention/control) established by Poisson regression analysis for the consecutive 6-mo intervals were as follows: 0 to 6 mo, RR = 1.17 (P = 0.84); 6 to 12 mo, RR = 1.07 (P = 0.91); 12 to 18 mo, RR = 1.07 (P = 0.93); 18 to 24 mo, RR = 0.73 (P = 0.68); 24 to 30 mo, RR = 1.06 (P = 0.96); 30 to 36 mo, RR = 0.93 (P = 0.98). Overall, the control group exhibited a 1-yr survival of 85.5% (CI, 82.2 to 88.7%) and a 2-yr survival of 68.3% (CI, 64.2 to 72.9%). Similar values were observed for the intervention group, with a 1-yr survival of 83.9% (CI, 80.6 to 87.2%) and a 2-yr survival of 69.3% (CI, 65.1 to 73.6%). The as-treated analysis revealed results similar to those obtained with the ITT analysis (overall RR = 0.93; CI, 0.71 to 1.22; P = 0.6121).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Life-table intent-to-treat (ITT) analysis of patient survival, comparing the study groups. The P value was 0.9842 (log-rank test). RR, relative risk; CI, confidence interval.

In additional ITT analyses, mortality rates for the two groups remained similar when patients within each group were stratified according to a variety of measures known to be associated with patient survival (e.g., age, diabetes mellitus, serum albumin levels, nPNA values, and anuria). Age had a significant effect on outcomes for both groups. Patients <50 yr of age exhibited significantly better survival than did those ≥50 yr of age (P < 0.0001) (Figure 4). In both age strata, however, there was no difference in survival between the control and intervention groups. As expected, diabetic status negatively affected survival (P < 0.0001). However, within the diabetic or nondiabetic stratum, no effect of the intervention on survival could be discerned (Figure 5). Serum albumin levels had an effect on survival (P < 0.0001), with higher values conferring a survival advantage (Figure 6). Within the strata for this measure, however, survival was similar for the control and intervention groups. Life-table ITT analyses of patient survival stratified according to nPNA values (>0.8 g/kg per d or <0.8 g/kg per d) and study group revealed a significant nPNA effect (P = 0.0001) but no significant overall treatment effect (P = 0.8514). Similarly, there was no significant difference in patient survival between the intervention and control groups when the two were compared for the subset of functionally anephric patients (GFR of <1 ml/min).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Life-table ITT analysis of patient survival stratified according to age and study group. The age effect was significant at P = 0.0001. The overall treatment effect was NS (P = 0.5146).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5. Life-table ITT analysis of patient survival stratified according to diabetic status and study group. The diabetic status effect was significant at P = 0.0001. The overall treatment effect was NS (P = 0.7797). DM, diabetes mellitus.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6. Life-table ITT analysis of patient survival stratified according to serum albumin levels (S. Alb) and study group. The albumin effect was significant at P = 0.0001. The overall treatment effect was NS (P = 0.5204).

A variety of factors were observed to have no independent effects on survival, and no differences in survival between the intervention and control groups were discernible in subsets stratified according to these measures, including case type (incident versus prevalent), gender, and baseline peritoneal transport characteristics (characterized with the dialysis adequacy and transport test into the four standard classifications of high, high average, low average, and low transporters). The relationship between body size and survival was examined in great detail. Three indices of body size (stratified by tertiles) were examined, i.e., BSA, total body water (estimated by using Watson’s formula [3,15]), and body mass index. No differences in survival between the intervention and control groups were discernible in subsets stratified according to any of these measures of body size. Additionally, no center effect on the results was discerned.

To further examine the potential interactions between total body water and Kt/V, we examined the RR of death on the basis of total body water tertiles and pKt/V quintiles, using the middle total body water tertile and the middle pKt/V quintile as references and ignoring group assignments. This analysis was performed with a time-dependent Cox regression analysis adjusted for gender, age, diabetes mellitus, baseline albumin levels, and GFR. As demonstrated in Figure 7, we observed no evidence of a size effect or a dose effect within the range of values studied.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7. Results of time-dependent Cox regression analysis adjusted for gender, age, diabetes mellitus, baseline albumin levels, and GFR. RR and 95% CI are plotted according to time-dependent pKt/V values (divided into quintiles) and baseline total body water (TBW) values (divided into tertiles).

There were a total of 157 deaths in the control group and 159 deaths in the intervention group. Acute myocardial infarction was the most common cause of death in both groups (control, 22.4%; intervention, 27.8%; P = NS). Greater proportions of patients in the control group died as a result of congestive heart failure (13.4% versus 5.7% in the intervention group, P < 0.05) or a combination of uremia/hyperkalemia/acidosis (12.2% versus 5.1% in the intervention group, P < 0.05). Generalized infections, strokes, and peritonitis were equally common as causes of death in the two groups.

Multivariate Cox regression analysis in this trial demonstrated that several factors were powerful predictors of outcomes for the study population as a whole. Listed in Table 3 are the results of a Cox regression analysis performed by ignoring the effect of the treatment group. The Cox regression model was chosen to mirror, as closely as possible, earlier analyses performed with the Canada-United States (CANUSA) study data, in which the effects of peritoneal and renal clearances on patient survival were assessed independently of each other (47). With the exception of subjective global assessments, the model summarized in Table 3 includes the factors identified in the CANUSA study as being significantly associated with patient survival. Age, diabetes mellitus, serum albumin levels, residual renal function, and, to a lesser extent, nPNA values were all identified as significant factors associated with patient survival.

Secondary Outcomes

The results of selected laboratory measures averaged across the study period are summarized in Table 4. The two groups were similar with respect to most of the measures. pCrCl and pKt/V were different by design, and the values for the two groups remained significantly separated. Serum albumin levels averaged for the duration of the study were slightly higher for the intervention group. This finding was thought to be attributable to the slightly higher baseline value for the intervention group, because the two groups were similar when changes in serum albumin levels from baseline values were considered. A minimal (approximately 100 ml/d) but statistically significant increase in peritoneal ultrafiltration was observed in the intervention group. Similar results were obtained with a conditional linear mixed-effects model, which takes into account the possible effects of nonignorable missing data attributable to patient withdrawal.

Although overall withdrawal and technique survival was similar for the two groups, the specific causes of withdrawal were different. More patients in the control group withdrew from the study because of uremia [24 patients (5%) in the control group versus no patients in the intervention group, P < 0.0001], and more patients in the intervention group withdrew from the study because of peritoneal exchange volume-related discomfort [17 patients (3.5%) in the intervention group versus one patient (0.2%) in the control group, P < 0.001]. Similar numbers of patients received transplants in the two groups [control, 37 patients (7.6%); intervention, 26 patients (5.4%); P = NS] or were lost to follow-up monitoring [control, 48 patients (9.9%); intervention, 42 patients (8.7%); P = NS]. Technique survival, as determined by life-table analysis, was not affected by treatment group, diabetic status, or baseline peritoneal transport characteristics (as assessed with the dialysis adequacy and transport test).

Hospitalization rates were similar for the two groups, in both unadjusted numbers of admissions per patient per year and rates adjusted for age, gender, diabetic status, serum albumin concentration, and previous time on dialysis; only the adjusted analysis is described here (control, 1.03 admissions/patient per yr; intervention, 1.17 admissions/patient per yr; P = 0.166). The adjusted numbers of hospital days were also similar for the two groups (control, 6.8 d/patient per yr; intervention, 7.2 d/patient per yr; P = 0.593), as were the adjusted peritonitis rates (control, 24.4 patient-mo/episode; intervention, 23.3 patient-mo/episode; P = 0.622). Similarly, no differences in adjusted exit site infection rates were observed between the two groups (control, 64.9 patient-mo/episode; intervention, 51.9 patient-mo/episode; P = 0.326).

Appendix

The following institutions and investigators participated in the study: steering committee: Ramon Paniagua, M.D., Dante Amato, M.D., Ricardo Correa-Rotter, M.D., Alfonso Ramos, M.D., Edward Vonesh, Ph.D., John Moran, M.D., and Salim Mujais, M.D.; safety committee: Ramón Paniagua, M.D., Dante Amato, M.D., Aurora Maravilla, M.D., and Ma. de Jesús Ventura; coordination and data management: Ramon Paniagua, M.D., Dante Amato, M.D., Ma. de Jesús Ventura, and Alejandro B. Hinojosa-Rojas; clinical monitors: Ma. de Jesús Ventura, Olga Lozano, Clara Madonia, Silvia Garnica Salazar, and Inés Vaquera-Rodríguez; central laboratory: Ramón Paniagua, M.D., Ernesto Rodríguez, Marcela Ávila-Díaz, and Raquel Becerril-Becerril; advisory committee: Peter G. Blake, M.D., John M. Burkart, M.D., Bengt Lindholm, M.D., Karl D. Nolph, M.D., and Robert E. Wolfe, Ph.D.; managing committee: Laura Trujillo, Magdalena Nemer, and Angel Rodriguez; participating centers and investigators: Ricardo Correa-Rotter, M.D., Mario Cortés-Pérez, Eduardo Quintana-Piña, Josefina Loredo-Alonso, and Yolanda Martínez-Cerca (Department of Nephrology and Mineral Metabolism, Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán, Mexico City); Imelda Hernández-Reyes, M.D., and Ma. Estela Zepeda-Covarrubias (Hospital General 89, Instituto Mexicano del Seguro Social, Guadalajara, Jalisco); Enrique Hernández-Maldonado, M.D., and Magaly Cevallos-Fernández (Hospital de Especialidades, Centro Médico Nacional Ignacio García Téllez, Instituto Mexicano del Seguro Social, Mérida, Yucatán); Marcos Martínez-García, M.D., and Alicia Herrera-Molina (Hospital General 11, Instituto Mexicano del Seguro Social, Jalapa, Veracruz); Guillermo Valadez-Juvera, M.D., and Alma Leticia Valenzuela (Hospital de Especialidades 1, Centro Médico Nacional del Noroeste, Instituto Mexicano del Seguro Social, Ciudad Obregón, Sonora); José Alejandro García-Larumbe, M.D., and Bertha Espinoza-Pérez (Hospital General 30, Instituto Mexicano del Seguro Social, Mexicali, Baja California); José Luis Medina-Gómez, M.D., and Celestina Jiménez-Ronquillo (Hospital de Especialidades, Centro Médico Nacional Manuel Avila Camacho, Instituto Mexicano del Seguro Social, Puebla, Puebla); Eduardo Aguilar, M.D., and Javier Ortiz, Elizabeth Almanza-Medina (Hospital General 1, Instituto Mexicano del Seguro Social, Zacatecas, Zacatecas); José Martínez-Bárcenas, M.D., and Eloisa López-Pinales (Hospital General 33, Instituto Mexicano del Seguro Social, Monterrey, Nuevo León); Ma. Elena Hurtado-González, M.D., and Guadalupe Alcántara-Ortega (Hospital General 25, Instituto Mexicano del Seguro Social, Mexico City); Alejandra Cisneros-García, M.D., Juan Guillermo Oros, M.D., and Josefina Luna-Tapia (Hospital General 27, Instituto Mexicano del Seguro Social, Mexico City); Jorge Prieto-Fierro, M.D., and Leonor Abularach-Nevares (Hospital de Especialidades, Centro Médico Nacional, Instituto Mexicano del Seguro Social, Torreón, Coahuila); Abraham Sanjurjo-Gallardo, M.D., Dolores Gómez-Noriega, M.D., and Ma. de Lourdes Barrón-Andrade (Hospital General 32, Instituto Mexicano del Seguro Social, Mexico City); Mario Enríquez-Rivas, M.D., and Ma. Jesús Galaz-Valdez (Hospital General 2, Instituto Mexicano del Seguro Social, Hermosillo, Sonora); Cristina Avilés-Hernández, M.D., and Ma. de la Cruz Pérez-Hernández (Hospital de Especialidades, Centro Médico Nacional la Raza, Instituto Mexicano del Seguro Social, Mexico City); Gerardo Rodríguez del Villar, M.D., and Ma. Elena Ríos-Estrada (Hospital General 1, Instituto Mexicano del Seguro Social, Durango, Durango); Roberto Baca-Baca, M.D., and Ma. Graciela Pacheco-Hernández (Hospital General 1, Instituto Mexicano del Seguro Social, Querétaro, Querétaro); Ma. Juana Sil-Acosta, M.D., and Ma. Teresa Galindo-Salinas (Hospital General 1 Gabriel Mancera, Instituto Mexicano del Seguro Social, Mexico City); José Amado Chimalpopoca, M.D., and Ma. Isabel González-Yañez (Hospital General 8, Instituto Mexicano del Seguro Social, Mexico City); Francisco Belio-Caro, M.D., and Olivia Figueroa-Rosas (Hospital General 1, Instituto Mexicano del Seguro Social, Morelia, Michoacán); Julio Kaji-Kiyono, M.D., Lucía Isabel Aranda-Contreras, and Virginia Becerril-Morales (Hospital General Regional 1° de Octubre, Instituto de Seguridad y Servicios Sociales de los Trabajadores del Estado, Mexico City); Hugo Breien-Alcaraz, M.D., and Ana Gabriela Ruelas-Peralta (Hospital General Valentín Gómez Farías, Instituto de Seguridad y Servicios Sociales de los Trabajadores del Estado, Guadalajara, Jalisco); Juan Alberto Aguilar-Martínez, M.D., and Leticia Martínez-Guzmán (Hospital General Gaudencio González Garza, Centro Médico Nacional la Raza, Instituto Mexicano del Seguro Social, Mexico City); and Francisco Monteón, M.D., José Luis Camarena, M.D., and Alma Rosa Martínez (Hospital de Especialidades, Centro Médico Nacional de Occidente, Instituto Mexicano del Seguro Social, Guadalajara, Jalisco).