Peritoneal Dialysis International, Vol. 25, pp. 274–284Printed in Canada. All rights reserved.

0896-8608/05 $3.00 + .00Copyright © 2005 International Society for Peritoneal Dialysis

274

PERITONITIS-RELATED MORTALITY IN PATIENTS UNDERGOINGCHRONIC PERITONEAL DIALYSIS

Miguel Pérez Fontán,1,3 Ana Rodríguez-Carmona,1 Rafael García-Naveiro,1

Miguel Rosales,2,3 Pedro Villaverde,1 and Francisco Valdés1

Divisions of Nephrology1 and Public Health,2 Hospital Juan Canalejo;Department of Medicine,3 University of A Coruña, A Coruña, Spain

Correspondence to: M. Pérez Fontán, Servicio de Nefrología,Hospital Juan Canalejo, Xubias 84, 15006 A Coruña, Spain.

mfontan@canalejo.orgReceived 11 August 2004; accepted 15 November 2004.

Peritonitis is a well-known cause of mortality in perito-neal dialysis (PD) patients. We carried out a retrospectivestudy to disclose the clinical spectrum and risk profile of peri-tonitis-related mortality. We analyzed 693 episodes of infec-tious peritonitis suffered by 565 patients (follow-up 1149patient-years). Death was the final outcome in 41 cases (5.9%of episodes), peritonitis being directly implicated in 15.2%of the global mortality and 68.5% of the infectious mortalityobserved. In 41.5% of patients with peritonitis-related mor-tality, the immediate cause of death was a cardiovascularevent. Highest mortality rates corresponded to fungal(27.5%), enteric (19.3%), and Staphylococcus aureus(15.2%) peritonitis. Multivariate analysis disclosed that thebaseline risk of peritonitis-related mortality was significantlyhigher in female [relative risk (RR) 2.13, 95% confidenceinterval (CI) 1.24 – 4.09, p = 0.02], older (RR 1.10/year, CI1.06 – 1.14, p < 0.0005), and malnourished patients (RR 2.51,CI 1.21 – 5.23, p = 0.01) with high serum C-reactive protein(s-CRP) levels (RR 4.04, CI 1.45 – 11.32, p = 0.008) and a lowglomerular filtration rate (RR 0.75 per mL/minute, CI 0.64 –0.87, p < 0.0005). Analysis of risk after a single episode ofperitonitis and/or subanalysis restricted to peritonitis causedby more aggressive micro-organisms disclosed that overallcomorbidity [odds ratio (OR) 1.21, CI 1.05 – 1.71, p = 0.005],depression (OR 2.35, CI 1.14 – 4.84, p = 0.02), and time onPD at the time of the event (OR 1.02/month, CI 1.00 – 1.03,p = 0.02) were other predictors of mortality.

In summary, the etiologic agent is a definite marker ofperitonitis-related mortality but gender, age, residual renalfunction, inflammation (s-CRP), malnutrition, and depres-sion are other significant correlates of this outcome. Most ofthese risk factors are common to cardiovascular and perito-nitis-related mortality, which may explain the high incidenceof cardiovascular event as the immediate cause of death inpatients with peritonitis-related mortality.

Perit Dial Int 2005; 25:274–284 www.PDIConnect.com

KEY WORDS: Peritonitis; mortality; residual renal func-tion; serum C-reactive protein; depression.

Despite signif icant advances in prevention andtherapy during the past two decades, peritonitis is

still a significant cause of morbidity and technique fail-ure among patients undergoing peritoneal dialysis (PD)(1). Moreover, these infections carry a relatively low butdefinite risk of direct mortality, estimated in 1% – 6% ofepisodes in different studies (2). Peritonitis has beenconsidered to play a significant role in as many as onesixth of the deaths occurring in the course of PD therapy(3). The etiologic agent is undoubtedly a strong markerof outcome after a single episode of peritonitis, with thehighest mortality rates recorded after enteric (4–6), fun-gal (7,8), and, to a lesser extent, gram-negative (9,10)and Staphylococcus aureus (10,11) infections; coagulase-negative staphylococci represent the most benign agentsof peritonitis in terms of mortality (9–11).

Many studies have explored the prognostic spectrumof PD-related peritonitis (4–15), in most cases empha-sizing the relevance of factors such as the etiologicagent(s), the severity of the initial clinical presentation,or the efficacy of different therapeutic approaches. Hos-pitalization, peritoneal catheter loss, and technique fail-ure, with dropout to hemodialysis (HD), are the mostfrequently quoted serious consequences of these infec-tions, while direct peritonitis-related mortality has gen-erally been considered as a most severe but secondaryoutcome, due to the relatively low incidence of thisevent. On the other hand, few studies have analyzed thespecific clinical spectrum of peritonitis-related mortal-ity (3,16) and the association of this outcome with base-line or single-episode-related risk factors other than theetiologic agent or the immediate clinical aggressivenessof the infection. It is conceivable that the many debili-tating conditions that frequently concur in dialysis pa-tients may contribute to def ining the outcome of

275

PDI MAY 2005 – VOL. 25, NO. 3 PERITONITIS-RELATED MORTALITY IN PD

PD-related peritonitis, but this possibility has been in-sufficiently studied.

We performed a retrospective study to assess the clini-cal spectrum and risk profile of peritonitis-related mor-tality in a wide population of patients undergoing PD ina single center for a period exceeding 18 years.

POPULATION AND METHOD

We analyzed retrospectively the information recordedin an extensive prospective database of patients under-going PD in our unit between 1 January 1986 and 31March 2004, with the aims of establishing the incidence,clinical spectrum, and risk profile for peritonitis-relatedmortality during the aforementioned period.

STUDY POPULATION

During the study period, 565 patients underwent PDtherapy in our unit. Their main baseline characteristicsare displayed in Table 1. Total follow-up was 1149 pa-tient-years (range 1 – 125 months).

All patients used standard continuous ambulatory PD(CAPD) or automated PD systems. Y-set systems wereprogressively introduced starting in 1987, and only 36patients (6.4%) used traditional spike systems for CAPDexchanges at any time during follow-up.

DEFINITION OF VARIABLES

We defined “peritonitis-related mortality” as thedeath of a patient due to any immediate cause but underone of three circumstances:

1. During the course of a clinically active peritonitis(persistent abdominal pain and/or turbiddialysate);

2. During the week following complete clinical, bac-teriologic, and cytological remission of an episodeof peritonitis;

3. In the case of a refractory peritonitis demandingcatheter removal, before hospital discharge forreinitiation of regular dialysis therapy (PD or HD).

Peritoneal dialysis-related peritonitis was diag-nosed according to standard criteria. The term “com-plicated peritonitis” was used generically to includeany episode of infectious peritonitis deserving anapproach beyond simple antibiotic therapy. We diag-nosed “refractory peritonitis” in the presence of per-sistent clinical activity and/or bacteriologic positivityafter 5 days of appropriate antibiotic therapy. The

TABLE 1Baseline Characteristics of the Study Population.

Values expressed as [n (%)] unless otherwise.

Patients (n) 565Age (years) 57.9±11.1 (range 7–87)Gender (male/female) 310/255

(55/45)Underlying disease

Glomerular 60 (10.6)Interstitial 68 (12.0)Vascular 55 (9.7)Polycystic kidney diseases 42 (7.4)Systemic 23 (4.1)Diabetic nephropathy 171 (30.3)Other/unknown 146 (25.8)

Diabetes 190 (33.6)Major comorbidities 1.0±1.6

None 280 (49.6)1 145 (25.7)>1 140 (24.8)

GFR (mL/min) 4.4±3.2 (range 0–16.0)Malnutrition by Subjective Global Assessment

No 494 (87.5)Moderate 52 (9.2)Severe 19 (3.4)

Obesity (BMI>30 kg/m2) (n=529) 69 (13.0)Depression 48 (8.5)Chronic hemodialysis before PD 38 (6.7)Previous renal transplant 21 (3.7)Immunosuppression 38 (6.7)Mode of PD (CAPD/A PD) 344/221 (61/39)Need for a dialysis partner 265 (46.9)D/P creatininea 0.64±0.13 (range 0.30–0.97)Hemoglobin (g/dL) 10.1±1.7 (range 5.6–15.7)Plasma albumin (g/L) 37.2±5.7

>30 g/L 509 (90.1)≤30 g/L 56 (9.9)

Serum CRP (mg/L) (n=330) 6.7 (range 1–198)<7 mg/L 169 (51.2)≥7 mg/L 161 (48.8)

GFR = glomerular filtration rate; BMI = body mass index; PD =peritoneal dialysis; CAPD = continuous ambulatory PD; APD =automated PD; D/P = dialysate-to-plasma ratio; CRP = C-reac-tive protein.a After 240 minutes baseline peritoneal equilibration test.Values express mean (standard deviation), number of cases(%), or median (range) (C-reactive protein).

definition of “enteric peritonitis” was based on bac-teriologic criteria: growth in dialysate of one or moreanaerobic intestinal bacteria and/or at least twomicro-organisms characteristic of intestinal flora (in-cluding Enterobacteriaceae, Enterococcus spp, andyeasts, among others).

276

PÉREZ-FONTÁN et al. MAY 2005 – VOL. 25, NO. 3 PDI

The determinant role of the etiologic agent on peri-tonitis-related mortality may compound any strategy ofanalysis of risk factors for this complication of PD. Try-ing to overcome this methodological limitation, we per-formed a specific subanalysis restricted to episodes ofperitonitis caused by more aggressive micro-organisms,which were grouped under the term “severe peritonitis.”We included in this subset fungal (primary or second-ary) and enteric peritonitis, as well as infections causedby mycobacteria, Staphylococcus aureus, and Pseudomo-nas aeruginosa and related bacteria.

TREATMENT OF PERITONITIS

Our protocol of empiric antibiotic therapy for PD-related peritonitis has been based on intraperitoneal cip-rofloxacin during the past 16 years, and has beendescribed in detail elsewhere (17). During follow-up, an-tibiotic therapy may be modified according to bacterialantibiotic sensitivity and clinical course. Management ofcomplicated peritonitis includes catheter removal in caseof fungal (immediate) or refractory peritonitis. Indica-tion for exploratory laparotomy during enteric perito-nitis is individualized, based on clinical presentationand/or diagnostic workup (including routine abdominalcomputed tomographic scan). These criteria remainedstable during the study follow-up.

STRATEGY OF ANALYSIS

The analysis was carried out over two secondary datafiles based on patients and individual episodes of peri-tonitis respectively. The first was used to establish thebaseline risk profiles for presenting at least one episodeof infectious peritonitis (any type) or severe peritonitis,and for peritonitis-related mortality. The second file al-lowed mathematic modeling of the risk of death associ-ated with a single episode of peritonitis (any) or severeperitonitis.

The baseline variables considered for analysis in-cluded age, gender, underlying renal disease, diabetesmellitus, mode of PD (assignment according to intention-to-treat criteria in case of change), partial or completeincapacity for self-dialysis (requiring a dialysis partner),peritoneal transport characteristics at the start of PD[as estimated from dialysate-to-plasma ratio (D/P) ofcreatinine after 240 minutes in a standard peritonealequilibration test (PET)], glomerular filtration rate (GFR;estimated from mean renal clearance), major comor-bidities, malnutrition [standard Subjective Global As-sessment (SGA)], obesity (body mass index > 30 kg/m2),depression (ongoing antidepressant therapy at the start

of PD therapy), procedure other than conservativetherapy (HD, previous renal transplant), previous or cur-rent immunosuppressive drug therapy, and hemoglobin,plasma albumin (by automatic analyzer), and serumC-reactive protein (s-CRP) levels (by immunoturbidi-metry; Roche Diagnostics, Mannheim, Germany). Thecomorbidity score was applied as the sum of all the majorcomorbidities (both cardiovascular and non-cardiovas-cular) present at the time of initiation of PD therapy,without correction for severity. Depression and malnu-trition received specific attention and were not com-puted in the general comorbidity score. Given the lowprevalence of severe malnutrition, moderate and severemalnutrition were grouped as one binary variable(present/absent).

For the analysis of risk after a single episode of peri-tonitis, some baseline variables, including age, accu-mulated comorbidity, mode of PD, malnutrition,obesity, antidepressant and immunosuppressivetherapy, and hemoglobin and albumin values, wereupdated to the last control available before the event.Other variables considered for this part of the analysisincluded time on PD at the time of the event, etiologicagent, baseline peritoneal cell count, and some basicclinical markers of aggressiveness (hospital admission,peritoneal catheter removed, laparotomy performed,relapse, superinfection, and simultaneous exit-siteinfection).

All variables were available for study in at least 95%of the cases, with the exceptions of D/P creatinine (n =323), GFR (n = 508), body mass index (n = 529), and s-CRP(n = 330).

STATISTICS

Univariate statistical analysis was based on Student’st-test (unpaired data), chi-square distribution, analysisof variance, and Mann–Whitney’s test. The multivariateapproach included Cox’s proportional hazard model (pa-tient-based risk profiles) and logistic regression analy-sis (peritonitis-based risk profiles). In the former case,patients were censored in case of renal transplant, lossto follow-up, or death due to causes other than perito-nitis (peritonitis-related mortality). The strategy ofmultivariate analysis included creating a best-fit modelfor the explored event and then introducing other vari-ables, alone or in combination, to obtain adjusted esti-mations of the independent relative risk (RR; Cox’smodel) or odds ratio (OR; logistic regression) associatedwith these secondary variables. Statistical analysis wasproduced with the help of the SPSS 12.0 software (SPSSInc., Chicago, Illinois, USA).

277

PDI MAY 2005 – VOL. 25, NO. 3 PERITONITIS-RELATED MORTALITY IN PD

RESULTS

GENERAL OVERVIEW

We recorded 733 episodes of peritonitis during fol-low-up, but 40 cases were excluded from analysis due tononinfectious etiology (eosinophilic, chemical, icodex-trin). The incidence of peritonitis declined progressivelyduring the study follow-up (Figure 1), but the absolutenumber of infections increased during the same perioddue to the growing number of patients undergoing PDin our unit (20 patients treated in 1986, compared with123 in 2003). The etiologic spectrum of infectious peri-tonitis and the corresponding accumulated mortalityrates are presented in Table 2. The distribution of peri-tonitis by etiologic agent was relatively homogeneousin time, with the exceptions of Staphylococcus aureusinfections (21 of 46 episodes before the end of 1990,when mupirocin was introduced for treatment of carri-ers) and enteric peritonitis, which, for unclear reasons,had a peak incidence during the period 1998–2000. Fig-ure 2 depicts the time courses of global and peritonitis-related mortalities. Our criteria for the diagnosis ofsevere peritonitis was met in 159 episodes of peritoni-tis (including 18 superinfections by yeasts) suffered by127 patients.

RISK OF PRESENTING PERITONITIS

Only 267 patients (47.3%) remained free of peri-tonitis by the end of their follow-up. Univariate analy-sis of the baseline risk profile for presenting at leastone episode of peritonitis disclosed that this eventwas more likely in older patients (mean age 60.1 yearsif peritonitis, compared with 55.3 if no peritonitis,p < 0.001) undergoing CAPD (56.4% presented peri-tonitis, compared with 46.4% of patients undergoingautomated PD, p = 0.012), needing a dialysis partner

(58.0% with peritonitis, compared with 47.5% of pa-tients able to perform self-dialysis, p = 0.012), previ-ously treated with HD (71.1% had per itonitis,compared with 51.1% of patients with PD as first renalreplacement therapy, p = 0.014), with a poorer GFR(mean 3.9 vs 4.9 mL/minute, p = 0.001), lower he-moglobin levels (mean 9.9 vs 10.2 g/dL, p = 0.03),higher s-CRP levels (median 10.3 vs 5.2 mg/L, p =0.001 Mann–Whitney), and a longer follow-up on PDtherapy (mean 32.2 vs 16.0 months, p < 0.001); othervariables were not significant (NS).

TABLE 2Etiologic Agents of Peritonitis. Values expressed as [n (%)].

Micro-organism Episodes Mortality

Staphylococcus aureusa 46 (6.6) 7c (15.2)Coagulase-negative staphylococcus 206 (29.7) 1 (0.5)Enterococcus sppa 22 (3.2) 1 (4.5)Streptococcus spp 113 (16.3) 5d (4.4)Other gram-positive bacteria 23 (3.3) 0 (0)Enterobacteriaceae 85 (12.3) 4 (4.7)Pseudomonas aeruginosa and

related gram-negative bacteria 9 (1.3) 1 (11.1)Other gram-negative bacteria 11 (1.6) 0 (0)Polymicrobial non-enteric 26 (3.8) 1c (3.8)Enteric 62 (8.9) 12 (19.4)Fungal (primary)b 22 (3.2) 6 (27.3)Tuberculosis 3 (0.4) 2 (66.7)Negative culture 65 (9.4) 1 (1.5)Total 693 (100) 41 (5.9)

a Vancomycin sensitive in all cases.b Other 18 peritonitis by yeasts were the consequence of su-

perinfections of primary bacterial peritonitis. In all, 11/40patients (27.5%) died after fungal peritonitis, either primaryor secondary.

c One superinfection by yeast.d Three superinfections by yeasts.

Figure 2 — Accumulated global and peritonitis-related mor-tality rates in different phases of follow-up.

Figure 1 — Time course of the incidence of peritonitis duringthe study period.

278

PÉREZ-FONTÁN et al. MAY 2005 – VOL. 25, NO. 3 PDI

The baseline risk of suffering at least one instance ofsevere peritonitis appeared to be higher in women(26.0% suffered severe peritonitis, compared with14.6% of males, p = 0.001), older patients (mean age60.3 vs 57.2 years, p = 0.07), needing a dialysis partner(25.2% had severe peritonitis, compared with 14.9% ofpatients able to perform self-dialysis, p = 0.003), previ-ously treated with HD (42.1% compared with 18.5% ofpatients coming from conservative therapy, p = 0.001),with more comorbidities (mean 1.2 vs 0.9 per patient,p = 0.013), poorer GFR (mean 3.2 vs 4.7 mL/min, p =0.001), a depressive background (31.3% compared with18.6% of non-depressed patients, p = 0.016), lower he-moglobin levels (mean 9.6 vs 10.2 g/dL, p = 0.02),higher baseline s-CRP levels (median 12.2 vs 6.5 mg/L,p = 0.06 Mann–Whitney), and a longer follow-up on PDtherapy (mean 32.9 vs 22.0 months, p < 0.001); othervariables were NS.

The results of multivariate analysis of the risk profilesfor presenting at least one episode of peritonitis or se-vere peritonitis are presented in Table 3. A baseline s-CRPlevel ≥7 mg/L carried a relative risk of 1.62 (95% confi-dence interval 1.10 – 2.38, p = 0.015) of presenting atleast one episode of peritonitis during follow-up.

SPECTRUM OF PERITONITIS-RELATED MORTALITY

In all, 41 patients died as a consequence of peritoni-tis. In 24 cases (58.5%), peritonitis was considered themain cause of death; in 13 other cases (31.7%), the in-fection had a clearly significant role as a cause of deathbut other previous conditions, including dementia/fail-ure to thrive (n = 9), cardiovascular disorders (n = 4),and neoplastic/ischemic abdominal disease (n = 4), con-

tributed significantly to the final outcome. Finally, in4 other cases (9.8%), the role of peritonitis was uncer-tain but probably not determinant as a cause of death.Table 4 displays the immediate circumstances of deathin the 41 patients who suffered peritonitis-related mor-tality. Remarkably, the immediate cause of demise was acardiovascular event in 41.5% of the cases. Six patientsdied suddenly during the course of a HD session duringthe days following removal of the peritoneal catheter.

BASELINE RISK PROFILE OF PERITONITIS-RELATED MORTALITY

Table 5 depicts the main baseline characteristics ofpatients with peritonitis-related mortality, comparedwith survivors and patients dying for reasons other thanperitonitis. The results of multivariate analysis of thebaseline risk profile for peritonitis-related mortality arepresented in Table 6. A baseline s-CRP level ≥7 mg/L por-tended a RR of 4.04 (95% confidence interval 1.45 –11.32, p = 0.008) of peritonitis-related mortality.

TABLE 4Immediate Cause of Death in Patients with

Peritonitis-Related Mortality (n = 41)

Cause N (%)

Peritonitis-related sepsis 12 (29.3)General deterioration/withdrawal from dialysis 12 (29.3)Sudden death/lethal arrhythmia 8 (19.5)Acute pulmonary edema 4 (9.8)Myocardial infarction 2 (4.9)Lower limb gangrene 2 (4.9)Secondary mesenteric ischemia 1 (2.4)

TABLE 3Baseline Risk Profile for Presenting at Least One Episode of Infectious

Peritonitis (All) or Severe Peritonitis. Multivariate Analysis.

All SevereRR 95% CI p Value RR 95% CI p Value

Age (per year) 1.01a 1.00–1.02 0.04 1.01a 1.00–1.03 0.09Female sex 1.04 0.82–1.26 0.88 2.03a 1.29–3.16 0.002Automated PD (vs CAPD) 0.85a 0.59–0.95 0.02 1.03 0.67–1.63 0.90Glomerular filtration rate (per mL/min) 0.96a 0.92–0.99 0.034 0.88a 0.81–0.95 0.003Serum C-reactive protein (log10) 1.56a 1.12–2.19 0.009 1.39 0.93–2.51 0.13Depression 1.13 0.85–1.50 0.37 1.44a 0.99–2.13 0.052Incidence of peritonitis (per episode/patient-year) — — — 1.62 1.34–2.04 0.0005

RR = relative risk; 95% CI = 95% confidence interval; PD = peritoneal dialysis; CAPD = continuous ambulatory PD.a Included in best model.Cox’s model. The table includes only the variables that were significant in at least one model.

279

PDI MAY 2005 – VOL. 25, NO. 3 PERITONITIS-RELATED MORTALITY IN PD

RISK PROFILE OF MORTALITY AFTER A SINGLE EPISODE OFPERITONITIS

Univariate analysis of the risk profile for mortalityafter a single episode of infectious peritonitis indicatedthat this outcome was associated with older age (mean73.3 years for lethal vs 61.7 for nonlethal infection, p <0.001), female sex (63.4% of lethal infections vs 45.8%of nonlethal infections, p = 0.02), and overallcomorbidity (mean 1.6 points if lethal infection, com-pared with 0.9 if nonlethal infection, p = 0.02) and mal-nutrition at the time of the episode (31.7% of lethalinfections vs 12.2% of nonlethal infections, p = 0.001).As expected, lethal peritonitis was accompanied by fea-tures of clinical aggressiveness (hospital admission, peri-toneal catheter removal, laparotomy performed; all p <

0.01 vs nonlethal peritonitis). Finally, there was a minortrend to higher peritoneal leukocyte counts in lethalperitonitis (median 1865 cells/mm3, compared with 1280in nonlethal peritonitis; p = 0.10, Mann–Whitney).Table 7 presents the results of multivariate analysis ofrisk factors of mortality after a single episode of perito-nitis (all) or severe peritonitis.

DISCUSSION

Our results show that peritonitis represents a signifi-cant cause of mortality in patients undergoing PDtherapy. The accumulated peritonitis-related mortalityrates observed in our center (15.2% of deaths and 68.5%of infectious deaths) match well with values reportedpreviously (3). On the other hand, peritonitis-related

TABLE 5Baseline Risk of Peritonitis-Related Mortality: Comparison with Patients

Dying from Other Causes and Survivors. Univariate Analysis.

Peritonitis-related mortality Death from other causes Survivors(n=41) (n=200) (n=324)

Age (years) 71.1±9.2 64.7±10.5a 51.8±17.3a,d

Gender (male/female) 15/26 (36.6/63.4) 120/80 (60/40)b 175/149 (54.0/46.0)e

Diabetes 11 (26.8) 94 (47)c 87 (26.8)Comorbidities (n) 1.1±1.2 1.4±1.2 0.6±1.0f

Glomerular filtration rate (mL/min) 2.5±2.2 3.9±3.2b 4.9±3.2f

Malnutrition 11 (26.8) 38 (18.9) 22 (6.8)c

Obesity 4 (9.8) 22 (12) 43 (13.3)Depression 7 (17.1) 24 (12.0) 17 (5.2)Previous hemodialysis 5 (12.2) 17 (8.5) 16 (4.9)Previous renal transplant 0 8 (4.0) 13 (4.0)Immunosuppression 3 (7.3) 14 (7.0) 21 (6.5)Mode of PD (CAPD/APD) 23/18 (56.1/43.9) 130/70 (65.0/35.0) 191/133 (59.0/41.0)Need for a dialysis partner 33 (80.5) 126 (63.0)b 106 (32.7)c

D/P creatinine: baseline PET 0.68±0.09 0.68±0.11 0.63±0.13d

Hemoglobin (g/dL) 9.5±1.8 9.8±1.6 10.3±1.8f

Baseline plasma albumin (g/L) 37.1±5.5 36.2±5.5 38.0±5.7Serum C-reactive protein (mg/L) 19.4 (3.3–98) 10.3 (1.0–198) 5.2 (1.0–48)c

Follow-up on PD (months) 25.1±20.7 26.6±21.8 23.2±20.9Time to first peritonitis (months) 11.0±9.7 15.7±15.0b 14.3±15.0Number of peritonitis (per patient) 2.4±2.5 1.2±1.7b 1.2±1.8b

PD = peritoneal dialysis; CAPD = continuous ambulatory PD; APD = automated PD; D/P = dialysate-to-plasma ratio; PET = perito-neal equilibration test.a p < 0.001 versus peritonitis-related mortality.b p < 0.01 versus peritonitis-related mortality.c p < 0.01 versus any other group.d p < 0.02 versus mortality for other causes.e p < 0.05 versus peritonitis-related mortality.f p < 0.05 versus any other group.Figures denote number of cases (%), mean±standard deviation, or median (serum C-reactive protein). Comparisons by chi-square,Student’s t-test, and Mann–Whitney test.

280

PÉREZ-FONTÁN et al. MAY 2005 – VOL. 25, NO. 3 PDI

TABLE 7Mortality After a Single Episode of Peritonitis (All) or Severe Peritonitis. Multivariate Analysis. Logistic Regression Analysis.

All peritonitis (n=693) Severe peritonitis (n=159)OR 95% CI p Value OR 95% CI p Value

Age (per year) 1.11a 1.06–1.15 0.0005 1.15a 1.07–1.23 0.0005Female sex 2.41a 1.18–4.83 0.015 2.08 0.77–5.59 0.14Comorbidities (per point) 1.22 0.89–1.44 0.15 1.21a 1.05–1.71 0.005Malnutrition 2.89a 1.71–5.17 0.0005 1.30 0.57–2.95 0.54Depression 2.35a 1.14–4.84 0.02 2.02 0.72–5.69 0.19Time on PD until peritonitis (per month) 1.02a 1.00–1.03 0.02 1.02a 1.00–1.05 0.04Automated PD (vs CAPD) 1.48 0.75–2.94 0.26 1.25 0.46–3.38 0.67Need for a dialysis partner 1.16 0.54–2.48 0.71 0.74 0.27–2.05 0.56Diabetes 0.68 0.32–1.46 0.33 0.60 0.21–1.70 0.34Obesity 1.44 0.68–3.6 0.40 1.40 0.59–4.29 0.70Immunosuppression 0.75 0.22–2.63 0.61 0.62 0.11–3.51 0.59Plasma albumin (per g/L) 1.07 0.98–1.17 0.20 1.09 0.99–1.20 0.07First episode of peritonitis 0.97 0.58–1.62 0.91 0.86 0.38–1.93 0.71Hospital admission 44.85 16.11–124.88 0.0005 10.38 2.83–38.05 0.0005Catheter removed 12.89 6.10–27.23 0.0005 2.87 1.09–7.55 0.03Laparotomy performed 17.78 6.04–52.30 0.0005 6.95 1.86–26.03 0.004Relapse 1.08 0.53–2.21 0.84 0.64 0.20–2.06 0.46Superinfection 2.10 1.15–3.82 0.016 0.90 0.43–1.68 0.64Simultaneous exit-site infection 2.78 0.75–10.29 0.13 1.04 0.21–5.06 0.97Dialysate cell count at the start of peritonitis (log10) 1.74 0.77–3.17 0.19 2.26 0.71–6.15 0.16

OR = odds ratio; 95% CI = 95% confidence interval; PD = peritoneal dialysis; CAPD = continuous ambulatory PD.a Variable included in best model.

TABLE 6Baseline Risk Profile for Peritonitis-Related and All-Cause Mortalities. Multivariate Analysis. Cox’s Model.

Peritonitis-related mortality All-cause mortalityRR 95% CI p Value RR 95% CI p Value

Age (per year) 1.10a 1.06–1.14 0.0005 1.03a 1.02–1.04 0.0005Female sex 2.13a 1.24–4.09 0.02 1.07 0.81–1.41 0.65Glomerular filtration rate (per mL/min) 0.75a 0.64–0.87 0.0005 0.89a 0.84–0.93 0.0005Malnutrition 2.51a 1.21–5.23 0.011 1.65a 1.17–2.32 0.004Serum C-reactive protein (log10) 3.51a 1.46–8.45 0.005 1.25a 1.05–2.55 0.03Incidence of peritonitis (per episode/patient/year) 1.88 1.49–2.36 0.0005 1.13 0.98–1.31 0.08Diabetes 0.95 0.47–1.94 0.89 1.41a 1.07–1.86 0.014Comorbidity (per point) 1.11 0.88–1.40 0.37 1.23a 1.10–1.36 0.0005Depression 1.65 0.90–3.01 0.11 1.64a 1.25–2.15 0.0005Immunosuppression 1.43 0.42–4.83 0.57 1.76a 1.03–2.99 0.038Plasma albumin (per g/L) 1.01 0.94–1.09 0.73 0.97a 0.94–0.99 0.014D/P creatinine after 240 minutes; baseline PET (per point) 1.03 0.96–1.10 0.41 1.03a 1.00–1.05 0.018Automated PD (vs CAPD) 1.33 0.71–2.50 0.38 0.93 0.70–1.24 0.63Need for dialysis partner 1.61 0.65–4.00 0.30 1.25 0.92–1.70 0.15Previous hemodialysis 1.85 0.62–5.59 0.27 1.04 0.58–1.87 0.88Obesity 1.50 0.49–5.50 0.42 0.55 0.29–1.11 0.10Hemoglobin (per g/dL) 0.83 0.67–1.10 0.15 0.95 0.87–1.04 0.24

RR = relative risk; 95% CI = 95% confidence interval; PD = peritoneal dialysis; D/P = dialysate-to-plasma ratio; PET = peritonealequilibration test; CAPD = continuous ambulatory PD.a Variable included in best model.

281

PDI MAY 2005 – VOL. 25, NO. 3 PERITONITIS-RELATED MORTALITY IN PD

mortality rates did not change markedly during a longstudy follow-up (Figure 2) and did not improve signifi-cantly on results reported in older studies (3,16). Thereare two possible explanations for this seemingly unex-pected finding. First, technical advances in PD therapyhave been relatively successful at preventing some of theless aggressive forms of peritonitis (including coagulase-negative staphylococci infections) but, with the possibleexception of S. aureus (1), their impact on the incidenceof the most severe forms of peritonitis (namely fungaland enteric infections) has been small, if any (2). Sec-ond, the clinical condition (age, diabetes, comorbidity)of patients starting PD therapy has declined markedly inpast years (18). The combination of relatively higher in-cidences of more aggressive forms of peritonitis affect-ing progressively more debilitated patients may underliethe lack of improvement in peritonitis-related mortalityrates, as observed in our study.

Our study confirms also the definite role of the etio-logic agent in the outcome of PD-related peritonitis. In-fections by S. aureus (10,11), Pseudomonas aeruginosa(3,9,14), fungi (3,7,8), mycobacteria (19), and entericmicro-organisms (5,6,20,21), which together formed thesevere peritonitis group, are well known for their sig-nificant morbidity and mortality rates. On the contrary,single gram-negative bacteria and vancomycin-sensitiveEnterococcus spp may be less consistently associated witha complicated course and mortality after peritonitis, asshown by our own results (Table 2); consequently, weexcluded these two groups from the subset of severeperitonitis.

It is noteworthy that, in 41.5% of patients with peri-tonitis-related mortality, the immediate cause of deathwas a cardiovascular event, representing indirect evidenceof the impact of preexisting cardiovascular disease on theoutcome of PD-related peritonitis. This association hasbeen suggested by some previous studies (3,16,20) andour results provide further evidence of the high incidenceof cardiovascular mortality during the period surround-ing an episode of severe peritonitis. Our analysis of therisk profile for peritonitis-related mortality provides newclues to understanding this correlation.

In past years, several studies have clearly demon-strated that residual renal function is a strong correlateof overall survival of PD patients (22–25). Our study isthe first to demonstrate that baseline GFR has a specificand consistent inverse correlation with the later inci-dence of peritonitis (Table 3) and peritonitis-relatedmortality (Table 6). It is evident that a lower number ofbag exchanges in patients with a significant GFR mayreduce the chances for contamination and subsequentinfection, but this explanation seems clearly insufficient

to justify such a strong correlation. On the other hand,low or absent GFR at the start of dialysis is a hallmark ofdelayed referral (a variable not explored in this study)(26) and secondary selection of PD (after chronic HD),and is associated with inadequacy of dialysis, malnutri-tion (22), inflammation (27), and volume overload (28),among other complications. It is true that the correla-tion between GFR and peritonitis outcomes persistedafter controlling for the effects of malnutrition (SGA)and s-CRP levels (Tables 3 and 6), but residual renal func-tion may still have predicted the risk of peritonitis byperforming as a marker of poor medical and nonmedicalconditions of PD patients.

Baseline s-CRP levels were strongly predictive of thelater risk of presenting both peritonitis and severe peri-tonitis (Table 3), as well as of peritonitis-related mor-tality (Table 6). Previous attempts to establish thiscorrelation have been unsuccessful (29), but peritonitisis generally recognized as a potential source of inflam-mation in patients undergoing PD therapy (30). Severalstudies (29,31–34) have explored the correlation be-tween s-CRP and general mortality in PD patients, con-firming the well-known association between this markerof inflammation and cardiovascular risk in chronic renalfailure (35). Our study provides the first evidence of anassociation between a baseline inflammatory state andthe later incidence of peritonitis and peritonitis-relatedmortality.

The background of the correlation between a base-line inflammatory state and the following risk of perito-nitis and peritonitis-related mortality in PD patients isnot clear. We cannot even discard a chance association,as some subsets of patients could present, for indepen-dent reasons, increased s-CRP levels and high incidencesof peritonitis. Alternatively, inflammatory states arestrongly associated with cardiovascular disease, malnu-trition, and, in general, clinical deterioration of affectedpatients (30). Thus, similar to the case of GFR, inflam-mation may define a subgroup of debilitated patientsprone to more frequent infections and complicated clini-cal courses. A third, more speculative hypothesis may bebased on the potential links between inflammatorystates and intestinal permeability. First, it is conceiv-able that a functionally (e.g., due to wall edema in vol-ume overloaded patients) or anatomically (e.g., colonicdiverticulosis) disrupted intestinal wall, with recurrentmicrobial transmigration, may be a source of inflamma-tion in some patients. On the other side, we know, mainlyfrom studies in critical patients (36), that proinflam-matory cytokines may induce dysfunction of the gutmucosal barrier. In either of those cases, high s-CRP lev-els could be, in some patients, a marker for a disabled

282

PÉREZ-FONTÁN et al. MAY 2005 – VOL. 25, NO. 3 PDI

antimicrobial intestinal barrier, with a potentially in-creased risk of peritonitis by intestinal micro-organismsduring PD therapy. Our results are consistent but do notprovide evidence favoring this hypothesis.

Advancing age was strongly associated with the riskof presenting PD-related peritonitis (Table 3). Previousinformation on the relative incidence of these infectionsin elderly patients is controversial [reviewed in Ref. (37)]but our results demonstrate that older PD patients ex-perience a definite risk of severe peritonitis (Table 3)and peritonitis-related mortality (Tables 5, 6, 7). On theother hand, our study showed that female patients weremore likely to present severe peritonitis but not all-causeperitonitis (Table 3) than were males. As an expectedconsequence, the risk of peritonitis-related mortalitywas also higher in women (Tables 6 and 7). The explana-tion for this previously unreported finding is not clearbut the female genitourinary tract may act as a poten-tial reservoir for yeasts and other micro-organisms caus-ing aggressive peritonitis during PD therapy.

Depressive disorders are relatively frequent in patientson dialysis and have been demonstrated to be associ-ated with malnutrition (38), hospitalization (39), gen-eral mortality (39,40), and, in the case of PD patients, ahigh incidence of peritonitis (41). Our study confirmeda consistent association between depressive disordersand general mortality in PD patients and showed clear,previously unreported trends to correlation betweenbaseline depression and a following incidence of severeperitonitis (Table 3) and peritonitis-related mortality(Table 6). Moreover, the risk of death after a single epi-sode of peritonitis was significantly higher in patientswith a depressive background (Table 7).

Incidence of peritonitis was a strong independentpredictor of peritonitis-related mortality and showed aclear trend to predicting overall mortality (Table 6),which is in agreement with previous studies (3). On theother side, to our knowledge, the direct correlation be-tween time on PD at the time of the event and peritoni-tis-related mortality (Table 7) has not been reportedpreviously. The incidence of severe peritonitis was par-ticularly high in patients with a background of frequentinfections and a longer follow-up on PD (Table 3), anobservation that may provide an explanation for thatfinding.

Some relevant variables did not behave as predic-tors of peritonitis or peritonitis-related mortality anddeserve some comment. Our results do not indicate anincreased risk of presenting peritonitis in hypoalbumin-emic patients (42), neither do they support a predic-tive role of plasma albumin levels for peritonitis-relatedmortality, as suggested by previous studies based on

univariate strategies of analysis (4,12,16). The rela-tively low prevalence of severe hypoalbuminemiaamong our patients (Table 1) and better performanceby other markers of malnutrition and inflammation(SGA, s-CRP) in the regression models may explain thisdiscrepancy. On the other hand, the debate over therelative risk of PD-related peritonitis in patients withdiabetes mellitus is still open after many years (43),but our results do not show even a minor trend to suchan association. Finally, immunosuppressive drugtherapy was independently associated with all-causemortality (Table 6) but not with the risk of peritonitisor peritonitis-related mortality, and our results do notsupport recent reports (44) suggesting a high risk ofperitonitis-related mortality in patients starting PDwith a failing renal allograft, although, in this case, thevalidity of our conclusions may be limited by an evidentlack of statistical potency.

In summary, peritonitis represents a significant causeof mortality in patients undergoing PD therapy. The etio-logic agent is an essential predictor of peritonitis-related mortality but advancing age, female gender,poor residual renal function, a baseline inflammatorystate (as estimated from s-CRP levels), malnutrition, adepressive background, and a high incidence of perito-nitis are other significant predictors of this complica-tion. Time on dialysis at the time of the event isassociated with increased mortality after a single epi-sode of peritonitis. Most of these risk factors are com-mon to cardiovascular and peritonitis-related mortality,which may help to explain the high incidence of cardio-vascular events as an immediate cause of death in pa-tients with severe peritonitis.

REFERENCES

1. Troidle L, Gorban-Brennan N, Kliger A, Finkelstein F. Con-tinuous peritoneal dialysis-associated peritonitis: a re-view and current concepts. Semin Dial 2003; 16:428–37.

2. Fried LF, Piraino B. Peritonitis. In: Gokal R, Khanna R,Krediet R, Nolph KD, eds. Textbook of Peritoneal Dialysis.Dordrecht: Kluwer Academic; 2000: 545–64.

3. Fried LF, Bernardini J, Johnson JR, Piraino B. Peritonitisinfluences mortality in peritoneal dialysis patients. J AmSoc Nephrol 1996; 7:2176–82.

4. Tzamaloukas AH, Obermiller LE, Gibel LJ, Murata GH, WoodB, Simon D, et al. Peritonitis associated with intraab-dominal pathology in continuous ambulatory peritonealdialysis patients. Perit Dial Int 1993; 13(Suppl 2): S335–7.

5. Piraino B, Hamburger RJ. Complicated multiorganismperitonitis. Perit Dial Int 1995; 15:394–400.

6. Kern EO, Newman LN, Cacho CP, Schulak JA, Weiss MF. Ab-dominal catastrophe revisited: the risk and outcome of

283

PDI MAY 2005 – VOL. 25, NO. 3 PERITONITIS-RELATED MORTALITY IN PD

enteral peritoneal contamination. Perit Dial Int 2002; 22:323–34.

7. Goldie SJ, Kiernan-Troidle L, Torres C, Gorban-Brennan N,Dunne D, Kliger AS, et al. Fungal peritonitis in a largechronic peritoneal dialysis population: a report of 55 epi-sodes. Am J Kidney Dis 1996; 28:86–91.

8. Wang AY, Yu AW, Li PK, Lam PK, Leung CB, Lai KN, et al.Factors predicting outcome of fungal peritonitis in peri-toneal dialysis: analysis of a 9-year experience of fungalperitonitis in a single center. Am J Kidney Dis 2000; 36:1183–92.

9. Bunke CM, Brier ME, Golper TA. Outcomes of single organ-ism peritonitis in peritoneal dialysis: gram negatives ver-sus gram positives in the Network 9 Peritonitis Study. KidneyInt 1997; 52:524–9.

10. Troidle L, Gorban-Brennan N, Kliger A, Finkelstein F. Dif-fering outcomes of gram-positive and gram-negative peri-tonitis. Am J Kidney Dis 1998; 32:623–8.

11. Peacock SJ, Howe PA, Day NP, Crook DW, Winearls CG,Berendt A. Outcome following staphylococcal peritonitis.Perit Dial Int 2000; 20:215–19.

12. Murata GH, Fox L, Tzamaloukas AH. Predicting the courseof peritonitis in patients receiving continuous ambulatoryperitoneal dialysis. Arch Intern Med 1993; 25:2317–21.

13. Tzamaloukas AH, Murata GH, Fox L. Peritoneal catheterloss and death in continuous ambulatory peritoneal di-alysis: correlation with clinical and biochemical param-eters. Perit Dial Int 1993; 13(Suppl 2):S338–40.

14. Krishnan M, Thodis E, Ikonomopoulos D, Vidgen E, Chu M,Bargman JM, et al. Predictors of outcome following bac-terial peritonitis in peritoneal dialysis. Perit Dial Int 2002;22:573–81.

15. Choi P, Nemati E, Banerjee A, Preston E, Levy J, Brown E.Peritoneal dialysis catheter removal for acute peritoni-tis: a retrospective analysis of factors associated withcatheter removal and prolonged postoperative hospital-ization. Am J Kidney Dis 2004; 43:103–11.

16. Digenis GE, Abraham G, Savin E, Blake P, Dombros N,Sombolos K, et al. Peritonitis-related deaths in continu-ous ambulatory peritoneal dialysis patients. Perit Dial Int1990; 10:45–7.

17. Pérez Fontán M, Rosales M, Fernández F, Moncalián J,Rivera CF, Alonso A, et al. Ciprofloxacin in the treatmentof gram-positive bacterial peritonitis in patients under-going CAPD. Perit Dial Int 1991; 11:233–6.

18. Prichard S. Cardiovascular risk in peritoneal dialysis.Contrib Nephrol 2003; 140:82–90.

19. Lui SL, Lo CY, Choy BY, Chan TM, Lo WK, Cheng IK. Optimaltreatment and long-term outcome of tuberculous perito-nitis complicating continuous ambulatory peritoneal di-alysis. Am J Kidney Dis 1996; 28:747–51.

20. Tzamaloukas AH, Hartshorne MF, Gibel LJ, Murata GH.Persistence of positive dialysate cultures after apparentcure of CAPD peritonitis. Adv Perit Dial 1993; 9:198–201.

21. Harwell CM, Newman LN, Cacho CP, Mulligan D, SchulakJA, Friedlander MA. Abdominal catastrophe: visceral in-

jury as a cause of peritonitis in patients treated by peri-toneal dialysis. Perit Dial Int 1997; 17:586–94.

22. Shemin D, Bostom AG, Lambert C, Hill C, Kitsen J, KligerAS. Residual renal function in a large cohort of PD patients:change over time, impact on mortality and nutrition. PeritDial Int 2000; 20:439–44.

23. Bargman JM, Thorpe KE, Churchill DN. Relative contribu-tion of residual renal function and peritoneal clearanceto adequacy of dialysis: a reanalysis of the CANUSA study.J Am Soc Nephrol 2001; 12:2158–62.

24. Rocco MV, Frankenfield DL, Prowant B, Frederick P,Flanigan MJ. Risk factors for early mortality in U.S. peri-toneal dialysis patients: impact of residual renal function.Perit Dial Int 2002; 22:371–9.

25. Termorshuizen F, Korevaar JC, Dekker FW, van Manen JG,Boeschoten EW, Krediet RT. The relative importance ofresidual renal function compared with peritoneal clear-ance for patient survival and quality of life: an analysis ofthe Netherlands Cooperative Study on the Adequacy ofDialysis (NECOSAD)-2. Am J Kidney Dis 2003; 41:1293–302.

26. Churchill DN. Strategies to improve clinical outcomes inperitoneal dialysis patients: delivered dose and membranetransport. Am J Kidney Dis 1998; 32(Suppl 4):S58–62.

27. Pecoits-Filho R, Heimburger O, Barany P, Suliman M,Fehrman-Ekholm I, Lindholm B, et al. Associations be-tween circulating inflammatory markers and residual renalfunction in CRF patients. Am J Kidney Dis 2003; 41:1212–18.

28. Konings CJ, Kooman JP, Schonck M, Struijk DG, GladziwaU, Hoorntje SJ, et al. Fluid status in CAPD patients is re-lated to peritoneal transport and residual renal function:evidence from a longitudinal study. Nephrol Dial Transplant2003; 18:797–803.

29. Noh H, Lee SW, Kang SW, Shin SK, Choi KH, Lee HY, et al.Serum C-reactive protein: a predictor of mortality in con-tinuous ambulatory peritoneal dialysis patients. Perit DialInt 1998; 18:387–94.

30. Kaysen GA. The microinflammatory state in uremia: causesand potential consequences. J Am Soc Nephrol 2001; 12:1549–57.

31. Haubitz M, Brunkhorst R. C-reactive protein and chronicChlamydia pneumoniae infection. Long term predictors forcardiovascular disease and survival in patients on perito-neal dialysis. Nephrol Dial Transplant 2001; 16:809–15.

32. Herzig KA, Purdie DM, Chang W, Brown AM, Hawley CM,Campbell SB, et al. Is C-reactive protein a useful predictorof outcome in peritoneal dialysis patients? J Am SocNephrol 2001; 12:814–21.

33. Chung SH, Heimburger O, Stenvinkel P, Wang T, LindholmB. Influence of peritoneal transport rate, inflammation,and fluid removal on nutritional status and clinical out-come in prevalent peritoneal dialysis patients. Perit DialInt 2003; 23:174–83.

34. Wang AY, Woo J, Lam CW, Wang M, Sea MM, Lui SF, et al. Issingle time point C-reactive protein predictive of outcome

284

PÉREZ-FONTÁN et al. MAY 2005 – VOL. 25, NO. 3 PDI

in peritoneal dialysis patients? J Am Soc Nephrol 2003;14:1871–9.

35. Stenvinkel P, Alvestrand A. Inflammation in end-stagerenal disease: sources, consequences and therapy. SeminDial 2002; 15:329–37.

36. Fink MP. Intestinal epithelial permeability: update on thepathogenesis of gut mucosal barrier dysfunction in criti-cal illness. Curr Opin Crit Care 2003; 9:143–51.

37. Grapsa E, Oreopoulos DG. Continuous ambulatory perito-neal dialysis in the elderly. In: Gokal R, Khanna R, KredietR, Nolph KD, eds. Textbook of Peritoneal Dialysis.Dordrecht: Kluwer Academic; 2000: 419–33.

38. Koo JR, Yoon JW, Kim SG, Lee YK, Oh KH, Kim GH, et al.Association of depression with malnutrition in chronichemodialysis patients. Am J Kidney Dis 2003; 41:1037–42.

39. Lopes AA, Bragg J, Young E, Goodkin D, Mapes D, CombesC, et al. Depression as a predictor of mortality and hospi-talization among hemodialysis patients in the UnitedStates and Europe. Kidney Int 2002; 62:199–207.

40. Kimmel PL, Peterson RA, Weihs KL, Simmens SJ, AlleyneS, Cruz I, et al. Multiple measurements of depression pre-dict mortality in a longitudinal study of chronic hemodi-alysis outpatients. Kidney Int 2000; 57:2093–8.

41. Troidle L, Watnick S, Wuerth DB, Groban-Brennan N, KligerA, Finkelstein F. Depression and its association with peri-tonitis in long-term peritoneal dialysis patients. Am J Kid-ney Dis 2003; 42:350–4.

42. Wang Q, Bernardini J, Piraino B, Fried L. Albumin at thestart of peritoneal dialysis predicts the development ofperitonitis. Am J Kidney Dis 2003; 41:664–9.

43. Misra M, Khanna R. Peritoneal dialysis in diabetic end-stage renal disease. In: Gokal R, Khanna R, Krediet R,Nolph KD, eds. Textbook of Peritoneal Dialysis. Dordrecht:Kluwer Academic; 2000: 647–65.

44. Sasal J, Naimark D, Klassen J, Shea J, Bargman JM. Laterenal transplant failure: an adverse prognostic factor atinitiation of peritoneal dialysis. Perit Dial Int 2001; 21:405–10.