Candida as a risk factor for mortality in peritonitis*

Candida as a risk factor for mortality in peritonitis* Philippe Montravers, MD, PhD; Hervé Dupont, MD, PhD; Remy Gauzit, MD; Benoit Veber, MD; Christian Auboyer, MD; Patrick Blin, MD, MSc; Christophe Hennequin, MD, PhD; Claude Martin, MD

Objective: The clinical significance of Candida cultured from peritoneal fluid specimens remains a matter of debate. None of the studies that have addressed this issue have clearly distinguished between community-acquired peritonitis and nosocomial peritonitis. The current study tried to differentiate the pathogenic role of Candida in these two clinical settings and assess its importance on outcome. Design: A multiple-center, retrospective, case-control study was conducted in intensive care unit patients. The interaction between mortality rates and type of patients was assessed. In the case of a significant interaction, a separate analysis of mortality and morbidity was planned. Setting: Seventeen intensive care units in teaching and nonteaching hospitals. Patients: Cases were patients operated on for peritonitis with Candida cultured from the peritoneal fluid, whereas controls were operated patients free from yeast. Cases and controls were matched for type of infection, Simplified Acute Physiology Score II, age, and time period of hospitalization. Interventions: None. Measurements and Main Results: The following characteristics were collected: demographic variables, underlying disease, se-

A

large amount of data have been published on the role of Candida in intra-abdominal infections, but this remains a controversial issue (1, 2). Although isolation of these organisms was associated with increased mortality and a complicated postoperative course in many stud-

*See also p. 902. From Département d’Anesthesie Réanimation (DAR), CHU Bichat-Claude Bernard, AP-HP, Université Paris VII (PM); DAR, CHU Nord Amiens (HD); DAR, CHU Hotel Dieu, AP-HP (RG); DAR, CHU Charles Nicolle, Rouen (BV); DAR, CHU St Etienne (CA); Département de Pharmacologie, Université Victor Ségalen, Bordeaux (PB); Unité de Génétique Moléculaire des Levures, Institut Pasteur, Paris (CH); DAR, CHU Nord, Marseille (CM). Supported, in part, by a grant from Pfizer Laboratories France. PM, HD, RG, CH, and CM have received speaking fees from Pfizer France and MSD Chibret; BV and CA have received speaking fees from Pfizer France. Copyright © 2006 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: 10.1097/01.CCM.0000201889.39443.D2

646

verity score, site of infection, microbiological features, and antiinfective treatments. Survival was defined as the main outcome criterion and morbidity variables as secondary criteria. Odds ratios of mortality were calculated. Matching was achieved in 91 cases and 168 controls. Matching criteria, clinical characteristics, and mortality rate were not statistically different between cases and controls. A significant interaction was demonstrated between mortality rates and type of infection, leading to separate analysis of patients with community-acquired peritonitis and nosocomial peritonitis. The subgroup analysis demonstrated an increased mortality rate only in nosocomial peritonitis with fungal isolates (48% vs. 28% in controls, p < .01). Upper gastrointestinal tract site (odds ratio, 4.9; 95% confidence interval, 1.6 –14.8) and isolation of Candida species (odds ratio, 3.0; 95% confidence interval, 1.3– 6.7, p < .001) were found to be independent risk factors of mortality in nosocomial peritonitis patients. Conclusions: Isolation of Candida species appears to be an independent risk factor of mortality in nosocomial peritonitis but not in community-acquired peritonitis. (Crit Care Med 2006; 34:646–652) KEY WORDS: peritonitis; Candida; nosocomial infections; community-acquired infection; outcome; pathogenicity

ies (3–7), the need to treat Candida is still controversial, frequently phrased in terms of pathogenic or nonpathogenic organisms and essentially based on clinical grounds. This debate is reinforced by the fact that patients with Candida often die of complications attributed to infection despite antifungal therapy (3, 4). Many confounding factors could influence outcome, such as surgical complications, age, underlying diseases, or subsequent nosocomial infections (6, 8, 9). None of the studies that have addressed the role of Candida in peritonitis have clearly distinguished between community-acquired peritonitis (CAP) and nosocomial peritonitis (NP) or assessed the importance of the type of infection on outcome (4, 6, 7, 10, 11). Most of the recommendations published over recent years remain evasive on this issue (1, 12, 13), rarely giving therapeutic guidelines in these situations. Even the Infectious Disease Society of America’s guidelines for the treatment of complicated intra-

abdominal infections (2), although addressing this issue, were based on articles including an insufficient number of cases to demonstrate a role for the type of infection (3, 4). Based on this concern, the present study was designed to determine the influence on outcome of Candida cultures recovered from intra-abdominal specimens in patients with peritonitis, with special reference to the type of infection. A case-control study was therefore conducted in patients operated and admitted to the intensive care unit (ICU) with a diagnosis of peritonitis comparing patients with and without Candida cultures matched according to the type of infection (CAP or NP) and the severity of their illness.

MATERIALS AND METHODS Study Design. This retrospective, multiplecenter (15 teaching and three nonteaching hospitals), case-control study analyzed patients hospitalized in the ICU between January

Crit Care Med 2006 Vol. 34, No. 3

1997 and December 2000. Centers were selected based on the annual number of patients admitted to the ICU with the diagnosis of intra-abdominal infection and the capacity for routine identification of peritoneal microorganisms. Definitions. All cases and control patients included were adults (ⱖ18 yrs old) hospitalized in ICU who underwent laparotomy for a diagnosis of peritonitis for a disease diagnosed before admission or during the ICU stay. In each center, consecutive patients with a positive culture for Candida species in the peritoneal fluid obtained during surgery were recruited and constituted the cases. For each case in the same study center, the investigators screened at least one and ideally two controls defined as patients with no yeast culture on peritoneal fluid collected during surgery and no yeast culture during the hospital stay. The following hierarchy was used in the matching process: a) type of peritoneal infection (CAP or NP); b) Simplified Acute Physiology (SAPS) II score (14) at the time of surgery ⫾ 10; c) age ⫾10 yrs; d) year of hospitalization ⫾1 yr. The day of operation leading to Candida isolation was considered to be day 0. In cases of multiple reoperations, only the index operation leading to inclusion was considered. Nosocomial intra-abdominal infection was defined as an infection not present on admission that became evident ⱖ48 hrs after admission in patients hospitalized for a reason other than intra-abdominal infection (15). Antibiotic use in the study centers was defined by protocols for prophylactic and curative purposes, but prophylactic antifungal therapy was not used. Exclusion criteria were as follows: primary peritonitis (medical causes of intra-abdominal infection not requiring surgical therapy) (16), infected peritoneal dialysis, acute bowel perforation (ⱕ6 hrs), severe pancreatitis, gynecologic peritonitis, organ transplantation, and neutropenia ⬍500·mm⫺3. Absence of microbiological cultures was also an exclusion criterion. Patients with negative fungal cultures were considered to be controls. Data Collection. A report form was completed for each patient based on review of the medical charts. No informed consent was required by the Institutional Review Board because of the retrospective and nonrandomized nature of the study. The following data were collected: demographic characteristics (age, gender, and weight); and underlying disease (heart failure with New York Heart Association grade ⱖ3, chronic respiratory failure requiring home ventilatory support or oxygen, malignancy, cirrhosis, chronic renal failure with serum creatinine ⱖ300 ␮mol·L⫺1, diabetes mellitus, corticosteroids with prednisone ⬎40 mg/day, and splenectomy). On the day of surgery, the severity of infection was assessed using SAPS II and presence of organ dysfunction using the Organ Dysfunctions and/or Infection (ODIN) score (17). The site of perito-

Crit Care Med 2006 Vol. 34, No. 3

nitis was recorded, and the transverse mesocolon was considered to be the barrier between upper and lower gastrointestinal tract. Microbiological techniques applied for culture and identification were those routinely used in the microbiology laboratory. Antibiotic susceptibility was assessed by the disk diffusion method (18). Identification and susceptibility testing were performed by the hospital mycology department or microbiology laboratory. Microbiological and mycological results of cultured peritoneal fluid collected during surgery were recorded. Positive fungal cultures from extraperitoneal sites at the time of surgery were also recorded. Anti-Infective Treatments. Antibiotic prophylaxis administered during the fortnight before the index episode or curative antibiotic treatment administered during the previous week was recorded. Antimicrobial and antifungal treatments prescribed empirically and after identification and susceptibility testing were also recorded. The investigators noted appropriateness of initial antimicrobial therapy according to the culture results. Empirical antibiotic treatment was considered inappropriate when all the cultured bacteria were not targeted, except for Candida species. No protocol was available in the study centers for the treatment of Candida, and treatment was left to the attending physician’s discretion. Surgical Procedures. Surgical principles were similar in all centers. All patients underwent laparotomy. After peritoneal fluid sampling for microbiology, abundant peritoneal lavage was performed using sterile isotonic sodium chloride solutions. Ostomies were preferred to primary anastomosis. No openwound management was performed, and the abdomen was not irrigated after surgery. The investigators reported all surgical procedures. Outcome. Survival was defined as the main outcome criterion. The following secondary outcome criteria were reported by the investigators: time to death, ICU length of stay, duration of mechanical ventilation, need for relaparotomy, number of reoperations, and digestive and extradigestive complications. Cause of death was assessed by the investigators as related or not related to the initial infection. Statistical Analysis. Statistical analysis was performed using SAS software (version 8.1, SAS Institute, NC). Sample size estimation was based on a 30% control mortality, an odds ratio to 2.5 with a two-tailed ␣ of 5%, and a power of 80%. At least 64 cases and 128 controls were required to achieve this power. Patient characteristics were expressed as mean ⫾ SD or proportions and were compared using analysis of variance, Kruskal-Wallis (ordinal or nonnormal quantitative variables), chi-square, and Fisher’s exact tests. Bonferroni post hoc test was used for multiple comparisons. Interaction between mortality rates and type of patients (CAP and NP) was assessed using analysis of maximum likelihood estimates. In the case of a significant interaction, a separate

analysis of mortality and morbidity was planned in the subgroups (CAP and NP). Odds ratio of mortality and 95% confidence intervals (CIs) were calculated using a conditional forward stepwise logistic regression. Only variables significant at a limit of 5% on univariate analysis were proposed in the multivariate model. A value of p ⬍ .05 was considered significant.

RESULTS Matching Process. During the study period, 109 patients with a positive culture for Candida species in the peritoneal fluid obtained during surgery were selected as eligible cases for the study and 211 patients satisfying the inclusion criteria were selected as controls (mean number of cases and controls per center, 6 ⫾ 4 and 12 ⫾ 9, respectively). The results of matching for the criteria listed previously were as follows: 191 of the 211 controls were matched to cases for diagnosis (90%), 202 controls were matched for severity score (96%), 197 controls were matched for age (93%), and 204 controls were matched for length of hospital stay (97%). Overall, 18 cases (six CAP and 12 NP) and 43 controls were excluded and successful matching was achieved in 259 (81%) patients, constituting the study population: 91 cases (33 [36%] CAP and 58 [64%] NP) and 168 controls (62 [37%] CAP and 106 [63%] NP). The mean age (67 ⫾ 13 and 68 ⫾ 13 yrs, respectively) and SAPS II (49 ⫾ 13 and 47 ⫾ 13, respectively) were similar in cases and controls. The clinical variables compared on admission are presented in Table 1. Previous Antibiotic Treatments. Antimicrobial prophylaxis was administered during the fortnight preceding the index operation in 26 cases and 66 controls (29% vs. 39%, respectively). Antibiotic treatment for suspected or defined infection was administered during the previous week in 50 cases and 55 controls (55% vs. 33%, p ⬍ .01). Microbiological Results. Microorganisms cultured from the peritoneal fluid are shown in Table 2. All cultured yeasts were Candida species (Table 2). Candida species was the only organism cultured in 26 (29%) cases. The mean number of cultured organisms was similar in the two groups (2.6 ⫾ 1.5 vs. 2.4 ⫾ 1.4 in cases and controls, respectively). A total of 48 organisms were cultured from blood (21 and 27 in cases and control groups, respectively), including four episodes of candidemia (three Candida albi647

Table 1. Clinical characteristics of the study population and patients with nosocomial peritonitis expressed as number (%) of patients or mean ⫾ Study Population

Male gender Underlying disease No underlying disease Cardiovascular or respiratory disease Malignancy and immunosuppression Chronic renal failure Cirrhosis Diabetes mellitus Coming from Home Medical care facility Hospital Length of stay in hospital before surgery, days Interval between admission to ICU and surgery, days Signs of severity Cardiovascular failure Acute respiratory failure Renal failure Neurologic failure Liver failure Upper gastrointestinal tract site Stomach/duodenum Biliary tract Lower gastrointestinal tract site Small bowel Colon Other Type of injury Perforation Necrosis Anastomosis leakage No cause Surgical management Bowel resection Bowel anastomosis Ostomy Bowel suture

SD

Nosocomial Peritonitis

Cases (n ⫽ 91)

Controls (n ⫽ 168)

Cases (n ⫽ 58)

Controls (n ⫽ 106)

71 (78)

128 (76)

38 (66)

77 (73)

30 (33) 18 (20) 34 (37) — 2 (2) 1 (1)

55 (33) 18 (11)a 80 (48) 5 (3) 4 (2) 4 (2)

16 (28) 7 (12) 21 (36) — 2 (3) 1 (2)

32 (30) 10 (9) 63 (59) 2 (3) 2 (3) 3 (3)

25 (27) 3 (3) 81 (89) 12 ⫾ 13 3⫾6

62 (37) 3 (2) 144 (86) 12 ⫾ 13 1 ⫾ 3a

— 4 (7) 54 (93) 15 ⫾ 14 4⫾7

— 6 (6) 100 (94) 13 ⫾ 12 3⫾4

11 (12) 84 (92) 4 (4) 13 (14) 6 (7) 48 (53) 37 (41) 11 (12) 58 (64) 25 (27) 27 (30) 6 (7)

15 (9) 149 (89) 12 (7) 19 (11) 9 (5) 44 (26)b 25 (15)b 19 (11) 149 (89)b 45 (27) 95 (57)b 9 (5)

5 (9) 54 (93) 3 (5) 9 (16) 6 (14) 30 (52) 19 (33) 11 (19) 44 (76) 19 (33) 20 (34) 5 (9)

11 (10) 97 (92) 9 (8) 12 (11) 7 (9) 36 (34)a 19 (18)a 17 (16) 93 (88)a 29 (27) 58 (55)b 6 (6)

66 (73) 15 (16) 19 (21) 6 (7)

124 (74) 40 (24) 39 (23) 12 (7)

23 (40) 11 (19) 19 (33) 6 (10)

47 (44) 22 (21) 39 (37) 8 (8)

17 (19) 11 (12) 33 (36) 33 (36)

59 (35)b 17 (10) 105 (63)b 31 (18)b

12 (21) 14 (24) 16 (28) 13 (22)

28 (26) 24 (23) 46 (43)a 17 (16)

ICU, intensive care unit. p ⬍ .05; bp ⬍ .01 vs. cases.

a

cans). At the time of the index operation, Candida species were cultured from extraperitoneal sites in 27 and nine controls (30% vs. 5%, respectively, p ⬍ 0.01), mostly involving urinary (10 [11%] vs. 1 [1%]) and bronchopulmonary (8 [9%] vs. 7 [4%]) sites. Anti-Infective Treatments. Empirical antimicrobial treatment was administered in all patients. Inadequacy of empirical antibiotic treatment was similar in the two groups (37% and 31% in cases and controls, respectively). Empirical antifungal treatment was given in 38 (42%) cases (fluconazole in all but one case) and 14 (8%) controls (fluconazole in all but one case). After identification, antifungal treatment was administered in 57 (63%) cases (fluconazole in 49 patients and amphotericin B in 13 patients). Presence of 648

Candida species in extraperitoneal cultures (19 cases), fungemia (three cases), and absence of clinical improvement despite treatment with broad-spectrum antibiotics (15 cases) were the usual criteria to initiate antifungal therapy. Six patients initially treated by fluconazole subsequently received amphotericin B due to Candida glabrata (n ⫽ 5) and Candida krusei infections, whereas one patient initially treated by amphotericin B received fluconazole because of renal failure. Decreased susceptibility to azoles was reported in six C. glabrata strains and resistance in two C. krusei strains. Prognosis. A trend toward increased mortality was observed in the cases (34 [37%] deaths vs. 45 [26%] in the controls, p ⫽ .077). Other criteria are shown in Table 3. An interaction was demon-

strated between mortality rates and predefined type of patients (CAP and NP, hazard ratio, 5.3; 95% CI, 1.5–19.1; p ⫽ .01) leading to analysis of outcome of these subgroups as planned initially. Subgroup Analysis. In patients with CAP, the mean age (71 ⫾ 13 and 69 ⫾ 13 yrs, respectively) and SAPS II (47 ⫾ 13 and 46 ⫾ 12, respectively) were similar in cases and controls. The frequency of inadequate empirical antibiotic therapy was similar in both groups (seven of 32 [22%] in cases and 22 of 61 [36%] in controls). Mortality rate was similar between cases and controls (six deaths [19%] in cases vs. 15 [24%] in controls). Secondary outcome criteria were similar in the two groups: time to death (17 ⫾ 24 days vs. 15 ⫾ 11 days in controls), ICU length of stay (16 ⫾ 20 days vs. 14 ⫾ 15 days), Crit Care Med 2006 Vol. 34, No. 3

Table 2. Microorganisms isolated from peritoneal fluid expressed as number (%) of total number of organisms in the study population and patients with nosocomial peritonitis Study Population

Nosocomial Peritonitis

Organisms

Cases

Controls

Cases

Controls

Aerobes Gram-positive bacteria Staphylococci Streptococci Enterococcus species Miscellaneous Gram-negative bacteria Enterobacteriaceae Escherichia coli Klebsiella species Enterobacter species Serratia species Proteus and Morganella species Citrobacter species Nonfermenting Gram-negative bacteria Miscellaneous Anaerobes Clostridium species Bacteroides species Miscellaneous Yeasts Candida C. albicans C. glabrata C. tropicalis C. parapsilosis C. krusei Other Candida Total

120 (51) 57 (24) 16 (7) 12 (5) 27 (12) 2 (1) 63 (27) 50 (21) 25 (11) 5 (2) 9 (4) 2 (1) 6 (3) 3 (1) 8 (3) 5 (2) 21 (9) 4 (2) 9 (4) 8 (3) 94 (40) 94 (40) 62 (26) 15 (6) 5 (2) 4 (2) 2 (1) 6 (3) 235

318 (81)a 121 (31) 15 (4) 37 (10)b 65 (17) 8 (2) 197 (51)a 166 (43)a 105 (27)a 14 (4) 12 (3) 3 (1) 25 (6)b 7 (2) 17 (4) 14 (4) 71 (18)a 11 (3) 36 (9)b 24 (6) — — — — — — — — 389

75 (51) 35 (24) 13 (9) 3 (2) 19 (13) — 40 (27) 31 (21) 15 (10) 3 (2) 5 (3) 1 (1) 4 (3) 3 (2) 6 (4) 3 (2) 11 (13) 2 (1) 5 (3) 4 (3) 61 (41) 61 (41) 39 (27) 10 (7) 4 (3) 4 (3) 2 (1) 2 (1) 147

200 (81) 75 (30) 11 (4) 20 (8) 44 (18) 4 (2) 125 (51) 105 (43) 64 (26) 10 (4) 9 (4) 3 (1) 16 (7) 3 (1) 13 (5) 7 (3) 46 (19) 9 (4) 25 (10) 13 (5) — — — — — — — — 246

a

p ⬍ .01 vs. cases; bp ⬍ .05.

duration of mechanical ventilation (10 ⫾ 14 days in both groups), relaparotomy (6 [18%] and 10 [16%]), and digestive (1 [2%] control) and extradigestive complications (1 [3%] case) in cases and controls, respectively. Cause of death was assessed by the investigators as related to initial infection in three (50%) cases and 12 (80%) controls. Death was observed in one of the five cases (20%) and one of the three controls (33%) receiving empirical antifungal treatment. Six additional patients received an antifungal treatment after identification, whereas 22 patients remained untreated. Mortality in treated and untreated patients was similar (two deaths in treated patients [18%] and three in untreated patients [13%]). In patients with NP, the mean age (66 ⫾ 12 and 67 ⫾ 13 yrs, respectively) and SAPS II (48 ⫾ 12 in both groups) were similar in cases and controls. No differences for other criteria were demonstrated between cases and controls: time to death (23 ⫾ 23 vs. 29 ⫾ 46 days in controls), digestive (8 [14%] vs. 7 [7%]) and extradigestive complications (4 [7%] vs. 2) (2). Crit Care Med 2006 Vol. 34, No. 3

In these patients with NP, inadequacy of empirical antibiotic treatment was similar in the two groups (40% and 36% in cases and controls, respectively). Empirical antifungal treatment was given in 33 (57%) cases and 11 (10%) controls. After identification, antifungal treatment was administered in 46 (79%) cases. Death occurred in 17 of the 33 (52%) patients who received empirical antifungal treatment, whereas 11 of the 25 (44%) patients initially untreated patients died. Similarly, 23 of the 46 (50%) patients who received antifungal treatment after identification of Candida species died vs. five of the 12 (42%) untreated cases. Univariate and multivariate analysis of outcome in NP patients is shown in Table 4. Upper gastrointestinal site and case group were significantly associated with mortality after multivariate analysis.

DISCUSSION To the best of our knowledge, this study is the first to demonstrate that the prognosis of Candida peritonitis de-

pends on the type of infection. In the present study, the significantly increased mortality and morbidity observed in the group of NP with Candida isolation in the peritoneal fluid argue in favor of its pathogenic role. We also consider that the discrepancy observed between CAP and NP patients, despite a similar initial severity, supports the demonstration of the importance of Candida in NP. However, because of the small number of cases of CAP included in this study, the nonsignificant result obtained for outcome does not allow any firm conclusions on the pathogenic role of Candida in this subgroup. The feasibility of a controlled randomized trial designed to demonstrate the pathogenicity of Candida in the course of peritonitis is highly questionable and such a study has never been published. Some of the reasons include a) the marked variability in the incidence of fungal peritonitis; b) enrollment of a sufficient number of patients requiring a large number of study centers; c) the need to stratify patients according to their severity and the type 649

Table 3. Outcome of the study population and patients with nosocomial peritonitis according to initial diagnosis and expressed as number (%) of patients or mean ⫾ SD Study Population

Subsequent reoperation Duration of mechanical ventilation, days Length of stay in ICU, days Death Interval between index operation and death, days Cause of death Death related to peritoneal infection Other infectious cause Noninfectious cause

Nosocomial Peritonitis

Cases

Controls

Cases

Controls

32 (35) 16 ⫾ 17 23 ⫾ 24 34 (37) 22 ⫾ 22

41 (24) 11 ⫾ 14b 16 ⫾ 16b 45 (26) 21 ⫾ 34

26 (45) 18 ⫾ 17 26 ⫾ 25 28 (48) 23 ⫾ 23

31 (29)a 13 ⫾ 16b 18 ⫾ 18b 30 (28)a 29 ⫾ 46

26 (29) 3 (3) 2 (2)

30 (18)a 7 (4) 7 (4)

23 (82) 3 (1) 2 (7)

18 (60)b 7 (23) 5 (17)

ICU, intensive care unit. a p ⬍ .05; bp ⬍ .01 vs. cases. Table 4. Univariate and multivariate analysis with regard to deaths of patients with nosocomial peritonitis (n ⫽ 164) Univariate Analysis

Multivariate Analysis

Risk Factors

Odds Ratio (95% CI)

p Value

Adjusted Odds Ratio (95% CI)

p Value

Case group Upper gastrointestinal tract site Empirical antifungal treatment Inappropriate empirical antibiotic treatment

2.4 (1.2–4.6) 2.1 (1.1–4.1) 1.9 (0.9–3.9) 2.2 (1.1–4.3)

.01 .02 .07 .02

3.0 (1.3–6.7) 4.9 (1.6–14.8) — 1.6 (0.6–4.3)

.009 .005 — .3

CI, confidence interval.

of infection; and d) the long time required to achieve the end point. We therefore decided that a retrospective, multiple-center case-control study, including multivariate analysis, would be an appropriate study design. By selecting controls matched for cases on potentially confounding variables, the question can be addressed with a smaller sample than without matching (19). Although matching eliminates confounding factors for the matched variables, other potential confounding factors are usually also present for which the cases and controls are not matched, justifying multivariate analysis. Finally, the number of patients included gives a representative population, avoiding the influence of a predominant center, which could have easily modified the results. Similarly to community-acquired pneumonia and nosocomial pneumonia, which are clearly different in terms of type of infection and prognosis, our hypothesis was that CAP and NP do not constitute the same disease and consequently that these patients do not behave in the same way and do not share the same outcome, which is why an interaction analysis between groups 650

(CAP and NP) was planned in multivariate analysis using maximum likelihood estimates. As mentioned in the definitions, we only considered as nosocomial intraabdominal infections patients with an infection not present on admission that became evident ⱖ48 hrs after admission. Thus, we assume that the number of misdiagnosed infections was limited. Due to the matching process, such a population should be balanced in cases and control patients. Finally, the number of these patients might have been very limited as suggested by the mean hospital stay before surgery in case patients. At the time of the index operation, physicians were unable to clinically predict the presence or absence of Candida in peritoneal samples, and empirical antifungal treatment was only administered in one half of cases and in several controls. Candida colonization in extraperitoneal sites and previous antibiotic therapy, variables significantly increased in cases, could be proposed to differentiate patients at risk of fungal peritoneal culture, as previously suggested by the experts (1, 12, 13, 20). However, the notion of colonization is not always available at the time of operation. Identification of

Candida on direct examination of peritoneal fluid could also be proposed (7) but was not taken into account in this study as it was not systematically performed in all centers at the time of the study. One of the more perplexing problems in the management of disease in patients with positive Candida samples is whether they require antifungal therapy. Candida is usually regarded as a pathogen when isolated from a patient with an intraabdominal abscess or NP (2, 4). These recommendations are based on reports of a dramatic increase in death rates in the absence of antifungal treatment (3, 4). On the other hand, presence of Candida species in biological samples is frequently discussed in terms of infection (i.e., responsible of the infection) or colonization (no direct link with the current infection). Interestingly, the definition for colonization or infection in the course of peritonitis is somewhat unclear. In the literature and in the Infectious Disease Society of American’s guidelines (1, 2), the experts omitted a clear definition for colonization, outlining the difficulty of this issue. In the present study, curative antifungal treatment was not prescribed in all patients after identification of CanCrit Care Med 2006 Vol. 34, No. 3

T

his study shows that isolation of Candida in perito-

neal specimens of nosocomial peritonitis appears to be an independent risk factor for mortality.

dida species, even in the case of NP. Interestingly, the mortality rate observed in patients receiving antifungal therapy was similar to that observed in untreated patients, questioning the real efficacy of antifungal treatment. This point has been previously reported by other authors, who observed high death rates due to complications of infection despite antifungal treatment (3, 4). In a cohort of Candida peritonitis due to peptic ulcer perforation, Lee et al. (21) administered antifungal therapy to only 35% of patients, following treatment guidelines, and 62.5% of these treated patients died from septic shock. Although many studies have assessed the efficacy of antifungal agents in the treatment of candidemia or deep fungal infections, no convincing clinical trial has ever demonstrated a positive impact of antifungal treatment on morbidity and mortality in fungal peritonitis. It can therefore be speculated that positive fungal culture in the course of peritonitis could be a surrogate marker of increased severity and morbidity, without being the cause of these complications. Most of the information available on clinical features or prognosis of Candida in the course of upper gastrointestinal tract peritonitis is based on reports involving a limited number of observations usually in CAP (4, 6, 10, 21, 22). In NP, upper gastrointestinal perforation was not, until recently, considered to be a risk factor for Candida infection. Eggimann et al. (23), evaluating fluconazole prophylaxis for the prevention of intra-abdominal Candida infections in high-risk surgical patients, included only two patients (10% of the study population) with upper gastrointestinal tract perforation. Dupont et al. (7) recently reported a high proportion of NP originating from the upCrit Care Med 2006 Vol. 34, No. 3

per gastrointestinal tract. These authors also reported that an upper gastrointestinal site was a risk factor of mortality (7). The upper gastrointestinal tract site of peritonitis was also independently associated with fungal isolation in a recent study validating a predictive score of yeast isolation in severe intra-abdominal infections (24). Our results confirm this observation, suggesting that this anatomical site could be a risk factor for Candida peritonitis. The inadequacy of empirical antibiotic treatment is a demonstrated factor of higher mortality rate in peritonitis (5, 25, 26). This inadequacy does not appear to play a pivotal role in our results, as the inadequacy of empirical antibiotic treatment was similar in the two groups. Because of the importance of this variable, it was forcibly included in the multivariate analysis without demonstrating any significant effect on mortality.

3.

4.

5.

6.

7.

8.

CONCLUSIONS This study shows that isolation of Candida in peritoneal specimens of NP appears to be an independent risk factor for mortality. This result, not modified by adjustment for adequacy of empirical antibiotic treatment, is a strong argument in favor of a pathogenic role of Candida in NP. This is reinforced by the absence of increased mortality in the CAP group. However, the need for curative antifungal treatment remains a matter of debate that is not resolved by our results.

9.

10.

11.

12.

13.

ACKNOWLEDGMENTS Study investigators (all cities are in France): G. Plantefeve (Argenteuil), M. Ehrhard (Paris), M. Fleyfel (Lille), J. L. Gerard (Caen), Y. M. Guillou (Rennes), R. Jospe (Saint Etienne), D. Jusserand (Rouen), F. Lagarigue (Chambray les Tours), M. J. Laisné (Paris), T. Lazard (Paris), J. Y. Lefrant (Nimes), J. L. Le Guillou (Paris), T. Lherm (Corbeil), H. Dupont and E. Lobjoie (Amiens), P. Mahul (Saint Etienne), I. Mohammedi (Lyon), P. F. Perrigault (Montpellier), J. C. Rigal (Nantes).

14.

15.

16. 17.

REFERENCES 1. Pappas PG, Rex JH, Sobel JD, et al: Guidelines for treatment of candidiasis. Clin Infect Dis 2004; 38:161–189 2. Solomkin JS, Mazuski JE, Baron EJ, et al: Guidelines for the selection of anti-infective

18.

agents for complicated intra-abdominal infections. Clin Infect Dis 2003; 37:997–1005 Solomkin JS, Flohr AB, Quie PG, et al: The role of Candida in intraperitoneal infections. Surgery 1980; 88:524 –530 Calandra T, Bille J, Schneider R, et al: Clinical significance of Candida isolated from peritoneum in surgical patients. Lancet 1989; ii:1437–1440 Montravers P, Gauzit R, Muller C, et al: Emergence of antibiotic-resistant bacteria in cases of peritonitis after intraabdominal surgery affects the efficacy of empirical antimicrobial therapy. Clin Infect Dis 1996; 23: 486 – 494 Sandven P, Qvist H, Skovlund E, et al: Significance of Candida recovered from intraoperative specimens in patients with intraabdominal perforations. Crit Care Med 2002; 30:541–547 Dupont H, Paugam-Burtz C, Muller-Serieys C, et al: Predictive factors of mortality due to polymicrobial peritonitis with Candida isolation in peritoneal fluid in critically ill patients. Arch Surg 2002; 137:1341–1346 Pacelli F, Doglietto GB, Alfieri S, et al: Prognosis in intra-abdominal infections. Multivariate analysis on 604 patients. Arch Surg 1996; 131:641– 645 Bohnen JM, Mustard RA, Oxholm SE, et al: APACHE II score and abdominal sepsis. Arch Surg 1988; 123:225–229 Alden SM, Frank E, Flancbaum L: Abdominal candidiasis in surgical patients. Am Surgeon 1989; 55:45– 49 Soutter DI, Todd TR: Systemic candidiasis in a surgical intensive care unit. Can J Surg 1986; 29:197–199 Edwards JE Jr, Bodey GP, Bowden RA, et al: International Conference for the Development of a Consensus on the Management and Prevention of Severe Candidal Infections. Clin Infect Dis 1997; 25:43–59 Buchner T, Fegeler W, Bernhardt H, et al: Treatment of severe Candida infections in high-risk patients in Germany: Consensus formed by a panel of interdisciplinary investigators. Eur J Clin Microbiol Infect Dis 2002; 21:337–352 Le Gall JR, Lemeshow S, Saulnier F: A new simplified acute physiology score (SAPSII) based on a European-North American multicenter study. JAMA 1993; 270:29057–29068 Garner JS, Jarvis WR, Emori TG, et al: CDC definitions for nosocomial infections. In: APIC Infection Control and Applied Epidemiology: Principles and Practice. Olmsted RN (Ed). St. Louis, MO, Mosby, 1996, pp A1–A20 Wittmann DH: Intraabdominal infection. World J Surg 1990; 14:145–147 Fagon JY, Chastre J, Novara A, et al: Characterization of intensive care unit patients using a model based on the presence or absence of organ dysfunctions and/or infection: The ODIN model. Intensive Care Med 1993; 19: 137–144 Soussy CJ, Carret G, Cavallo JD, et al: Antibiogram Committee of the French Microbi-

651

ology Society. Report 2000 –2001. Pathol Biol (Paris) 2000; 48:832– 871 19. Katz MH: Multivariate Analysis. Cambridge, UK, Cambridge University Press, 1999 20. British Society for Antimicrobial Chemotherapy Working Party: Management of deep Candida infection in surgical and intensive care unit patients. Intensive Care Med 1994; 20:522–528 21. Lee SC, Fung CP, Chen HY, et al: Candida peritonitis due to peptic ulcer perforation:

Incidence rate, risk factors, prognosis and susceptibility to fluconazole and amphotericin B. Diagn Microbiol Infect Dis 2002; 44: 23–27 22. Peoples JB: Candida and perforated peptic ulcers. Surgery 1986; 100:758 –764 23. Eggimann P, Francioli P, Bille J, et al: Fluconazole prophylaxis prevents intra-abdominal candidiasis in high-risk surgical patients. Crit Care Med 1999; 27:1066 –1072 24. Dupont H, Bourichon A, Paugam-Burtz C, et al:

Can yeast isolation in peritoneal fluid be predicted in intensive care unit patients with peritonitis? Crit Care Med 2003; 31:752–757 25. Mosdell DM, Morris DM, Voltura A, et al: Antibiotic treatment for surgical peritonitis. Ann Surg 1991; 214:543–549 26. Falagas ME, Barefoot L, Griffith J, et al: Risk factors leading to clinical failure in the treatment of intra-abdominal or skin/soft tissue infections. Eur J Clin Microbiol Infect Dis 1996; 15:913–921

ACCM Guidelines on SCCM Website The Guidelines and Practice Parameters developed by the American College of Critical Care Medicine are now available online at http://www.sccm.org/professional_resources/guidelines/ index.asp. The printed version of the Guidelines, provided in a binder, is also available through the SCCM Bookstore, located at http://www.sccm.org/pubs/sccmbookstore.html. Please watch the Website to stay updated on the ACCM Guidelines and Practice Parameters.

652

Crit Care Med 2006 Vol. 34, No. 3