Journal of Critical Care (2008) 23, 11–17

The clinical significance of Candida colonization ofrespiratory tract secretions in critically ill patientsB

Marie-Soleil Delislea, David R. Williamsonb,c, Marc M. Perreaultc,d,Martin Albertb,c, Xuran Jiange, Daren K. Heylande,⁎

aHôpital de l'Enfant-Jésus, Québec, Québec, CanadabHôpital du Sacré-Cœur de Montréal, Montréal, Québec, CanadacUniversité de Montréal, Montréal, Québec, CanadadCentre Universitaire de Santé McGill, Montréal, Québec, CanadaeQueens University, Kingston, Ontario, Canada

g

0d

Keywords:Candida;colonization;Candida airwaycolonization;

Pneumonia;Ventilator-associatedpneumonia;

Critical care

AbstractPurpose: Clinical uncertainty exists regarding the significance of colonization confined to respiratorytract secretions with Candida sp in critically ill patients. Our objectives were to describe suchcolonization, its associated risk factors, and to examine the clinical outcomes in patients with aclinical suspicion of ventilator-associated pneumonia with isolated Candida colonization compared tothose without.Materials and Methods: In a retrospective analysis of the Canadian ventilator-associated pneumoniastudy, patients were divided into 2 groups according to the isolated presence or absence of Candidain the respiratory tract enrollment culture. We compared length of mechanical ventilation, intensivecare unit and hospital stay, and mortality outcomes between groups. We used multiple logisticregression analysis to determine factors independently associated with Candida colonization andhospital mortality.Results: Of the 639 eligible patients, 114 (17.8%) were colonized with Candida in the enrollmentculture. A multivariate analysis identified female sex (odds ratio [OR], 1.65; 95% confidence interval[CI], 1.02-2.65), number of comorbidities (OR, 1.35; 95% CI, 1.08-1.71), worsening or persistentinfiltrate at randomization (OR, 1.92; 95% CI, 1.09-1.38), antibiotics started within 3 days ofrandomization (OR, 3.16; 95% CI, 1.71-5.83), and on antibiotics at randomization but all startedmore than 3 days before randomization (OR, 3.04; 95% CI, 1.68-5.50) as variables associated withCandida respiratory tract colonization. A significant increase in median hospital stay (59.9 vs 38.6 days,P = .006) and hospital mortality (34.2% vs 21.0%, P = .003) was observed in patients with Candidacolonization. In a multivariate model, Candida colonization of the respiratory tract was independentlyassociated with hospital mortality (OR, 2.47; 95% CI, 1.39-4.37).Conclusion: Respiratory tract Candida colonization is associated with worse clinical outcomes and isindependently associated with increased hospital mortality. However, it is unclear whether Candida

☆ This study was supported by grants from the Canadian Institutes of Health Research and Physicians' Services Incorporated of Ontario, and by unrestrictedrants from AstraZeneca, Bayer, and Merck Frosst.⁎ Corresponding author. Tel.: +613 549 6666x3339, fax: +1 613 548 1351.E-mail address: dkh2@queensu.ca (D.K. Heyland).

883-9441/$ – see front matter © 2008 Elsevier Inc. All rights reserved.oi:10.1016/j.jcrc.2008.01.005

12 M.-S. Delisle et al.

colonization is causally related to poor outcomes or whether it is a marker for increased morbidityand mortality.© 2008 Elsevier Inc. All rights reserved.

1. Introduction using a 2 × 2 factorial design to either bronchoscopy or

Fungal infection and colonization are becoming majorclinical challenges in patients in the intensive care unit (ICU)[1-12]. Candida colonization from various sites has beenassociated with prolonged ICU and hospital stays as well asincreased health care costs in mixed medical, surgical, andtrauma populations [10]. Although some studies addressCandida colonization mainly in surgical ICU patients, fewexamine other ICU populations. Wey et al [13] identifiedCandida colonization at one or multiple sites as anindependent risk factor for candidemia. Although numerousrisk factors for candidemia have been reported, none of theserisk factors have been studied for association with anincreased risk of Candida colonization at any site [13-16].

Isolation of Candida sp from respiratory tract secretions isfrequent in mechanically ventilated, non-immunocompro-mised patients [9,11,17-20]. Studies assessing the signifi-cance of colonization of respiratory tract secretionsspecifically are scarce and leave the matter unclear[9,11,17-20]. The clinical relevance of isolated respiratorytract secretion Candida colonization in contrast to coloniza-tion of multiple sites has also not been established. Recently,in amixedmedical and surgical population, Azoulay et al [21]found an association between Candida colonization of therespiratory tract secretions and a prolonged period ofmechanical ventilation (MV), longer ICU and hospital stay,and an increased risk for PseudomonasVAP but no differencein mortality. Of note, 39.7% of patients in that study hadextrapulmonary Candida colonization. No studies havedescribed the risk factors and outcomes of patients withisolated Candida colonization of respiratory tract secretionsin a North American setting.

The primary objective of this study of patients with aclinical suspicion of a VAP was to describe the characteristicsof patients with isolated Candida colonization (confined tothe respiratory tract secretions only) and to determine the riskfactors associated with Candida colonization. Our secondaryobjectives were to compare patients with and without Can-dida colonization in terms of duration of MV, ICU andhospital stay; and ICU, hospital, and 28-day mortality.

2. Materials and methods

2.1. Design

We conducted a retrospective cohort study on a subset ofpatients from the Canadian randomized, multicenter VAPtrial [22]. In this trial, ICU patients with a clinical suspicionof a VAP from 28 participating centers were randomized

endotracheal aspirate (ETA) groups and to receive treatmentwith either 2 broad-spectrum antibiotics or a single broad-spectrum antibiotic. Subjects were followed prospectivelyfor 28 days or until death or hospital discharge.

2.2. Population

Adult patients were included in the VAP trial if they wereadmitted to an ICU for more than 96 hours, if they weremechanically ventilated for more than 48 hours, and if theyfulfilled criteria for a clinical suspicion of VAP. The criteriafor clinical suspicion of pneumonia had to be met within48 hours of enrollment [22]. Immunocompromised patients,defined as post–organ transplantation, HIV or neutropenic(b1000 absolute neutrophils) patients, or those receivingcorticosteroids more than 20 mg/d of prednisone orequivalent for more than 6 months, were excluded, as wellas patients infected or colonized with Pseudomonas speciesor methicillin-resistant Staphylococcus aureus, and thoseunlikely to be discharged from ICU within 3 weeks. Thecomplete VAP study design and results have been publishedelsewhere [22].

Within12hours of randomization, respiratory tract cultures(either bronchoalveolar lavage [BAL] or ETA)were obtained.For this analysis, we divided patients in 2 groups according towhether Candida was present or not in the enrollmentspecimen. Because we focused on Candida colonizationconfined to respiratory tract secretions, patientswith a positiveCandida culture at any site in the 7-day period precedingrandomizationand fromany siteother than the respiratory tracton day of randomization were excluded from this analysis.Cultures other than respiratory samples obtained at randomi-zation were performed according to clinicians' judgment.

Sample procurement and laboratory processing werestandardized. The BAL and ETA were performed in astandardized manner according to conventional techniques.Both ETA and BAL specimens were transported immedi-ately to the laboratory.

2.3. Data collection

Upon enrollment in the VAP trial, demographics such asage, sex, height, weight, primary admission diagnosiscategory, number of comorbid diseases, APACHE (Acutephysiology and chronic health evaluation) II score, type ofICU, date and time of randomization, admission to thehospital, admission to the ICU, and start of MV wererecorded for all patients [23]. Daily monitoring was thenperformed on all patients for signs and symptoms of

13Significance of Candida colonization of respiratory tract secretions

infection, organ dysfunction, and other ICU-acquiredcomplications [24]. During hospitalization, newly acquireddiagnoses, duration of MV, need for vasopressors, andhemodialysis were recorded. Culture results from all sitesand antibiotic use for the 7 days before enrollment andthroughout the study period were recorded. We recordedsecondary outcome measures of days of MV, in the ICU andin the hospital, as well as occurrence of fungemia and ICU,in-hospital, and 28-day mortality.

2.4. Statistical analysis

Baseline patient characteristics were compared betweenpatients with and without the presence of Candida in therespiratory tract enrollment culture. Categorical variables aredescribed as counts and percentages and were tested using theχ2 test. Continuous variables are described as means withSDs and compared using independent t test. Mortalityoutcomes between the colonized patients vs noncolonizedpatients were compared using Fisher exact test. Length of

Fig. 1 Patient

MV, length of stay, and mortality are described using Kaplan-Meier estimates of the median and quartiles and werecompared between groups using the log-rank test. To identifyfactors independently associated with Candida colonizationand increased mortality in colonized patients vs noncolonizedpatients, we used logistic regression including patientcharacteristics with a univariate model significance level of0.20: age, sex, APACHE II score, number of comorbidities,presence vs absence of Candida in respiratory tract secre-tions, admission diagnosis category (sepsis, trauma, andneurologic condition assessed independently), BAL vs ETAdiagnostic method, antibiotic monotherapy vs combinationtherapy, use of vasopressors during hospitalization, anti-biotics started within 3 days before randomization vs noantibiotic, antibiotics started prior to 3 days before randomi-zation vs no antibiotics, multiple organ dysfuncfion score onday 1, and PaO2/FiO2 ratio also on day 1. Variables identifiedfrom the univariate model with a significance level of 0.20 orless were included in a multivariate analysis, in addition toage, sex, APACHE II score, bronchoscopy vs ETA groups,and combination antibiotic therapy vs monotherapy to

flowchart.

14 M.-S. Delisle et al.

determine relevant risk factors. All tests are two-sided, and aP value of less than .05 was considered statisticallysignificant. Analyses were completed with SAS Version 9.1(SAS Institute, Cary, NC).

3. Results

Of the 740 patients enrolled in the VAP study betweenMay 2000 and February 2005, we included 639 for analysis.One patient withdrew consent and was not consideredfurther. We excluded an additional 100 patients: 79 hadCandida-negative respiratory tract secretions sample onrandomization day, but positive respiratory tract secretion

Table 1 Baseline patient characteristics

Colonized

n = 114

Age 59.6 ± 16.8Male sex (n [%]) 71 (62.3)Admission category (n [%])Medical 72 (63.2)Surgical 42 (36.8)Primary diagnosis system (n [%])Cardiovascular 33 (28.9)Gastrointestinal 9 (7.9)Neurologic 7 (6.1)Renal 1 (0.9)Respiratory 25 (21.9)Sepsis 8 (7.0)Trauma 24 (21.1)Other condition 7 (6.1)No. of comorbidities0 24 (21.1)1 28 (24.6)2 28 (24.6)3 34 (29.8)APACHE II 20.4 ± 6.4PaO2/Fio2 201.1 ± 77.0Day 1 MODS 6.0 ± 2.7Days on MV 7.7 ± 4.5Days in ICU 8.0 ± 4.3CXR at enrollment (n [%])New infiltrate 20 (17.5)Worsening orpersistentinfiltrate

94 (82.5)

On antibiotics at randomization (n [%])Not on an any within 3 d 21 (18.4)Yes but none startedwithin 3 d

53 (46.5)

New antibiotics startedwithin 3 d

40 (35.1)

On antifungal agents 8 (7.0)On vasopressors 32 (28.1)

MODS indicates multiple organ dysfunction score.

sample within 7 days before randomization, or had Candida-positive culture from other sites within that same period or onday of randomization; 21 were excluded because of Candidafound in their respiratory tract secretions on the day ofrandomization and a positive Candida culture from othersites within 7 days before or on the day of randomization.

We found Candida colonization confined to respiratorytract secretions on the day of randomization in 114 (17.8%)of 639 patients (Fig. 1). The proportion of patients whosepositive Candida cultures were isolated via the ETA or BALarms of the VAP trial were 15.6% (n = 49) and 20.0%(n = 65), respectively (P = .15).

Various types of Candida sp were isolated from res-piratory tract secretion specimens collected at randomization.

Noncolonized Total P

n = 525 N = 639

58.9 ± 18.4 59.0 ± 18.1 .71377 (71.8) 448 (70.1) .04

.69321 (61.1) 393 (61.5)204 (38.9) 246 (38.5)

.007128 (24.2) 161 (25.2) .3144 (8.4) 53 (8.3) .8780 (15.2) 87 (13.6) .013 (0.6) 4 (0.6) .6875 (14.3) 100 (15.6) .7111 (2.1) 19 (3.0) .04146 (27.8) 170 (26.6) .00538 (7.2) 45 (7.0) .14

.04174 (33.1) 198 (31.0)136 (25.9) 164 (25.7)95 (18.1) 123 (19.2)120 (22.9) 154 (24.1)19.8 ± 6.2 19.9 ± 6.3 .37221.4 ± 84.5 217.6 ± 83.5 .025.5 ± 3.0 5.6 ± 2.9 .117.5 ± 5.6 7.5 ± 5.4 .827.6 ± 5.6 7.6 ± 5.3 .37

.001172 (32.8) 192 (30.0)353 (67.2) 447 (70.0)

b.001227 (43.2) 248 (38.8)155 (29.5) 208 (32.6)

143 (27.2) 183 (28.6)

14 (2.7) 22 (3.4) .02112 (21.3) 144 (22.5) .12

Table 2 Mortality outcomes of patients with and withoutCandida colonization of respiratory tract secretions

Colonized Noncolonized RR (95% CI) P

n = 114 n = 525 (Colonized/noncolonized)

28-dmortality

27 (23.7) 86 (16.4) 1.45 (0.99-2.12) .08

ICUmortality

24 (21.1) 73 (13.9) 1.51 (1.00-2.29) .06

Hospitalmortality

39 (34.2) 110 (21.0) 1.63 (1.20-2.21) .003

RR indicates relative risk.

15Significance of Candida colonization of respiratory tract secretions

Candida albicans was accounted for in 65.3% ofsamples. Torulopsis glabrata was isolated in 1.3% ofsamples, whereas other non-albicans sp were found in6.7% of airway specimen. One specimen (0.7%) was iden-tified to have non–Candida albicans yeast species, andfinally, 26% of samples grew yeast not further speciatedby the local laboratory. All non–albicans sp were isolatedin colonized patients who were on antifungals at timeof randomization.

Patients with and without isolated Candida colonizationof respiratory tract secretions at randomization arecompared in Table 1. Male patients were more frequentlyfound in the noncolonized group (71.8% vs 62.3%, P =.04). Patients in the colonized group were more frequentlyadmitted for sepsis (7.0% vs 2.1%, P = .005) andrespiratory conditions (21.9% vs 14.3%, P = .04) butless often for neurologic conditions (6.1% vs 15.2%, P =.01). A significantly lower PaO2/FiO2 ratio was observed inthe colonized group (201.1 vs 221.4, P = .02). Morepatients in the colonized group had worsening or persistentinfiltrate as opposed to a new infiltrate at the time thatVAP was suspected. They were also more frequently onantibiotics (81.6% vs 56.7%, P b .001) and antifungalagents (7.0% vs 2.7%, P = .02) at randomization. Thosewith colonization had greater number of comorbidconditions (P = .04). Bacterial VAP was documentedthrough BAL or ETA culture results in 72 (63.2%)colonized patients compared to 417 (79.4%) patients inthe noncolonized group (P b .001).

A multivariate analysis identified the following variablesas carrying a significantly higher risk of respiratory tract

Table 3 Clinical outcomes of patients with and without Candida col

Colonized

n (%) Median (IQR)

Time on ventilator (d) 91 (79.8) 10.9 (3.9, 32.7ICU length of stay (d) 88 (77.2) 14.1 (6.2, 43.2Hospital length of stay (d) 73 (64.0) 59.9 (24.2, –)

IQR indicates interquartile range.

secretion Candida colonization: female sex (odds ratio[OR], 1.65; 95% confidence interval [CI], 1.02-2.65; P =.04), number of comorbidities (OR, 1.35; 95% CI, 1.08-1.71; P = .01), worsening or persistent infiltrate atrandomization (OR, 1.92; 95% CI, 1.09-1.38; P = .02),antibiotics started within 3 days of randomization (OR,3.16; 95% CI, 1.71-5.83; P = .02), and on antibiotics atrandomization but all started more than 3 days beforerandomization (OR, 3.04; 95% CI, 1.68-5.50; P = .02).Candida colonization of respiratory tract secretions wasless common in patients admitted for neurologic reasons(OR, 0.35; 95% CI, 0.14-0.87; P = .02).

We present mortality data in Table 2, whereas results forother clinical outcomes are outlined in Table 3. Asignificant increase in hospital mortality (34.2% vs 21.0%P = .003) was observed in the colonized group. Hospitalstay was also significantly longer (59.9 vs 38.6 days, P =.006) in this group. All other outcomes tended to be worsein the colonized group but were not significantly differentbetween groups. Candidemia subsequently developed in 1and 4 patients from the colonized and noncolonized groups,respectively. Of note, 34 patients (29.8%) in the colonizedgroup and 72 (13.7%) in the noncolonized group receivedantifungals at some point during the course of the VAP trial(P b .001). Fluconazole was the most commonly usedagent. Mortality in colonized patients who receivedantifungals was not significantly different from patientswho did not receive such treatment (38.2% vs 32.5%, P =.67). In addition, no significant difference in mortality wasfound in patients excluded from the study because ofmultiple sites Candida colonization compared with patientswith respiratory tract Candida colonization only (41.9% vs34.2%, P = .33).

Logistic regression showed a significant associationbetween Candida colonization of respiratory tract secretionsand hospital mortality (OR, 2.38; 95% CI, 1.38-4.11). Otherfactors significantly independently associated with increasedhospital mortality were older age (OR, 1.06; 95% CI, 1.04-1.08), increased number of comorbidities (OR, 1.38; 95% CI,1.11 -1.70), and admission for neurologic disorders (OR,2.48; 95% CI, 1.28-4.78).

Although patients with pseudomonal colonization orinfection at randomization were excluded from the VAPtrial, 1.8% (n = 2) of patients in the colonized group and1.3% (n = 7) of patients in the noncolonized group developedsubsequent pseudomonal superinfection (P = .67).

onization of respiratory tract secretions

Noncolonized P

n (%) Median (IQR)

) 447 (85.1) 8.1 (3.8, 18.1) .06] 439 (83.6) 11.6 (6.1, 25.2) .07

407 (77.5) 38.6 (21.0, 111.0) .006

16 M.-S. Delisle et al.

4. Discussion

We conducted a retrospective analysis of data from a largerandomized trial of diagnostic strategies and empiricantibiotics in patients with a clinical suspicion of VAP toassess whether isolated Candida colonization of respiratorytract secretions impacts on patient outcomes. Our mainfinding is that isolated Candida colonization of respiratorytract secretions is associated with a longer hospital length ofstay and increased hospital mortality.

Previous studies have suggested that Candida coloniza-tion without Candida infection or candidemia is associatedwith worse clinical outcomes in nonsurgical populations.Olaechea et al demonstrated that critically ill patients withCandida colonization, defined as the presence of Candidasp in nonsignificant samples obtained from the urine,stomach, oropharynx, or tracheal aspirates, experienced aprolonged ICU stay by 6.2 days (OR, 1.69; 95% CI, 1.53-1.87,P b .001) and an extended hospital stay by 8.6 days (OR,1.27; 95% CI, 1.16-1.40, P b .001), resulting in increasedhealth care costs [10]. This interesting finding left unan-swered the question of whether Candida colonization at aspecific site was significantly associated with worse clinicaloutcomes. Recently, Azoulay et al [21] have observed thatpatients with Candida airway colonization, defined asrecovery of Candida from the respiratory tract, had asignificantly longer period of MV (13 vs 6 days, P b .0001),ICU stay (17 vs 9 days, P b .0001), and hospital stay (36 vs 22days, P b .0001) as compared to patients not colonized withCandida from the respiratory tract. However, they found nodifference in ICU or hospital mortality. Our results show atrend for longer duration of MV and length of ICU stay, aswell as for increased ICU and 28-day mortality. Though notsignificant, these results are in keeping with those obtained byAzoulay et al. Furthermore, 39.7% of patients with Candidaairway colonization in that study had extrapulmonarycolonization, raising the issue of different findings dependingon patient populations. To our knowledge, no previous studyhas demonstrated a relationship between respiratory tractCandida colonization and increased mortality.

Admission characteristics other than respiratory tractsecretion colonization by Candida also identify patients withaworseoutcome.Colonizedpatientsweremoreoftenadmittedwith pulmonary conditions, were more often on antibiotics,tended to have more comorbid diseases, have a lower PaO2/FiO2 ratio, andweremoreoften identified ashavingapersistentor worsening infiltrate. These conditions suggest thatcolonized patients were sicker upon admission to ICU, yetAPACHE II score calculated at that time was similar betweengroups. This is similar to findings by Azoulay et al, whichobserved that patients colonized in the respiratory tract havebeen more often admitted for acute respiratory failure, weremore infected at admission, yet were less often admitted forcoma and less often directly admitted to ICU. These commonfindings suggest that patients with Candida colonization ofrespiratory tract secretions are at risk of worse outcome with

possibly numerous contributing factors. However, whenmultivariate logistic regression was performed, only Can-dida colonization of respiratory tract secretions, number ofcomorbidities, and admission for neurologic reasons werefound to be associated with increased hospital mortality.

Amultivariate analysis suggests that patients who developCandida colonization of respiratory tract secretions aregenerally sicker and have recently received antibiotics thanthose who do not. Further prospective trials are required toconfirm which patient characteristics are risk factors forrespiratory tract secretionsCandida colonization andwhethersuch colonization is independently associated with worseclinical outcomes. The question also remains as to whetherisolated Candida colonization of respiratory tract secretionscarries a different burden of illness than multiple-site Can-dida colonization. In our study, mortality between these 2groups of patients did not differ significantly, but this wouldneed to be further explored in future trials, along withcomparison of other outcomes such as length of ICU orhospital stay, length of MV, or incidence of subsequentpseudomonal VAP.

The issue as to whether respiratory tract secretions Can-dida colonization is associated with or responsible for anincreased burden of illness remains unclear. The retro-spective design of our analysis does not provide an answer.However, the fact that few patients in either study groupdeveloped candidemia possibly indicates that it may not be amajor cause of death in this population. The publishedliterature suggests that isolation of Candida from therespiratory tract correlates poorly with candidiasis [18,19].This further raises the question as to whether initiatingantifungal treatment would be beneficial, which can only beaddressed with rigorous randomized trials.

An association between Pseudomonas and the presenceof Candida sp has been suggested previous trials [21,25]. Inour study, a comparable proportion of patients in each groupdeveloped subsequent pseudomonal pneumonia superinfec-tion. The fact that patients with pseudomonal colonizationand infection were excluded from the VAP trial by designand the small number of patients prevents firm conclusionsto be drawn on the matter.

The strengths of our study include the large number ofpatients enrolled, the prospective data collection in amulticenter, randomized controlled setting, and airwayspecimen collection done by protocol. We collected data onboth invasive and noninvasive techniques to obtain micro-biological specimens on mechanically ventilated patients.

There are several limitations to this analysis including theassumption that isolation of Candida from respiratory tractculture specimens indicates colonization and not infection.However, this is in concordance with similar studies [10,21].As a normal colonizer of the gastrointestinal tract and oralcavity, Candida spreads along the respiratory tract afterintubation or aspiration of gastric content [12,21]. It is thenusually observed in small number in respiratory tractquantitative cultures without clinical evidence of pneumonia

17Significance of Candida colonization of respiratory tract secretions

[12,21]. It was not possible in this study to examine the risk ofaspiration pneumonia or other respiratory conditions asso-ciated with Candida colonization of the respiratory tractbecause these were not documented in theVAP trial. Candidalinvasion of the lung parenchyma after hematogenousdissemination would be frequently associated with realclinical pneumonia but unlikely to have happened in thecourse of our trial as only 5 patients developed documentedcandidemia [12,18,19,21]. Clinical suspicion of VAP wasgenerally attributed to bacterial causative agents. We did notroutinely perform surveillance cultures of all other sites to besure that patients were truly only colonized in the respiratorytract. This study was done in the context of a clinicalsuspicion of VAP and does not generalize to patients withoutsuch a suspicion. Moreover, a proportion of patients in eachgroup received antifungal treatment. The decision to initiateor stop treatment was at clinicians' discretion and could bedone at any time during the 28-day study period. Therefore,the impact of antifungal treatment on patient outcomes and onour results could not be adequately measured within thisstudy, or controlled for as a confounding variable.

5. Conclusion

Our study demonstrates an association between isolatedCandida colonization of respiratory tract secretions andincreased hospital mortality and hospital length of stay.Whether Candida colonization of respiratory tract secretionsis a marker of disease severity or contributes directly tomortality and prolonged hospital stay requires furtherevaluation. One way to address this question is to conducta prospective randomized trial of pharmacologic strategies toeliminate or reduce the colonization of Candida and evaluatethe effect of such an intervention on clinically importantoutcomes in critically ill mechanically ventilated patientswith or without a clinical suspicion of pneumonia.

Acknowledgments

We thank the ICU nurses and physicians in participatingcenters, and research coordinators of the Canadian CriticalCare Trials Group listed in the appendix of the originalarticle [22].

References

[1] Banerjee SN, Emori TG, Culver DH, Gaynes RP, Jarvis WR, Horan T,et al. Secular trends in nosocomial primary bloodstream infections inthe United States, 1980-1989. Am J Med 1991;91:86S-9S.

[2] Beck-Sagué CM, Jarvis W. Secular trends in the epidemiology ofnosocomial fungal infections in the United States, 1980-1990. NationalNosocomial Infections Surveillance System. J Infect Dis 1993;167:1247-51.

[3] Vincent JL, Bihari DJ, Suter PM, et al. The prevalence of nosocomialinfection in intensive care units in Europe: results of the EuropeanPrevalence of Infection in Intensive Care (EPIC) Study; EPICInternational Advisory Committee. JAMA 1995;274:639-44.

[4] Rangel-Frausto MS, Wiblin T, Blumberg HM, Saiman L, Patterson J,Rinaldi M, et al. National epidemiology of mycoses survey (NEMIS):variations in rate of bloodstream infections due to Candida sp. inseven surgical intensive care units and six neonatal intensive care units.Clin Infect Dis 1999;29:253-8.

[5] Wey SB, Mori M, Pfaller MA, Woolson RF, Wenzel RP. Hospital-acquired candidemia: the attributable mortality and excess length ofstay. Arch Intern Med 1988;148:2642-5.

[6] Wenzel RP. Nosocomial candidemia: risk factors and attributablemortality. Clin Infect Dis 1995;20:1531-4.

[7] Rentz AM, Halpern MT, Bowden R. The impact of candidemia onlength of hospital stay, outcome and overall cost of illness. Clin InfectDis 1998;27:781-8.

[8] Charles PE, Doise JM, Quenot JP, Aube H, Dalle F, Chavanet P, et al.Candidemia in critically ill patients : difference of outcome betweenmedical and surgical patients. Intern Care Med 2003;29:2162-9.

[9] Pittet D, Monod MM, Suter PM, Frenk E, Auckenthaler R. Candidacolonization and subsequent infections in critically ill surgical patients.Ann Surg 1994;220(6):751-8.

[10] Olaechea PM, Palomar M, Léon-Gil C, Álvarez-Lerma F, Jordá R,Nolla-Salas J, et al. Economic impact of Candida colonization andCandida infection in the critically ill patient. Eur J Clin MicrobiolInfect Dis 2004;23:323-30.

[11] Charles PE, Dalle F, Aube H, Doise JM, Quenot JP, Aho LS. Candidaspp. colonization significance in critically ill medical patients: aprospective study. Intensive Care Med 2005;31:393-400.

[12] Azoulay E, Cohen Y, Zahar JR, Garrouste-Orgeas M, Adrie C, MoineP, et al. Practices in non neutropenic ICU patients with Candida-positive airway specimens. Intensive Care Med 2004;30:1384-9.

[13] Wey SB, Mori M, Pfaller MA, Woolson RF, Wenzel RP. Risk factorsfor hospital-acquired candidemia. A matched case-control study. ArchIntern Med 1989;149:2349-53.

[14] Fraser VJ, Jones M, Dunkel J, Storfer S, Medoff G, Dunagan WC.Candidemia in a tertiary care hospital: epidemiology, risk factors andpredictors of mortality. Clin Infect Dis 1992;15:414-21.

[15] Jarvis WR. Epidemiology of nosocomial fungal infections, withemphasis on Candida species. Clin Infect Dis 1995;20:1526-30.

[16] Bross J, Talbot GH, Maislin G, Hurwitz S, Strom BL. Risk factorsfor nosocomial candidemia: a case-control study in adults withoutleukemia. Am J Med 1989;87:614-20.

[17] Eggiman P, Pittet D. Candidoses du sujet non neutropénique: de lacolonisation à l'infection. Ann Fr Anesth Réanim 2001;20:382-8.

[18] El-Ebiary M, Torres A, Fàbregas N, Puig de la Bellacasa J, González J,Ramirez J, et al. Significance of the isolation of Candida species fromrespiratory samples in critically ill, non neutropenic patients. Am JRespir Crit Care Med 1997;156:583-90.

[19] Rello J, Esandi ME, Díaz E, Mariscal D, Gallego M, Vallès J. The roleof Candida sp isolates from bronchoscopic samples in nonneutropenicpatients. Chest 1998;114:146-9.

[20] Wood C, Mueller EW, Croce MA, et al. Candida sp. isolated frombronchoalveolar lavage: clinical significance in critically ill traumapatients. Intensive Care Med 2006;32(4):599-603.

[21] AzoulayE, Timsit JF, TaffletM, deLassenceA,DarmonM,Zahar JR, etal. Candida colonization of the respiratory tract and subsequent Pseu-domonas ventilator-associated pneumonia. Chest 2006;129:110-7.

[22] Heyland D, Dodek P, Muscedere J, Day A. A randomized trial ofdiagnostic techniques for ventilator-associated pneumonia. N Engl JMed 2006;355:2619-30.

[23] Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: aseverity of disease classification system. Crit Care Med 1985;13:818-29.

[24] Marshall JC, Cook DJ, Christou NV, Bernard GR, Sprung CL,Sibbald WJ. Multiple organ dysfunction score: a reliable descriptorof a complex clinical outcome. Crit Care Med 1995;23:1638-52.

[25] Nseir S, Jozefowicz E, Cavestri B, et al. Impact of antifungal treatmenton Candida-Pseudomonas interaction: a preliminary retrospectivecase-control study. Intensive Care Med 2007;33(1):137-42.