Significance of the Isolation of Candida Species from ... - ATS Journals

Significance of the Isolation of Candida Species from Respiratory Samples in Critically Ill, Non-neutropenic Patients An Immediate Postmortem Histologic Study MUSTAFA EL-EBIARY, ANTONI TORRES, NEUS FÀBREGAS, JORGE PUIG de la BELLACASA, JULIÀ GONZÁLEZ, JOSEP RAMIREZ, DOLORES del BAÑO, CARMEN HERNÁNDEZ, and M. T. JIMÉNEZ de ANTA Serveis de Pneumologia i Al.lèrgia Respiratòria, Anestesia, Microbiologia, Anatomia Patologica, Hospital Clínic, Departament de Medicina, Universitat de Barcelona, Spain

The diagnosis of pulmonary candidiasis is still controversial. We undertook a prospective study on 25 non-neutropenic, mechanically ventilated (. 72 h) patients who died in our ICU with the aim of assessing the incidence and significance of the isolation of Candida species from quantitative cultures of immediate postmortem lung biopsies and different respiratory sampling techniques. Immediate postmortem respiratory samples (endotracheal aspirate, protected specimen brush [PSB], bronchoalveolar lavage [BAL], blind biopsies [average 14/patient], and bilateral bronchoscopically guided biopsies [two per patient]) were taken from all patients. Lung tissue specimens were histologically examined. Respiratory samples were classified as having Candida or otherwise. Ten (40%) patients had at least one pulmonary biopsy yielding Candida spp. Among these 10 patients with Candida isolates, only two had definite pulmonary candidiasis. A total of 470 microorganisms were isolated from 280 of 375 (77%) lung biopsy samples in all 25 patients. Candida species represented 9% (n 5 40) of the isolates, corresponding to 10 patients (40%). In the 10 patients in whom Candida species was isolated from pulmonary biopsies, this was always associated with the isolation of the same microorganism from one of the sampling methods. Quantitative cultures of Candida species from different sampling methods correlated well among each other but could not discriminate the presence from absence of Candida pneumonia. A logistic regression model adjusted for the presence of antibiotics, days of antibiotic treatment, mechanical ventilation period, age, ARDS, parenteral nutrition, and gender did not show any independent risk factor for developing positive pulmonary samples for Candida species. The incidence of Candida isolation from pulmonary biopsies in critically ill mechanically ventilated, non-neutropenic patients who die is high (40%). However, the incidence of definite Candida pneumonia was 8%. We also found that Candida colonization is uniform throughout the different lung regions, and that the presence of Candida in respiratory samples, independently of quantitative cultures, is not a good marker of Candida pneumonia in critically ill, non-neutropenic, non-AIDS patients. El-Ebiary M, Torres A, Fàbregas N, de la Bellacasa JP, González J, Ramirez J, del Baño D, Hernández C, Jiménez de Anta MT. Significance of the isolation of Candida species from respiratory samples in critically ill, non-neutropenic patients: an immediate postmortem hisAM J RESPIR CRIT CARE MED 1997;156:583–590. tologic study.

(Received in original form December 2, 1996 and in revised form February 25, 1997) Supported in part by grants CIRIT/Fundació Clínic (Comissió Interdepartamental per a la Recerca i Tecnologia), and FIS (Fondo de Investigación Sanitaria) 94/ 0583 and I1D 96/0024. Dr. del Baño was a Research Fellow from Residencia Sanitaria Virgen de la Arrixaca, Murcia, Spain. Correspondence and requests for reprints should be addressed to Dr. A. Torres, Servei de Pneumologia, Hospital Clínic, c/Villarroel 170, 08036 Barcelona, Spain. Am J Respir Crit Care Med

Vol. 156. pp. 583–590, 1997

Fungal infections account for nearly 8% of all nosocomial infections; Candida is the responsible agent in 80% of the cases (1). This microorganism once considered as a minor pathogen is now among the most commonly cultured microorganisms in intensive care units (ICU). This fact is in part attributable to the prolonged antibiotic therapy, the more frequent use of surgery, instrumentation, and the greater number of immunocompromised patients being hospitalized, as well as the extensive use of intensive care facilities. Now ICU physicians have to routinely include Candida in their differential diagnoses when dealing with nosocomial infections.

584

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE

The isolation of Candida species from the respiratory secretions is frequent in mechanically ventilated patients (2). This occurs as a result of seeding of the lungs secondary to hematogenous dissemination, or it may follow the aspiration of colonized oropharyngeal or gastric contents (3). Nevertheless, invasive lung infection by Candida species is rare in nonimmunocompromised subjects. The criteria for the diagnosis of pulmonary candidiasis are still controversial. The isolation of Candida from cultures of sputum, endotracheal aspirates, bronchoscopic samples, percutaneous lung needle aspirates, and even lung tissue may only represent colonization of the tracheobronchial tree. In addition, the value of quantitative cultures of respiratory samples to diagnose Candida pneumonia is unknown. Thus, the frequency of this infection has not been well established in mechanically ventilated patients. Despite the debate about the diagnosis of pulmonary candidiasis, the definite diagnosis of pulmonary candidiasis still rests on histologic demonstration of the yeast in lung tissue with associated inflammation (4). We performed a prospective study upon immunocompetent, mechanically ventilated patients who died in our unit with the aim of assessing the incidence and significance of the isolation of Candida species in quantitative cultures of different respiratory sampling techniques (endotracheal aspirates, protected specimen brush, bronchoalveolar lavage), comparing them with the histology and microbiology of immediate postmortem pulmonary biopsies.

VOL. 156

1997

TABLE 1 CLINICAL CHARACTERISTICS OF PATIENTS WITH AND WITHOUT POSITIVE RESPIRATORY SAMPLES FOR CANDIDA SPECIES

Age, yr Sex (M/F) APACHE II* MV† period, days Temperature, 8C Leukocytes, 3109/L PaO2/FIO2 Antibiotic treatment (yes/no) No. of antibiotics Antibiotic treatment period, days Parenteral nutrition (yes/no) TPN‡ period, days Cause of admission Acute respiratory failure Stroke Postoperative Cranial trauma Cause of death Multiple organ failure Hypoxemia Cerebral death

With (n 5 10)

Without (n 5 15)

p Value

60 6 13 9/1 21 6 7 14 6 13 36.1 6 1 16.8 6 9.0 207 6 126 5/5 462 15 6 14 4/6 9 6 11

53 6 18 7/8 19 6 2 13 6 12 36.1 6 1 12.7 6 7.6 210 6 136 13/2 462 12 6 14 2/13 768

0.22 0.04 0.90 0.91 1 0.08 0.66 0.17 1 0.68 0.17 0.8

6 2 1 1

8 3 2 2

4 1 5

5 3 7

* APACHE II 5 acute physiology and chronic health evaluation. † MV 5 mechanical ventilation. ‡ TPN 5 total parenteral nutrition.

METHODS PATIENTS Twenty-five patients who died in our intensive care unit, and who had been mechanically ventilated for . 72 h were studied. In 20 of the 25 patients, pulmonary infection on the day of death was clinically suspected. The causes of admission to the ICU are summarized in Table 1. Twenty-one of the 25 patients (84%) had either diffuse (seven patients) or localized (14 patients) chest radiograph infiltrates 24 h before death. In four patients there were no infiltrates. Chest radiographs were reviewed on the day of the study. The clinical diagnoses of the pulmonary infiltrates on the day of death were: pneumonia in 14 patients, pneumonia and adult respiratory distress syndrome (ARDS) in five, congestive heart failure in one, and alveolar hemorrhage and pneumonia in one. Seventeen patients received prior antibiotic therapy (mean duration of antibiotic treatment was 15 6 15 d), whereas the remaining eight subjects did not receive antibiotics within the 48 h prior to death. Because patients with immunosuppression or hematologic neoplasia may have a unique etiology, immune response, or presentation of pulmonary infection these patients were excluded from this study. In addition, patients with neutropenia, defined as a total neutrophil count , 1,000/ml, were excluded. In each case, family members gave informed written consent for the study to be performed and permission from the Ethical Committee of our center was granted.

Study Protocol Five sampling techniques were performed immediately (taken within a range of 0 to 90 min) after death without discontinuing mechanical ventilation throughout the procedures (FIO2 was set at 1.0). First, endotracheal aspirate samples were obtained by sterile means from all patients, using a 22-in, No. 14 Fr suction catheter (Pennine Healthcare, Derby, UK) and collected in a mucus collector. Second, bilateral fiberoptic bronchoscopies were performed using different bronchoscopes for each lung following standard techniques. Using a special adaptor, the fiberoptic bronchoscope (BF30; Olympus, New Hyde Park, NY) was introduced through the endotracheal tube 10 min later without bronchial suctioning. A protected specimen brush (PSB) sample (Microbiology brush; Mill-Rose Laboratories, Inc., Mentor, OH) was retrieved from the area of maximal inflammation corresponding to

the area of chest X-ray abnormality. Third, protected bronchoalveolar lavage (BAL) specimens (Mill-Rose Laboratories, Inc.) were retrieved by instilling an aliquot of 20 ml sterile saline which was discarded, followed by four aliquots each of 30 ml sterile saline (mean 6 SD recovered BAL fluid was 26 6 16 ml). Fourth, guided lung biopsy sampling was taken as follows: one guided biopsy from each lung, through thoracotomy incisions, was obtained with the aid of the light of the bronchoscope. The bronchoscope was guided to the area of maximal visual inflammation corresponding to chest radiographic infiltrates. In the absence of infiltrates the bronchoscope was guided to the lower lobes. Fifth, bedside blinded (not guided by fiberoptic bronchoscope) bilateral pulmonary biopsies, following strict aseptic techniques, were obtained through the same thoracic incisions. Three fragments were obtained from both superior and inferior lobes from each lung and two from the middle lobe. The thoracotomy incision was done at the fifth intercostal space from the midclavicular to the midaxillary lines. The average size of the bronchoscopically guided biopsies and that of the blind biopsy samples was 2 3 2 3 2 cm each. Biopsy samples were always obtained from peripheral zones of the lung. Each fragment was sectioned into two pieces, one for quantitative microbial cultures, the other for histopathological processing. Samples from 25 patients included endotracheal aspirates (EA) (n 5 25), protected specimen brush (PSB) (n 5 47: 24 left and 23 right), protected bronchoalveolar lavage (BAL) (n 5 47: 24 left and 23 right), guided pulmonary biopsies (n 5 47: 24 left and 23 right), and blind lung biopsies (n 5 47 lungs: 328 biopsies). Three patients had been previously pneumonectomized.

Microbiological Processing Biopsy samples were sent to the Microbiology Laboratory for fungal and bacterial isolation and identification. Tissue specimens were placed, together with sterile sand and 3 ml of sterile saline, in a mortar (Vidrafoc®, ICIV, S.A., Barcelona) and homogenized. Serial dilutions (1021, 1022, 1023) of each sample were prepared in sterile normal saline. One hundred microliters of each dilution of endotracheal aspirate samples, PSB, BAL and biopsy samples were inoculated onto the following agar media: 5% sheep blood, chocolate, Centers for Disease Control and Prevention (CDC) blood, McConkey, blood charcoal yeast extract (BCYE-a), and Sabouraud-dextrose agar. All cultures were incubated at 378 C under aerobic and anaerobic conditions and

El-Ebiary, Torres, Fàbregas, et al.: Isolation of Candida from Respiratory Samples in CO2-enriched atmosphere. Cultures were evaluated for growth 24 h and 48 h later and discarded, if negative, 5 d after, except for CDC and Wilkins-Chalgren that were evaluated at 7 d and for Sabouraud at 6 wk. All microorganisms isolated were identified by standard laboratory methods (5). Results are expressed as colony forming units per gram of tissue (or per milliliter) (cfu · g21 5 number of colonies 3 dilution factor 3 inoculation factor).

Histopathological Study All lung tissue specimens were embedded in paraffin and stained with hematoxylin/eosin, periodic acid Schiff, May-Grünwald Giemsa and Gram stains. Pathologists were unaware of the results of clinical or other laboratory data. Candida microorganisms were identified histologically on the basis of typical morphologic features in the stained sections. The histologic diagnosis of Candida pneumonia was established according to two morphologic varieties: hematogenous pulmonary candidiasis characterized by small (2 to 4 mm) miliary nodules randomly distributed in the pulmonary parenchyma. The nodules have central necrosis with varying amounts of acute inflammation. Clusters of both pseudohyphae and budding yeasts can be found in the center of the nodules. The pseudohyphae consist of elongated blastopores that are arranged in chains resembling sausage links; the yeasts are round, ranging from 3 mm to 6 mm. The second form of pulmonary candidiasis is related to aspiration. Distribution of the lesions with clusters of pseudohyphae and/or yeasts around the bronchioles is considered to be evidence of aspiration. There may be an associated acute inflammatory cell infiltrate with bronchopneumonia and abscess formation (6–8). On the other hand, pathological criteria for unspecific pneumonia included foci with accumulation of polymorphonuclear leukocytes in the capillaries and adjacent alveolar spaces corresponding to various degrees of evolution and extension (9). For the purpose of the study, definite Candida pneumonia was considered if pulmonary biopsy samples were compatible with Candida pneumonia, or Candida species were isolated from pleural fluid.

Statistical Analysis Results are expressed as mean 6 SD. The t-test was employed for the comparison of quantitative variables (Mann-Whitney’s tests if the distribution was not normal). The chi square test (Fisher’s exact test when needed) was used to compare proportions. Spearman’s correlation was used to correlate log10 of the counts of the different sampling techniques. A logistic regression model was applied to adjust for confounding variables for assessing risk of developing positive respiratory samples for Candida. Common pitfalls associated with multivariable regression were avoided as described by Concato and coworkers (10). Continuous variables were categorized. The adjusted relative risk for developing positive respiratory samples for Candida was estimated for the following variables: presence of antibiotics (yes/no), days of antibiotic treatment (> 7 or , 7), mechanical ventilation period (> 5 or , 5), age, ARDS (yes/no), parenteral nutrition (yes/no), and sex (male/ female). These variables were selected according to their univariate analysis p value, or because they were deemed biologically plausible. The significance levels were set to 0.05. Diagnostic value for the sampling techniques (any counts for Candida spp.) were calculated in two different ways: using the presence or absence of unspecific histological pneumonia as described above, or using the definite confirmation of pulmonary candidiasis (histologic Candida pneumonia, positive pleural fluid culture).

RESULTS Study Population

Patients were categorized as having positive pulmonary biopsy cultures for Candida spp. or otherwise. Ten patients (40%) had at least one pulmonary biopsy sample yielding Candida spp.; 15 did not (60%). Only two patients (8%) had definite pulmonary candidiasis (compatible postmortem histology with Candida pneumonia in one case, and positive antemortem pleural fluid culture for Candida in the second). General characteristics at the inclusion in the study of both patients with and without Candida spp. and causes of ICU ad-

585

mission and death are described in Table 1. Sex was the only statistically significant variable when comparing both groups. No differences were found when comparing APACHE II scores, leukocyte count, and administration of parenteral nutrition between patients with and without Candida-positive biopsy samples (Table 1). The mean 6 SD duration of antibiotic treatment before death was 15 6 15 and 12 6 14 d for patients with and without positive pulmonary samples for Candida, respectively (p 5 0.61). Among the 10 patients with pulmonary biopsy samples for Candida spp., five received antibiotics while in five patients antibiotics were withheld 48 h before death. The average number of antibiotics given to each patient from either group was 4 6 2 drugs. Two patients of 25 received prior antifungal treatment before death: the first subject was treated with fluconazole on an empiric basis (in this particular case, Candida pneumonia due to Candida krusei was confirmed histologically after death) and the second received amphotericin B due to the antemortem isolation of Aspergillus fumigatus from a PSB sample. In this particular case, Aspergillus fumigatus was isolated in all samples from both lungs in the postmortem study. Additionally, postmortem histologic examination confirmed the presence of histologic pneumonia for Aspergillus in this case. For evaluating risk factors associated with the development of positive pulmonary samples for Candida, a logistic regression model was fit, adjusted for the following variables: presence of antibiotics (yes/no), days of antibiotic treatment (> 7 or , 7), mechanical ventilation period (> 5 or , 5), age, ARDS (yes/no), parenteral nutrition (yes/no), and gender (male/female). None of the seven variables demonstrated to be an independent risk factor for developing positive pulmonary samples for Candida species. Histopathological Findings

Samples from patients with positive postmortem lung cultures for Candida (n 5 10). Only lung biopsies (both blind and guided) samples from Patient 20 were compatible histologically with pulmonary candidiasis. The isolated yeast in this particular patient was Candida krusei in both antemortem and postmortem samples. Seventy-two of 144 (50%) pulmonary biopsy samples obtained from the 10 patients with positive pulmonary cultures for Candida were compatible with unspecific histologic pneumonia. Candida species was isolated from 44 of 72 of the positive biopsy samples for unspecific histologic pneumonia. On the other hand, Candida species was isolated from 35 of the remaining 72 biopsy samples without unspecific histologic pneumonia. Only two patients had negative lung biopsies for pneumonia in all specimens. The remaining eight patients had at least one positive sample for pneumonia. Mean 6 SD positive blind biopsy samples for unspecific histological pneumonia per patient was 5 6 5. With regard to guided lung biopsies, five of 18 (28%) lungs did not show unspecific histologic pneumonia. Other associated pulmonary diseases found at autopsy are described in Table 3. The most frequent associated conditions were diffuse alveolar damage (n 5 5), followed by alveolar hemorrhage (n 5 2). Samples from patients with negative lung cultures for Candida (n 5 15). None of the lung biopsy samples retrieved from these patients showed histologic Candida pneumonia. One hundred and two of 231 (44%) pulmonary biopsy samples obtained from the 15 patients with negative cultures for Candida were compatible with unspecific histologic pneumonia. All patients from this group had at least one blind biopsy compatible with unspecific histologic pneumonia. With regard to guided

586


lung biopsies, 11 of 29 (38%) lungs did not show histologic pneumonia. Microbiological Results

Growth of Candida species in any count was seen in 10 of 25 (40%) patients in at least one pulmonary biopsy sample (blind and bronchoscopically guided biopsies) (Table 2). In these 10 cases, Candida species was isolated on nine occasions from PSB; five from endotracheal aspirates; six from BAL; 13 from blind lung biopsies; and 12 from guided biopsies. Eight patients yielded Candida albicans, one yielded Torulopsis glabrata (Patient 6), and the other Candida krusei (Patient 20). In these 10 patients other microorganisms were always isolated from at least one respiratory sampling technique (Table 2). Among the 10 patients with positive pulmonary biopsies for Candida species, nine had premortem microbiologic results before death. Only three patients had positive antemortem samples for yeasts. The first (Patient 6) had positive endotracheal aspirate cultures of Torulopsis glabrata (4 3 103 cfu/ml), besides the patient had positive pleural fluid culture and gastric juice cultures for the same microorganism. The second (Patient 8) had positive pharyngeal swab for Candida albicans

VOL. 156

1997

(2 3 105 cfu/ml); and the third (Patient 20) had positive endotracheal aspirate and blood cultures for Candida krusei. Biopsy cultures. A total of 470 microorganisms were isolated in any count from 280 of 375 (77%) lung biopsy samples in all 25 patients. Forty percent (190 of 470) of the isolates were gram-negative bacilli, of which 72% (137 microorganisms) were Pseudomonas spp. Thirty-eight percent of the isolates (178 microorganisms) were gram-positive cocci, and 12% (56 microorganisms) included non pathogenic microorganisms (e.g., Neisseria spp., Staphylococcus epidermidis, Streptococcus sanguis). Ten percent (46 of 470) of the isolates represented fungi and yeasts: Candida spp. (40 isolates corresponding to 10 patients; 9%) and Aspergillus fumigatus (six isolates; 1%) (Figure 1). Other samples. A total of 65, 42, and 63 microorganisms were isolated from PSB, EA, and BAL, respectively. Candida species constituted 9 of 65 (14%), 5 of 42 (12%), and 6 of 63 (9%) of PSB, EA, and BAL cultures, respectively (Table 2). In only one case blood cultures were positive (Candida krusei). In this particular subject (Patient 20), Candida krusei was also isolated from all the respiratory samples. Qualitative and quantitative cultures for Candida spp.; agreement and correlations among techniques. Pulmonary sam-

TABLE 2 MICROORGANISMS ISOLATED FROM DIFFERENT TECHNIQUES FROM 10 PATIENTS WHO HAD CANDIDA SPECIES IN AT LEAST ONE PULMONARY BIOPSY SAMPLE No.

Side

1

Right

C. krusei 10 S. aureus 10

C. krusei 20

Left

C. krusei 10 S. aureus 10

A. fumigatus S. aureus 3 3 102

Right

S. aureus 2 3 102

Left

S. aureus 1 3 102

S. aureus 6 3 103 T. glabrata 1 3 103 S. aureus 3 3 103

7

Right

P. aeruginosa 2 3 10 C. albicans 5 3 103

8

Right Left

P. aeruginosa 1 3 104 P. aeruginosa 1 3 106

6

PSB

Blind Biopsies

4

Endotracheal Aspirate C. krusei 2 3 103 S. aureus 2 3 103

Right

—

S. aureus 2 3 107 T. glabrata 10

C. krusei 6 3 102 C. krusei NQ

S. aureus 4 3 103

T. glabrata 7 3 102

S. aureus 1 3 104

T. glabrata NQ

P. aeruginosa 9 3 10 C. albicans 10

P. aeruginosa 1 3 10 S. aureus 7 3 105 NO SAMPLE

P. aeruginosa 2 3 102 C. albicans 30

C. albicans NQ

P. aeruginosa 4 3 105 P. aeruginosa 8 3 104 C. albicans 10

P. aeruginosa 2 3 106


C. albicans 1 3 102 X. maltophila 7 3 102 X. maltophila 3 3 102

S. viridans 10


Right Left

S. pneumoniae 2 3 104

S. pneumoniae 5 3 103 C. albicans 3 3 102

19

Right Left

C. albicans 1 3 104 C. albicans 1 3 103

C. albicans 2 3 103 C. albicans 30

20

Right

C. krusei 3 3 102 P. aeruginosa 2 3 102 C. krusei 70 P. aeruginosa 40

Left

Bronchoscopically Guided Biopsies

7

6

Left 18

S. epidermidis 2 3 102 S. aureus 10 —

Left

16

BAL

C. albicans NQ — P. aeruginosa 1 3 102 NO SAMPLE S. pneumoniae 2 3 105

S. pneumoniae 4 3 103

C. albicans 3 3 102

C. albicans 3 3 105

C. albicans 2 3 104 C. albicans 30

C. albicans 7 3 103 C. albicans 2 3 103

C. krusei 3 3 102 P. aeruginosa 30 C. krusei 6 3 102 P. aeruginosa 70

C. krusei 2 3 105

P. aeruginosa 20

C. krusei 2 3 103

C. krusei 40 P. aeruginosa 30

C. krusei 3 3 102

22

Right Left

S. liquefaciens 1 3 102 S. liquefaciens 1 3 103

S. liquefaciens 20 S. liquefaciens 3 3 102 C. albicans 1 3 102

S. liquefaciens 1 3 107 S. liquefaciens 60

S. liquefaciens 10

C. albicans 1 3 102

25

Right

S. epidermidis 1 3 102 P. putida 8 3 102 S. epidermidis 1 3 104 C. albicans 1 3 103

S. epidermidis 2 3 102 C. albicans 60 —

S. epidermidis 7 3 103 C. albicans 1 3 102

S. epidermidis 8 3 103 C. albicans 1 3 103 S. epidermidis 2 3 103 C. albicans 20

C. albicans 6 3 102

Left

Definition of abbreviations: PSB 5 protected specimen brush; (—) 5 sterile culture. Patients from 1 through 16 were receiving broad-spectrum antibiotics. Figures after microbial names represent counts of the isolated species in cfu/ml for PSB and endotracheal aspirates, and in cfu/g for biopsy samples.

587

El-Ebiary, Torres, Fàbregas, et al.: Isolation of Candida from Respiratory Samples

Figure 1. Number and percentages of isolates from lung tissue cultures. Note that Candida spp. represented 9% of the total number of isolates. Others included non-pathogenic microorgansims (e.g., Neisseria spp., S. epidermidis, S. sanguis).

pling methods yielded the following geometrical mean 6 SD counts for Candida species: PSB (2.5 6 1.2 log10 cfu/ml, 95% CI: 1.4 to 3.6); EA (3.4 6 1.8 log10 cfu/ml, 95% CI: 1.5 to 5.2); blind pulmonary biopsies (2.0 6 0.7 log10 cfu/g, 95% CI: 1.5 to 2.5); BAL (2.3 6 1.3 log10 cfu/ml, 95% CI: 0.6 to 4.4); and guided pulmonary biopsies (2.2 6 1.0 log10 cfu/g, 95% CI: 1.4 to 3.0). There were no statistically significant differences among the means of quantitative Candida cultures of the five sampling techniques. (Figure 2). Table 4 shows both the qualitative agreement and the correlation coefficients as well as their p values between each pair of respiratory sampling techniques for quantitative cultures of Candida spp. Qualitative agreement among techniques ranged from 40 to 82% for all microorganisms. All five sampling tech-

niques showed good and significant correlation among each other. Diagnostic values of Candida cultures in different samples. Using the presence of unspecific histologic pneumonia as a gold standard, the sensitivities of the different diagnostic methods were: 62, 40, 17, 100, and 54%, for endotracheal aspirates, PSB, BAL, guided lung biopsies, and blind biopsies, respectively, while the specificities were: 100, 17, 55, 20, and 37%, for the same sample order, respectively. When calculating the diagnostic value parameters using the presence of confirmed pulmonary candidiasis as gold standard (histology or pleural fluid culture), the sensitivity of the different diagnostic methods were: 100, 50, 50, 75, and 100% for endotracheal aspirates, PSB, BAL, guided lung biopsies, and blind biopsies, respec-

TABLE 3 HISTOLOGIC FINDINGS FROM 10 PATIENTS WITH POSITIVE LUNG TISSUE SAMPLES FOR CANDIDA AND ANTEMORTEM CULTURES

No.

Side

1

Right Left Right Left Right Left Right Left Right Left Right Left Right Left Right Left Right Left Right Left

6 7 8 16 18 19 20† 22 25

Bronchoscopically Guided Biopsy Pneumonia Yes Yes Yes Yes No No No Yes Yes Yes Yes No Yes No Yes Yes Yes No

Associated Pulmonary Disease (autopsy)

Lung Tissue Slide with Pneumonia/Total Slides

Alveolar hemorrhage Hyaline membranes Interstitial fibrosis

2/8 1/6 3/8 1/7 None 6/8 NO SAMPLES DAD* 0/8 2/6 None 0/8 0/6 NO SAMPLES DAD (exudative stage) 5/6 Pulmonary hypertension 8/8 DAD 6/6 Pulmonary hypertension 8/8 6/6 DAD (fibroproliferative 8/8 stage). Fat embolism 0/6 Aspiration pneumonitis 3/8 and bronquiolitis 0/6

* DAD 5 diffuse alveolar damage. † Patient received IV fluconazol 400 mg/12 h based on the antemortem isolation of Candida krusei.

Antemortem Samples with Candida Negative Torulopsis glabrata in pleural fluid and gastric juice Negative C. albicans 2 3 105 cfu/ml in pharyngeal swab No samples

Negative Negative C. krusei in endotracheal aspirate Negative No samples

588


Figure 2. Geometrical means (6 SD) of Candida spp. cultures in endotracheal aspirates (EA), bronchoalveolar lavage (BAL), protected specimen brush samples (PSB), guided lung biopsy cultures, and blind lung biopsy cultures. Horizontal lines represent means. There were no statistically significant differences among the mean counts of the five sampling techniques.

tively, while the specificities were: 67, 55, 70, 33, and 20% for the same sample order, respectively.

DISCUSSION The main findings of our study are the following: (1) the incidence of Candida isolation from pulmonary biopsies in critically ill, mechanically ventilated, non-neutropenic patients who die is high (10 of 25; 40%), however, definite Candida pneumonia was established in two of 25 (8%) patients; (2) Candida colonization is uniform throughout the different lung regions; (3) the use of quantitative cultures of Candida in respiratory samples is not helpful for establishing this type of infection. The incidence of Candida pneumonia varies among different studies. Data from several studies including different subsets of patients show an incidence varying from 0.23 to 4.5% (4, 11, 12). However, it is important to note that most of these reports included cases with disseminated candidiasis (including lung involvement) or cases without histopathologic documentation of candidal infection. Information regarding Candida

VOL. 156

pneumonia in non-neutropenic patients requiring mechanical ventilation is scanty. In two studies (13, 14), the incidence of isolation of fungi—including yeasts—in ventilator-associated pneumonia was 3.5% and 4.5%, respectively. Fagon and coworkers (15) studying 205 ICU patients found that among patients infected by Candida albicans, none developed lower respiratory tract infection. In our histopathological study, the incidence of Candida isolation from pulmonary biopsies was 40% (10 of 25 patients), and the incidence of definite Candida pneumonia was 8% (two of 25 patients). Several histopathologic postmortem studies in critically ill patients have demonstrated that yeasts represent 7 to 17% of the microorganisms identified from pulmonary tissue cultures (16–19). From all these histopathologic studies, it seems that the presence of Candida in lung cultures is a frequent event in critically ill patients who die in the ICU. A number of risk factors for severe candidal infections have been identified in critical care patients. These include among others the number and duration of antibiotics, parenteral feeding, the use of multiple invasive devices, the length of ICU stay, APACHE II score and mechanical ventilation. Very little information regarding the risk factors for developing colonization or pulmonary Candida infection in mechanically ventilated patients has been described (15). Fagon and colleagues found that prior antibiotic treatment and steroid treatment were more frequent in critically ill patients with lung Candida infection or colonization (15). We performed a stepwise logistic regression model and found no risk factors that could have influenced the appearance of positive lung tissue cultures for Candida. However, this negative finding could be influenced by the small number of patients studied. Additionally, some variables could have been missed in our study since it was not initially designed to describe risk factors. In our study we found that the presence of Candida in the lung was uniformly distributed as assessed by the different sampling techniques performed bilaterally, and the multiple blind biopsy samples obtained. These findings suggest that Candida colonization is extensive throughout the respiratory tree in some critically ill patients undergoing mechanical ventilation. We ignore colonization in other organs, as this was not a primary goal of our study. Given the high postmortem rates of recovery of Candida spp. and the relatively infrequent antemortem recovery of this microorganism, extensive Candida colonization may be a terminal event, and probably, this could also be interpreted as a marker of poor prognosis (20). We believe that studies following up those patients with positive respiratory sample cultures isolated in vivo are needed to better

TABLE 4 AGREEMENT AND CORRELATIONS AMONG THE FIVE SAMPLING TECHNIQUES

Pair of Diagnostic Techniques PSB PSB PSB PSB Guided biopsy Guided biopsy Guided biopsy EA EA Blind biopsy

versus versus versus versus versus versus versus versus versus versus

EA BAL Blind biopsy Guided biopsy BAL Blind biopsy EA BAL Blind biopsy BAL

1997

Qualitative Agreement

Correlation Coefficient

p Value

Yes (%)

No (%)

0.7 0.88 0.7 0.8 0.95 0.85 0.93 0.8 0.7 0.88

0.01 , 0.0001 0.007 0.001 0.001 0.001 0.001 0.001 0.005 0.0001

8 (73%) 13 (65%) 9 (82%) 9 (82%) 10 (50%) 12 (60%) 8 (73%) 6 (54%) 6 (54%) 8 (40%)

3 (27%) 7 (35%) 11 (18%) 11 (18%) 10 (50%) 8 (40%) 3 (27%) 5 (46%) 5 (46%) 12 (60%)

Definition of abbreviations: PSB 5 protected specimen brush; EA 5 endotracheal aspirates; BAL 5 bronchoalveolar lavage.

El-Ebiary, Torres, Fàbregas, et al.: Isolation of Candida from Respiratory Samples

clarify the real meaning of these cultures. On the other hand, there was an excellent correlation between quantitative biopsy lung cultures for Candida and other sampling techniques. Conversely, we did not find such good correlations for other microorganisms causing ventilator-associated pneumonia in a previous postmortem study (17). A possible explanation for the discrepancies among studies is the absence of antifungal treatment in most of our patients before the study protocol. This does not apply to the latter study, in which all patients received prior antibacterial treatment, thus modifying the bacterial concentration in respiratory secretions and lung tissue. Criteria for the diagnosis of Candida pneumonia remain to be defined. Because there is no pathognomonic clinical picture for Candida pneumonia, diagnosis has relied for years on detecting the presence of Candida in sputum or bronchoscopic specimens. However, each of these methods is subject to error since Candida is a normal inhabitant of the mouth, and can be recovered from sputum in 20 to 55% of normal subjects (21). In our study, only two patients with positive Candida cultures from lung biopsies had true Candida lung infection. Not even the use of quantitative cultures could help in distinguishing infection from colonization in our patients. This contrasts with bacteria in which several thresholds have been established to help distinguish colonization from infection (22, 23). In our study the diagnostic performances of PSB, endotracheal aspirates, BAL, blind lung biopsies, and bronchoscopically guided lung biopsies were not optimal and varied among the different sampling techniques. Saito and colleagues reported that BAL had a sensitivity of 75% and a specificity of 100% for the diagnosis of Candida pneumonia in patients with acute leukemia and severe neutropenia (24). From our results we deduce that the use of the commonly available respiratory sampling methods (bronchoscopic or nonbronchoscopic) in mechanically ventilated patients is insufficient for the diagnosis of Candida pneumonia. Others have recommended the use of transbronchial biopsies, but this is formally contraindicated in mechanically ventilated patients. Caution should be taken when extrapolating conclusions from our study to in vivo patients, since our samples were taken after death. The significance of Candida cultured at autopsy is unclear. This yeast can be cultured at autopsy without histologic or clinical evidence of pulmonary disease or of fungal tissue invasion (25). In our immediate postmortem study, 10 out of 25 (40%) had positive biopsy cultures of Candida spp. However, in only one case was Candida infection confirmed histopathologically. The only certain method to establish that Candida is the primary lung pathogen is to demonstrate yeast or pseudohyphae in a lung biopsy specimen (25). The presence of mycelial forms is said to provide solid evidence of tissue invasion by Candida (26). Other authors feel there is no difference between yeast and mycelial forms in lung tissue for the definite diagnosis of Candida pneumonia (27). Histologic examination of lung tissue specimens screening for Candida ought to be performed carefully. In our study we performed three types of stains, hematoxylin/eosin, periodic acid Schiff, and May-Grünwald Giemsa, on our histologic preparations to screen for fungi. Silver stains such as Gomori methenamine are better special stains for screening fungi. On the other hand, we debate the significance of the isolation of Candida from lung tissue, accompanied by histologic changes of pneumonia, but without the histologic demonstration of the yeasts in lung tissue. Should this be considered as primary Candida pneumonia, then the incidence in our study would be higher (36%). Further studies are imperative to clarify this issue. Deciding when to initiate antifungal therapy for pulmonary candidiasis is often troublesome, because of the difficulty of

589

detecting true Candida pneumonia. In clinical practice, lung biopsies cannot be used for the management of patients with suspected Candida infection. Colonization at more than two sites has been proposed previously as the key to start therapy in high risk surgical patients, but even this rule is not really predictive of Candida infection as was recently observed by Pittet and coworkers (28, 29). What is clear now is the practice of treating patients with positive blood cultures and/or signs of hematogenous infection (e.g., Candida endophthalmitis) regardless of the presence or absence of neutropenia (30). Nevertheless, some patients with deep organ Candida infection may have blood cultures that are consistently negative (31). Surveillance studies of mechanically ventilated patients with positive respiratory samples for Candida species are required to establish criteria of when to start antifungal therapy in absence of blood cultures. In summary, the incidence of Candida isolation from pulmonary biopsies in critically ill mechanically ventilated, nonneutropenic patients who die in the ICU is high. The incidence of definite Candida pneumonia in this population is 8%. We also found that Candida colonization is uniform throughout the different lung regions, and that the presence of Candida in respiratory samples, independently of quantitative cultures, is not a good marker of Candida pneumonia. Future research regarding markers of Candida pneumonia, the significance of antemortem isolation of Candida in respiratory samples, and the investigation of the etiopathogenesis of this fungal pneumonia in mechanically ventilated patients is warranted. References 1. Edwards, J. E. 1991. Invasive candida infections: evolution of a pathogen. N. Engl. J. Med. 324:1060–1062. 2. Williams, D. M., J. A. Krick, and J. S. Remington. 1976. Pulmonary infection in the compromised host. Am. Rev. Respir. Dis. 114:359–394. 3. Murray, P. R., R. E. VanScoy, and G. D. Roberts. 1977. Should yeasts in respiratory secretions be identified? Mayo Clin. Proc. 52:42–45. 4. Haron, E., S. Vartivarian, E. Anaissie, R. Dekmezian, and G. Bodey. 1993. Primary Candida pneumonia. Medicine (Baltimore) 72:137–142. 5. Balows, A., and W. J. Harsler. 1991. Manual of Clinical Microbiology, 5th ed., Section III. American Society for Microbiology, Washington, DC. 209–553. 6. Dubois, P. J., R. L. Myerowitz, and C. M. Allen. 1977. Pathoradiologic correlation of pulmonary candidiasis in immunosuppressed patients. Cancer 40:1026–1036. 7. Katzenstein, A. L., and F. B. Askin. 1990. Infection: unusual pneumonias. In Surgical Pathology of Non-neoplastic Lung Disease, Ch. 10. 343–352. 8. Rose, H. D., and N. K. Sheth. 1978. Pulmonary candidiasis: a clinical and pathological correlation. Arch. Intern. Med. 138:964–965. 9. Fàbregas, N., A. Torres, M. El-Ebiary, J. Ramírez, C. Hernández, J. González, J. Puig de la Bellacasa, M. T. Jiménez de Anta, and R. Rodriguez-Roisin. 1996. Histopathological and microbiological aspects of ventilator-associated pneumonia. Anesthesiology 84:760–771. 10. Concato, J., A. R. Feinstein, and T. R. Holford. 1993. The risk of determining risk with multivariable models. Ann. Intern. Med. 118:201–210. 11. Masur, H., P. P. Rosen, and D. Armstrong. 1977. Pulmonary diseases caused by Candida species. Am. J. Med. 63:914–925. 12. Lihartova, A., and W. Chung. 1963. Bronchopulmonary moniliasis in the newborn. J. Clin. Pathol. 16: 56–60. 13. Rello, J., V. Ausina, M. Ricart, J. Castella, and G. Prats. 1993. Impact of previous antimicrobial therapy on the etiology and outcome of ventilator-associated pneumonia. Chest 104:1230–1235. 14. Torres, A., R. Aznar, J. M. Gatell, P. Jiménez, J. González, A. Ferrer, R. Celis, and R. Rodriguez-Roisin. 1990. Incidence, risk, and prognosis factors of nosocomial pneumonia in mechanically ventilated patients. Am. Rev. Respir. Dis. 142:523–528. 15. Fagon, J. Y., V. Lavarde, A. Novara, G. Mahieu, F. Stephan, A. Buu, and M. Safar. 1994. Nosocomial candida infections of the lower respiratory tract in ICU patients (abstract). Am. J. Respir. Crit. Care Med. A650. 16. Rouby, J. J., E. M. Martin de Lassale, P. Poete, M. H. Nicolas, L. Bodin,

590

17.

18.

19.

20.

21. 22.


V. Jarlier, Y. Le Charpentier, J. Grosset, and P. Viars. 1992. Nosocomial bronchopneumonia in the critically ill. Am. Rev. Respir. Dis. 146: 1059–1066. Torres, A., M. El-Ebiary, L. Padró, J. González, J. Puig de la Bellacasa, J. Ramirez, A. Xaubet, M. Ferrer, and R. Rodriguez-Roisin. 1994. Validation of different techniques for the diagnosis of ventilator-associated pneumonia. Am. J. Respir. Crit. Care Med. 149:324–331. Papazian, L., P. Thomas, L. Garbe, I. Guignon, X. Thirion, J. Charrel, C. Bollet, P. Fuentes, and F. Gouin. 1995. Bronchoscopic or blind sampling techniques for the diagnosis of ventilator-associated pneumonia. Am. J. Respir. Crit. Care Med. 152:1982–1991. Chastre, J., F. Viau, P. Brun, J. Pierre, M. C. Dauge, A. Bouchama, A. Akesbi, and C. Gibert. 1984. Prospective evaluation of the protected specimen brush for the diagnosis of pulmonary infections in ventilated patients. Am. Rev. Respir. Dis. 130:924–929. Pittet, D., and R. P. Wenzel. 1995. Nosocomial bloodstream infections: secular trends in rates, mortality and contribution to total hospital deaths. Arch. Intern. Med. 155:1177–1184. Baum, G. L. 1960. The significance of Candida albicans in human sputum. N. Engl. J. Med. 263:70–73. Torres, A., J. Puig de la Bellacasa, A. Xaubet, R. Rodriguez-Roisin, M. T. Jiménez de Anta, and A. Agustí-Vidal. 1989. Diagnostic value of quantitative cultures of bronchoalveolar lavage and telescoping plugged catheters in mechanically ventilated patients with bacterial pneumonia. Am. Rev. Respir. Dis. 140:306–310.

VOL. 156

1997

23. Chastre, J., J. Y. Fagon, M. Bornet-Lesco, S. Calvat, M. C. Dombret, and R. Al Khani. 1995. Evaluation of bronchoscopic techniques for the diagnosis of nosocomial pneumonia. Am. J. Respir. Crit. Care Med. 152: 231–240. 24. Saito, H. E., E. J. Anaisse, R. C. Morice, R. Dekmezian, and G. P. Bodey. 1988. Bronchoalveolar lavage in the diagnosis of pulmonary infiltrates in patients with acute leukemia. Chest 94:745–749. 25. Haley, L. D., and A. McCabe. 1950. A mycological study of 71 autopsies. Am. J. Clin. Path. 20:35–42. 26. Kozinn, P. J., and C. L. Taschdjian. 1966. Candida albicans: Saprophyte or pathogen? J.A.M.A. 198:170–174. 27. Taschdjian, C. L., P. J. Konzinn, and E. F. Toni. 1970. Opportunistic yeast infection with special reference to Candidiasis. Ann. N.Y. Acad. Sci. 174:606. 28. Pittet, D., M. Monod, P. M. Suter, E. Frenk, and R. Auckenthaler. 1994. Candida colonization and subsequent infections in critically ill surgical patients. Ann. Surg. 220:751–758. 29. Solomkin, J. S., A. B. Flohr, P. G. Quie, and R. L. Simmons. 1980. The role of Candida in intraperitoneal infections. Surgery 88:524–530. 30. Pittet, D., and J. Garbino. 1995. Fungal infections in the critically ill. Current Opinion in Critical Care 1:369–380. 31. British Society for Antimicrobial Chemotherapy. 1994. Management of deep Candida infection in surgical and intensive care unit patients. Intensive Care Med. 20:522–528.