INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The reported incidence of ventilator-associated pneumonia (VAP) is from 9% (1) to 40% (2). The high sensitivity and the specificity required for the diagnosis of pneumonia account for a large part of this variability. New quantitative microbiologic techniques based on the use of distal and protected specimens appear to give lower estimates of the incidence of pneumonia (1) than do classic criteria. Differences in patient selection and management and the application of preventive measures may also contribute to the discrepancies between reported incidences. In contrast, epidemiologic studies of nosocomial pathogens in intensive care units (ICU) describe consistent and high frequencies of Pseudomonas aeruginosa in VAP: from 17.2% (3) to 20.8% (4), and even 23% (5). Despite the advent of effective broad-spectrum antibiotics, the mortality of nosocomial lung infections caused by P. aeruginosa remains very high (6). Preventive measures used include: mechanical subglottic secretion drainage (7), use of sucralfate for prophylaxis for stress ulcer (8), selective digestive decontamination (SDD) (9, 10), and use of intratracheal antibiotics (10, 11). However, the efficacy of these measures has not been assessed for the reduction of the incidence of VAP with P. aeruginosa. At the Besançon University hospital the prevalence of P. aeruginosa in VAP in the surgical intensive care unit (SICU) was higher than that of other nosocomial pathogens such as Acinetobacter species or oxacillin-resistant Staphylococcus aureus. These observations prompted the present investigations aimed at documenting patterns of bronchopulmonary tract colonization with P. aeruginosa and assessing risks and routes of colonization and infection. This prospective study was performed with the main goal of identifying factors associated with P. aeruginosa ventilator-associated colonization so that interventions aimed at reducing the incidence of ventilator associated colonization could be developed. The second goal of this study was to identify which factors between bacteriologic characteristics and clinical features were possible interventional factors to reduce P. aeruginosa ventilator-associated pneumonia in the colonized patient.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Design

In a 12-mo prospective study, 6,026 specimens, collected from 190 patients admitted to the SICU (15 beds) of the University Hospital of Besançon, France, for a stay of more than 48 h, and who were mechanically ventilated, and 576 specimens collected from the environment of the patients in the unit, were screened for presence of P. aeruginosa. Stools, nasal and cutaneous (groin and armpit) swabs, tracheal secretions, and urine were collected as soon as possible after admission to the unit and then once a week throughout the patient's stay. All P. aeruginosa strains were typed by determination of DNA patterns by pulsed-field gel electrophoresis (PFGE). Data collected from patients were compared to identify risk factors for bronchopulmonary colonization with P. aeruginosa. Extracellular products of strains isolated from colonized patients and from patients with pneumonia were compared as were medical data collected from the two types of patients to identify causes of infection.

Bacteriologic Sampling

Specimens from patients were cultured as follows: aspirate tracheal secretions were washed and liquefied, and then cultured before and after two serial 100-fold dilutions on blood agar and Drigalsky agar (Diagnositics Pasteur, Marnes la Coquette, France); nasal, and cutaneous specimens were cultured on Columbia agar with 5% horse blood (Diagnostics Pasteur) at 41° C. Rectal swabs and urine, and environmental specimens were cultured on Mueller-Hinton agar and Drigalsky agar (Diagnostics Pasteur) at 41° C. Swabs were taken from the environment monthly: from sink plug-hole and tap handles in each of the 15 rooms and from the nursing station of the unit. Swabs were used to inoculate onto blood agar, which was incubated for as long as 48 h at 41° C.

Bacteriologic Methods

P. aeruginosa isolates were identified using the API 20NE system (Biomerieux, Lyon, France).

Genotyping Methods

Five identical colonies of each positive culture were genotyped as previously described by Talon and colleagues (12). DraI endonuclease was used because it gave clear and well-defined DNA profiles.

Analysis of DNA Relatedness

Electrophoretic restriction patterns were analyzed by scanning photographic negatives. GelCompar software was used for cluster analysis (Applied Maths^TM, Kortrijk, Belgium). Each strain was first compared with all other strains to calculate similarity using the Dice correlation coefficient. The strains were then grouped and the groups were depicted as a dendrogram using the UPGMA clustering algorithm (unweighted pair group method using arithmetic averages). Major restriction patterns were defined as those differing by more than three fragments with a similarity index < 85% according to Talon and colleagues (12). Major genotypes were labeled with numerals. A sample of SmaI restricted DNA of Staphylococcus aureus NCTC 8325 was included for each three or four plugs in each run as internal reference.

Enzyme Assays

Elastase activity, lecithinase, DNase, gelatinase, lipase, and hemolysin productions were first tested according to Molinari and colleagues (13). Clinical isolates were then examined quantitatively for levels of total protease, elastase and phospholipase C produced in vitro under defined conditions (14). P. aeruginosa strains PA103 and PAO1, identified secretory phenotypes, were used as references.

Data Collection and Baseline Data

All included patients were prospectively followed up daily until discharge from the SICU. The data collection was coordinated by a physician of the hygiene and nosocomial infection unit, who completed, for each included patient, standardized forms recording demographic characteristics, chronic diseases, diagnosis, clinical features, severity and laboratory findings at SICU admission, and treatment modalities (surgical and nursing procedures and medication) in the SICU with particular regard to respiratory support, antimicrobial treatment, and outcome. Prior antibiotics were first analyzed as a whole group and, in a second step as separate products: prior antibiotic Y/N, if yes, aminoglycosides Y/N, metronidazole Y/N, fluoroquinolone Y/N, fucidic acid Y/N, -lactamine Y/N, if -lactamine yes, potent antipseudomonal penicillin (piperacillin) Y/N, less potent antipseudomonal penicillin (aminopenicillin, penicillin G with or without inhibitor) Y/N, second generation cephalosporin Y/N, potent antipseudomonal third generation cephalosporin (ceftazidime) Y/N, less potent antipseudomonal third generation cephalosporin (cefotaxime) Y/N, imipenem Y/N.

Definitions

SICU acquired colonization. Patients with P. aeruginosa negative culture of the first tracheal secretion and positive culture of the subsequent tracheal secretion were defined as patients who had acquired colonization.

Pneumonia. Pneumonia was diagnosed by the staff physician when new and persistant chest radiographic pulmonary infiltrates, not otherwise explained, were detected, together with suggestive clinical (purulent sputum, fever > 38° C, impaired oxygenation, or white blood cell count > 10 × 10⁹/L) and microbiologic (repeated quantitative endotracheal aspirate positive with P. aeruginosa) findings.

Risk factors. Mechanical ventilation was defined as any period of respiratory support with tracheal intubation.

Statistical Analysis

Incidence of P. aeruginosa bronchopulmonary colonization and pneumonia. The main end points were the incidence of P. aeruginosa bronchopulmonary colonization and pneumonia among all patients in the SICU. First, the crude incidence was estimated as the total number of P. aeruginosa bronchopulmonary colonization and pneumonia cases divided by the total number of exposed patients. Second, time-failure methods, taking into account the various lengths of exposure in the SICU, and allowing to compute the hazard function, which estimates the instantaneous risk of developing colonization within fixed time intervals, were also used. The time of occurrence of colonization was calculated from the date of SICU admission within a maximal observation time in the SICU of 120 d. Patients who died without developing colonization were scored at the time of their death, and patients who were alive without developing colonization were scored on discharge. These estimations were based on the Kaplan-Meier method (15) and actuarial life table methods (16). We also estimated the times to occurrence of colonization with strains belonging to patterns isolated from single patients (called unique patterns) and with strains belonging to patterns isolated from several patients (called multiple patterns) using the Kaplan-Meier method; prognostic values were assessed by the log-rank test at the 5% level (17).

Risk factors for P. aeruginosa bronchopulmonary colonization and pneumonia. Two multivariate cohort analyses were carried out depending on the outcome considered: first, colonization Y/N (patient with tracheal secretions positive for P. aeruginosa whatever the quantitative results and without evidence of tissue invasion versus patients with no P. aeruginosa bronchopulmonary colonization); second infection versus colonization (patients with P. aeruginosa pneumonia versus patients with P. aeruginosa bronchopulmonary tract colonization but without pneumonia).

The characteristics and medical and surgical histories of the patient population included in the cohort analyses are given in Table 1. The same variables were considered to identify risk factors for the development of either colonization or pneumonia with P. aeruginosa using univariate conditional logistic regressions. Only variables significantly associated with colonization or infection are reported in Tables 3 and 6. Relative risks were estimated by exponentiation of regression coefficients and their 95% confidence intervals (CIs). To adjust for confounding factors, variables with a p value below the 10% significance level in univariate analysis were entered in multiple logistic regression models. The Khi2 statistic was used to perform correlation between the variables (p value below the 5% level was considered as significant). The statistical analysis was performed using B.M.D.P. software packages.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Incidence of Bronchopulmonary Tract Colonization

During the 1-yr study period, 304 patients were admitted to the SICU: 114 were excluded from the study because of duration of stay of less than 3 d. The pulmonary tract of 44 of the 190 patients (23.2%) were colonized with P. aeruginosa. The tracheal secretions of 34 patients contained 10⁷ or more CFU/ ml and those of 10 patients fewer than 10⁵ CFU/ml. Eight patients were colonized with P. aeruginosa on admission in the SICU and 36 patients acquired the colonization during the SICU stay, the mean of days for onset of acquired colonization was 18 d (range, 4 to 100 d). The rate of SICU-acquired bronchopulmonary colonization with P. aeruginosa was 10.36 per 1,000 d of mechanical ventilation.

The 7-d and 14-d Kaplan-Meier rates of colonization with P. aeruginosa were estimated to be 2.21 (standard deviation, SE = 1.09%) and 7.03% (SE = 2.2%), respectively. The instantaneous risk of colonization with P. aeruginosa over time was estimated to be 0.32% within the first week and 1.56% within the second week.

Thirteen patients suffered from pneumonia with P. aeruginosa: one patient was colonized on admission and 12 patients had acquired colonization during SICU hospitalization. Seven of the 12 patients who acquired colonization were colonized prior to development of VAP, with a range of days between colonization and VAP from 7 to 12. No P. aeruginosa pneumonia was observed before Day 5 after SICU admission. Cultures obtained from tracheal aspirates showed greater than 10⁷ organisms/ml. Of the 21 patients with tracheal secretion with > 10⁷ organisms/ml and without pneumonia, the colony counts of 14 dropped on subsequents samples (seven were not treated with antibiotic; three were treated with vancomicin; three were treated with potent antipseudomonal antibiotic: one with imipenem and two with piperacillin/clavulanic acid), seven remained > 10⁷ UFC when they were discharged from the SICU: three patients with subsequent samples with > 10⁷ UFC (two patients were not treated with antibiotic and one patient was treated with ceftazidime/amikacin) and four patients without subsequent samples before discharge.

Genotyping Results

One hundred seventeen nonrepetitive P. aeruginosa strains isolated from 66 patients (n = 110; 27 cutaneous specimens, 13 rectal swabs, 25 nasal swabs, 44 aspirate tracheal secretions, and 1 urine) and from the environment (n = 7) showed 68 distinct major DNA patterns. Among the bronchopulmonary tract isolates, there were 21 different unique DNA patterns isolated from single patients (20 from SICU-acquired colonization) and 10 multiple patterns (Figure 1). The sites of colonization with strains with these patterns, the number of colonized patients, and the chronology of isolation are shown in Table 2. Nine among the 10 patterns included only two to three patients, and one was a really epidemic pattern, including six patients. Each of the seven isolates from the environment had a different DNA pattern: three were isolated from the water supply in the rooms of three patients colonized with strains with the same DNA patten, four were not found in patients. Restriction fragment length polymorphisms (RFLP) of bronchopulmonary tract isolates were identical to those of other sites of colonization from the same patient except for three patients (one patient colonized with a unique pattern strain and two patients colonized with multiple pattern strains). The DNA pattern of P. aeruginosa pneumonia isolates were always similar to those found in these patients earlier during colonization. Among the 19 patients who acquired colonization with unique pattern stains in several sites during the SICU stay, 16 remained negative for fecal carriage, 12 for cutaneous sites and six for nasal sites. There was not significant lag between colonization of bronchopulmonary tract and nasal sites in nine patients and between colonization of bronchopulmonary tract and fecal carriage or cutaneous sites in one patient. Four, five and zero patients, respectively, became positive 1 wk or more after bronchopulmonary colonization for nasal, cutaneous, and fecal carriage.

View larger version (130K): [in this window] [in a new window]   Figure 1.   DNA patterns of P. aeruginosa bronchopulmonary isolates. Lanes 1, 5, 8, 11: DNA patterns of S. aureus NCTC 8325 after macrorestriction with SmaI. Lanes 2, 3, 4, 6, 7, 9, 10: DNA patterns of P. aeruginosa bronchopulmonary isolates from different patients after macrorestriction with DraI.

Risk Factors for Colonization

The duration of hospitalization in the SICU before colonization was not significantly different between patients colonized with unique DNA patterns and patients colonized with multiple DNA patterns: mean durations in days were 25.39 ± 3.37 and 37.75 ± 6.02, respectively (p = 0.0809). The variables significantly associated with colonization with unique DNA patterns, after univariate as well as multivariate analyses, are listed in Table . The risk of bronchopulmonary tract colonization increased with the length of hospitalization in the unit, with antimicrobial chemotherapy with third-generation cephalosporins poorly active against P. aeruginosa, in cases of surgical emergency, and in alcoholic patients. Among the patients colonized with multiple DNA patterns, there was a case-mix between index cases and SICU-acquired cases, so the analysis of risk factors was not performed.

Exoproteins in Culture Supernatant of Strains Isolated from Sputa

The main characteristics of eight bronchopulmonary tract isolates analyzed in an attempt to characterize the secretory phenotypes are summarized in Table 4. The amounts of exoproteins secreted during culture in liquid medium were very variable, from 0.00065 to 0.48 U/ml for alkaline protease, from 0.000 to 3.6 U/ml for elastase, and from 0.000 to 0.32 U/ml for phospholipase C. There was no correlation between the exoprotein concentrations during in vitro culture and either isolates responsible of colonization versus pneumonia or unique versus multiple DNA pattern isolates. There was no correlation between the production of the three different exoproteins in these subcultures. Two isolates (Isolates 4 and 6) with the same major DNA profile produced different amounts of exoproteins. The isolates produced greater amounts of elastase and of phospholipase when cultured at 30° C than when cultured at 37° C. 

Risk Factors for Pneumonia

The variables significantly associated with pneumonia after univariate as well as multivariate analyses are listed in Table . A clear association was found by multivariate analysis between pneumonia and treatment with metronidazole with a relative risk of 16 and between pneumonia and chronic obstructive pulmonary disease with a relative risk of 37.9. The correlation between the variable "Prior Receipt Metronidazole on admission to the SICU" and "Metronidazole during Hospitalization in the SICU" with each of the other multivariate candidate variables is shown in Table .

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The use of mechanical ventilation during hospitalization in the SICU was a risk factor for acquisition of pneumonia (1) and the pneumonia rate increases with length of exposure, according to Kaplan-Meier estimates: 7-d and 14-d pneumonia rates were 15.8 and 23.4%, respectively (18). Van Saene and colleagues (19) classify VAP into "early" and "late" onset, Day 4 being the cutoff. They found that Pseudomonadaceae are the most common isolates in "late" onset pneumonia; our data, like those of Rello and colleagues (20), agree with this idea. Kaplan-Meier rates of colonization were 7.03 after versus 2.21 before Day 8, a result confirmed by actuarial life table method and multivariate analysis of risk factors. The American Thoracic Society in its consensus statement (21) states that early ventilator-associated pneumonia and late-onset ventilator- associated pneumonia should be differentiated according to a time period of 5 d. Possibly early, intermediate, and late forms of ventilator-associated pneumonia using 6 to 13 d as cutoff should be described based on the pathogens occurring during this time period and their associated outcomes (22). But we only sampled every 7 d, therefore we did not know if P. aeruginosa ventilated-associated pneumonia was intermediate or late forms. Empirical therapy for these cases of nonearly pneumonia should include antipseudomonal agents until etiologic diagnosis is established.

An important goal of this study of sources and routes for bronchopulmonary tract colonization was to evaluate risk factors that may be amenable to preventive measures. Outbreaks with P. aeruginosa have been described in various settings, including ICUs, burn units, and oncology units. Several environmental sources of P. aeruginosa have been linked to outbreaks of infection: washing basin drains, toilets, and showers. Nevertheless, our studies suggest that colonization with P. aeruginosa occurs at least as often from endogenous as from exogenous sources, confirming the results of Kropec and colleagues (23). Ismaeel (24) did not find a link between intestinal colonization with P. aeruginosa and respiratory tract colonization and infection of ICU patients; strains isolated from the rectum were not the same as those subsequently found in the trachea on the basis of phenotypic markers (24). Niederman and colleagues (25) suggest that P. aeruginosa can first colonize the lower airway and subsequently the oropharynx, but no epidemiologic markers were used in this study. In our study, DNA typing of sequential isolates from individual patients confirmed the possible initial lower airway carriage of P. aeruginosa. Thus, the pattern and the route of tracheobronchial colonization with P. aeruginosa are dissimilar from those of enteric gram-negative bacteria. When P. aeruginosa strains were from exogenous sources, the analysis of the sequence of colonization and the results of the survey of the environment showed that the main route of colonization was cross-transmission from patient to patient probably via staff hands during endotracheal suctioning. Most hospital-acquired cross infection is believed to be transmitted by healthcare workers who fail to wash their hands adequately (26), whereas there is little evidence for a significant role of airborne transmission and of the inanimate environment (27). Routine handwashing before and between patient contact is simple and inexpensive, but nevertheless a very low level of compliance is reported, for several reasons (28). Thus, closed-suction system may be effective in prevention of bronchopulmonary tract cross colonization.

Several studies have reported risk factors for pneumonia in mechanically ventilated patients; few have documented the causative agents of such risk-factor-associated pneumonia. Fagon and colleagues (1) have reported that the distribution of infecting organisms responsible for VAP differs in patients who received prior antimicrobial therapy, with a large increase in multiresistant organisms such as P. aeruginosa or Acinetobacter spp. The previous antimicrobial use was associated with a 5-fold increase in the risk for the development of P. aeruginosa VAP (20, 29). The previous use of lesser potency antipseudomonal third-generation cephalosporins was the second most significant predictor of colonization by P. aeruginosa in our study (RR = 9.73). This suggests that the liberal use of broad-spectrum antimicrobial agents increases the risk of this life-threatening infection (20) and favors a more restrictive antibiotic policy in critically ill patients. The third significant predictor of P. aeruginosa bronchopulmonary colonization was alcoholism (RR = 4.56), a patient-specific risk factor, and the fourth, surgical emergency (RR = 4.55). These two risk factors identified using a multivariate statistical approach have not been reported in the literature.

In our study we differentiated between colonization and pneumonia in the risk factors analysis. The presence of COPD was the most significant predictor of infection by P. aeruginosa in colonized patients in our study (RR = 37.9). This risk factor was similarly identified by Celis and colleagues (30) in a multivariate analysis of nosocomial pneumonia in nonmechanically ventilated patients, by Jimenez and colleagues (31) in a study of pneumonia in a respiratory ICU, and by Rello and colleagues (20) in a study of risk factors for P. aeruginosa VAP. The loss of epithelium integrity and the impairment of mucosal clearance predispose patients with COPD to infection, probably by all gram-negative bacilli (32). Previous administration of metronidazole increased the risk of infection in colonized patients (RR = 16.0) in our study, but it was probably simply a confounder linked to some other risk factor that we did not identify and which is associated with the development of pneumonia (Table ).

We investigated the type of colonizing bacteria as levels of exoproteins secreted may influence the rate of pneumonia (33, 34). Negative DNase activity was more common in patients who were simply colonized with P. aeruginosa and conversely more positive in patients with pneumonia, but the difference was not significant (Fisher's exact test: p = 0.5). In our study, exoprotein production in vitro did not correlate with the presence or absence of pneumonia in vivo, but we only studied a small subset of strains. If these results were confirmed on a larger series, then analyses of exoproteins would be of no predictive value for pneumonia. For future advances in prevention, it is probably necessary to take into account the reciprocal importance of density of bacteria and of exoprotein production at the site of infection. A better understanding of interactions between patient epiemiologic characteristics, exoprotein activity in vivo, cell surface fibronectin, antimicrobial exposure, and the incidence of pneumonia should provide additional insight into the pathogenesis of P. aeruginosa VAP and lead to new preventive approaches.

In conclusion, our study may be of help to clinicians seeking to identify patients in the mechanically ventilated population who are at significantly greater risk of P. aeruginosa pneumonia. According to our findings, patients without prior lung disease who develop pneumonia during the first 48 h of mechanical ventilation can be empirically treated using a regimen without antipseudomonal activity.