INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

More than 50% of antibiotic use in intensive care units (ICUs) is instituted to treat presumed respiratory tract infections, especially in mechanically ventilated patients (1). Establishing the diagnosis of ventilator-associated pneumonia (VAP), however, is difficult (2). Generally, a combination of clinical parameters such as fever, leukocytosis, a positive culture of tracheal aspirate, and an infiltrate on chest X-ray are used for this purpose. Although the combination of these criteria may have a high sensitivity, specificity is low (3). As a result, many patients may unnecessarily receive antibiotics. The addition of quantitative cultures of samples obtained with bronchoscopic techniques, such as protected specimen brush (PSB) and bronchoalveolar lavage (BAL), to these clinical parameters increases specificity and may decrease unnecessary antibiotic use.

Although many studies have determined predictive values of positive or negative results from samples obtained by PSB and BAL for the diagnosis of pneumonia in selected groups of ventilated patients, little data is available on the clinical consequences of the implementation of these techniques in daily practice. We prospectively studied whether adding the results of PSB and BAL to the generally used criteria for diagnosing VAP influenced antibiotic prescription by intensive care physicians. In addition, we analyzed some costs associated with diagnosis and antimicrobial therapy of VAP.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The study was performed in the 16-bed general ICU of the University Hospital Maastricht. All patients admitted between January 1 and December 31, 1994 were included. Demographic characteristics and APACHE II scores were recorded on admission, and clinical parameters were followed daily. All episodes of infection and antibiotic use were registered prospectively.

In mechanically ventilated patients, a clinical suspicion of pneumonia was defined by the criteria listed in Table 1. These clinical, microbiologic, and radiographic criteria were used as an indication to perform bronchoscopy with PSB and BAL as described elsewhere (4). Depending on the results of bronchoscopy, the cases of a clinical suspicion of pneumonia were divided into three groups. Group 1 contained those patients for whom the clinical suspicion of pneumonia was microbiologically proven (e.g., when quantitative cultures of PSB yielded  10³ cfu/ml and/or of BAL yielded  10⁴ cfu/ml). Group 2 contained those patients in whom the clinical suspicion of pneumonia was not confirmed (e.g., when the quantitative cultures of PSB and BAL were sterile or below the above-mentioned cutoff points). Finally, Group 3 contained those patients that had not undergone bronchoscopy.

Antibiotic therapy was prescribed by the responsible attending critical care physician, and was in no way dictated by protocol. However, the choice of compounds was guided by the hospital formulary. In general, the results of gram staining of samples obtained by BAL were available within 2 h after bronchoscopy, preliminary culture results were available after 24 h, and define culture results with in vitro susceptibility testing results were available after 48 h. Empirical therapy was directed towards pathogens isolated from clinical and surveillance cultures of tracheal aspirates, and patients with VAP were treated with antibiotics for 2 wk. Surveillance cultures of tracheal aspirates were obtained on admission and twice weekly from all patients. In case of a suspicion of VAP, clinical cultures from tracheal aspirates were obtained unless surveillance cultures had been taken within 24 h. In the follow-up of each patient, the duration of antibiotic therapy and the recurrence of a new clinical suspicion of pneumonia was registered.

In case of a clinical suspicion of pneumonia, antibiotic therapy was only instituted after bronchoscopy had been performed. When a patient was receiving antibiotics > 48 h for another unrelated infection or for prophylaxis on the day that the criteria for a clinical suspicion of pneumonia were met, antibiotics were not changed until bronchoscopy.

Antibiotic consumption was analyzed in prescribed daily doses (PDD). The costs of antibiotic therapy were calculated using the prices per PDD paid by the hospital pharmacy, including the costs of monitoring serum antibiotic concentrations for aminoglycosides and vancomycin (average three times weekly). The costs of bronchoscopy as used in this study ($453.00) (1 U.S. dollar = 1.70 Dutch guilder) were the actual costs calculated by the hospital and included physician's charges and costs for material and microbiologic analysis.

Data are presented as mean values ± SD and analyzed by the chi-square test for categorical variables and Student's t test for parametric continuous variables.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

A total of 469 patients (515 admissions) were admitted to the ICU during the study period, and 274 (58%) patients were intubated and mechanically ventilated for a mean duration of 13 d. The majority of the patients were admitted because of medical (n = 182, ventilated n = 89) or surgical (n = 197, ventilated n = 140) reasons. The criteria for a clinical suspicion of pneumonia were met in 155 patients. Bronchoscopy was performed in 138 cases: BAL was done in 126 episodes, PSB in 104 cases; both procedures were done in 92 cases. No bronchoscopy was performed in 17 episodes because of medical contraindications such as hypoxemia despite ventilation with high positive end-expiratory pressure (PEEP) and high FI_O2 (n = 6), critical clinical situation (n = 5), neurologic objections (n = 2), coagulopathy (n = 1), or technical problems (n = 3). The mean APACHE II scores on admission were comparable for patients in each of the three groups: 19 ± 6 for Group 1, 19 ± 7 for Group 2, and 18 ± 6 for Group 3 (p = NS).

Microbiologic Results

Group 1 (patients in whom the clinical suspicion of pneumonia was microbiologically confirmed) contained 72 patients. These episodes of VAP were polymicrobial in 30 cases and were caused by Pseudomonas aeruginosa (n = 30), Staphylococcus aureus (n = 18), Enterobacteriaceae (n = 44), Streptococci (n = 7), Haemophilus influenzae (n = 7), and Xanthomonas maltophilia (n = 1). In 42 patients (58%), culture results of tracheal aspirates and bronchoscopy were concordant (e.g., the same species of microorganisms were isolated from tracheal aspirates and bronchoscopic samples); in 19 patients (26%), more species of bacteria were isolated from tracheal aspirates than from bronchoscopic samples; and in 11 patients (15%), species of bacteria were isolated at concentrations above the threshold for VAP from bronchoscopic samples that had not been cultured from tracheal aspirates. Twenty patients (28%) were on antibiotics when bronchoscopy was performed.

Group 2 (patients in whom the clinical suspicion was not confirmed microbiologically) contained 66 patients. In 35 patients (53%), culture results from tracheal aspirates and bronchoscopic samples were concordant; in 24 patients (36%), more species of bacteria were recovered from tracheal aspirates than from bronchoscopy; and in 7 patients (11%), bacteria were isolated from bronchoscopic samples that had not been cultured in tracheal aspirates. The diagnosis of VAP was not confirmed microbiologically because no bacteria grew from bronchoscopic samples (n = 37, 56%) or because quantitative culture results did not reach the predefined cutoff points (n = 29, 44%). Six patients without bacterial growth in previously obtained cultures of tracheal aspirates had nonsignificant bacterial growth in bronchoscopic samples. Twenty-eight patients (42%) were on antibiotics when bronchoscopy was performed, but in all cases for > 48 h.

Antibiotic Use

In Group 1, empiric antibiotic therapy was instituted directly after bronchoscopy in 40 (56%) patients (Group 1a) and, in the remaining 32 patients, antibiotics were started after culture results were available (Group 1b) (Figure 1). Antibiotic therapy was adjusted after culture results of bronchoscopy were available in 21 (53%) of 40 patients in Group 1a. In 11 patients, the spectrum of empiric antibiotic therapy, which was based on the bacteria isolated from previously obtained tracheal aspirates, could be narrowed. The reasons to adjust empiric antibiotic therapy in the remaining 10 patients were as follows: the occurrence of line sepsis at Day 2 (n = 2); abdominal sepsis at Day 4 (n = 1); a drug-induced rash (n = 1); the choice of an agent with better penetration because the patient also had intra-abdominal abscesses (n = 1); colonization of tracheal aspirates with a new pathogen at Day 5 (n = 1); clinical deterioration of the patient despite in vitro susceptibility for the empirical treatment of the pathogen causing VAP (n = 1); and in vitro resistance to the empiric antibiotic therapy (n = 1). Finally, in two patients, significant numbers of a pathogen that had not been isolated from tracheal aspirates were isolated from bronchoscopic samples and these pathogens were not covered by empirical therapy. Therefore, the changes in antibiotics did reflect a correction of inappropriate empirical antibiotic therapy in three patients. As a result, in patients in whom VAP was microbiologically confirmed by quantitative cultures from bronchoscopic samples, antibiotic therapy was adjusted because of these culture results in 14 patients (19%); by narrowing of the antibiotic spectrum in 11 patients because of a pathogen not isolated from tracheal aspirates in two patients and because of resistance to empirical therapy in another patient.

View larger version (25K): [in this window] [in a new window]   Figure 1.   Number of patients who underwent bronchoscopy, in whom the clinical suspicion of VAP was confirmed or not, and in whom antibiotic therapy was instituted or adjusted.

In patients in Group 2, empiric antibiotic therapy was instituted directly after bronchoscopy in 34 (52%) of 66 patients (Group 2a), and antibiotic therapy was discontinued within 48 h (when culture results were available) in 17 patients (Figure 1). The decision not to discontinue antibiotic therapy in the other 17 patients was made by the attending critical care physicians. In all cases, there was a high clinical suspicion of VAP and, in two patients, this decision was possibly influenced by a concomittant abdominal source of infection. No antibiotic therapy was instituted in 32 patients (Group 2b). However, in one of these patients, antibiotic therapy was started 2 d after bronchoscopy because of a deteriorating clinical condition of the patient, despite nonsignificant culture results. This patient was treated with antibiotics for 7 d with a good clinical response and was discharged from the ICU 15 d after initiating of therapy. Antibiotic therapy was instituted in 17 (100%) patients in Group 3.

When patients were analyzed with regard to institution of empirical antibiotic therapy, antibiotics were started immediately after bronchoscopy in 74 patients (Groups 1a and 2a). The clinical suspicion of VAP was subsequently confirmed in 40 of these patients (Group 1a); antibiotics were continued in all of them, and therapy was adjusted in 14 patients (35%). In Group 2a (VAP not confirmed), antibiotics were discontinued within 48 h in 17 patients (50%). In 64 patients, the decision to start antibiotics was postponed until culture results of bronchoscopic samples were available (Groups 1b and 2b). Among these patients, the diagnosis of VAP was confirmed in 32 patients (Group 1b), and all were treated with antibiotics. In Group 2b, one patient received antibiotics.

Consequently, among 66 patients in whom a clinical suspicion of pneumonia was not confirmed by bronchoscopic culture results (Group 2), 18 (27%) patients received a full course of antibiotic therapy, while 48 patients (73%) did not; 17 were treated for < 48 h, and 31 were not treated. Overall, antibiotic therapy was withheld in 48 (35%) of 138 patients who had undergone bronchoscopy (Groups 1 and 2) and in none of 17 patients who did not undergo bronchoscopy (Group 3) (p = 0.0002). In addition, in 14 of 72 patients (19%) in whom VAP was confirmed by bronchoscopy, empiric therapy based on previously obtained tracheal aspirates was adjusted. The mean duration of antibiotic therapy in the ICU was 10.6 d in Group 1, 8.4 d in Group 3, and 9.6 d for those patients in Group 2 (n = 18) who were treated despite nonsignificant culture results. The significant difference in duration of antibiotic therapy between Groups 1 and 3 (p < 0.05) was a result of the fact that more patients in Group 3 (10 of 17 [58%]) were discharged from the ICU while still receiving antibiotics as compared with patients in Group 1 (22 of 72 [31%]). In Group 2, nine (50%) of 18 patients were still treated with antibiotics for the clinical suspicion of pneumonia when discharged from the ICU.

Recurrence of a Clinical Suspicion of Pneumonia and Mortality

A second clinical suspicion of pneumonia occurred in seven (15%) of 48 patients in whom antibiotic therapy was withheld because of bronchoscopy results and in 18 (17%) of 107 patients who had been treated with antibiotics (p = NS). The duration until the second clinical suspicion of pneumonia was similar for patients who had or had not received antibiotic therapy: 15.8 and 13.1 d, respectively (p = NS).

Hospital mortality was 35% (17 of 48) in patients in whom antibiotic therapy had been withheld and 39% (42 of 107) in patients who had received antibiotics (p = NS). The APACHE II scores on admission were 19 ± 6 for patients who had been treated initially and 18 ± 6 for those in whom antibiotics had been withheld (p = NS). There were no significant differences in recurrence rates of clinical episodes of pneumonia or mortality rates between the three study groups (data not shown). For patients in Group 1, mortality rates were 50% (n = 7) for patients in whom initial antibiotic therapy was changed based on bronchoscopic culture results (n = 14), 35% (n = 9) for those with no change of therapy (n = 26), and 47% (n = 15) for those who received therapy when culture results were available. Statistical comparison of these groups did not approach significance (p  0.34). The delay in starting antibiotics in Group 1b did not influence outcome: the mortality rate in this group was 47% as compared with 40% in Group 1a (p = NS), and the APACHE II scores in both groups were 20 ± 6 and 19 ± 6, respectively (p = NS). For patients in Group 2, mortality rates were comparable for patients with no growth in bronchoscopic samples (28%) and those with growth below the cutoff point for VAP (37%) (p = NS).

Cost Analysis

The mean costs of antibiotic therapy per episode of VAP, calculated from all episodes of pneumonia that were treated, were $549.00, corresponding to $39.00 per day of treatment. The actual costs of an episode of bronchoscopy (including physicians' charges and microbiologic analysis) were $453.00. As a result, 48 episodes of suspected VAP were not treated (projected costs: $26,352.00) at the expense of 138 episodes of bronchoscopy (actual costs: $62,514.00). The costs of nursing and pharmacy administration as well as the costs of antibiotics that were given for > 48 h (total: $1,481.00) were not included in this analysis.

Under the circumstances tested, the introduction of bronchoscopic techniques would only be cost-effective if the costs of bronchoscopy would be $191.00, when the proportion of patients in whom therapy would be withheld would increase to 114 (83%) of 138, or when the costs of antibiotic therapy per episode of VAP would be $1,302.00, corresponding to $93.00 per day.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The results of this study suggest that implementation of bronchoscopic techniques to the criteria for diagnosing VAP may help to reduce antibiotic use in ICUs. In addition, empiric antimicrobial therapy was adjusted in 19% of the patients with proven VAP, in most cases by narrowing the antibiotic coverage. In patients with a clinical suspicion of pneumonia but with nonsignificant quantitative cultures from respiratory samples obtained by bronchoscopic techniques, withholding of antibiotic therapy was not associated with an increased mortality rate or a higher recurrence rate for VAP.

Treatment of respiratory infections accounts for the largest part of antibiotic consumption in ICUs. In a recent 1-d point-prevalence study of nosocomial infections in 1,047 ICUs in 17 European countries including 10,038 patients, 44.8% of all patients had one or more infection, mostly nosocomial, and 64.7% of these infections were respiratory tract related (1). On the day of study, 62.3% of all patients received antibiotics; 51% received more than one agent (1). However, in that study, bronchoscopic techniques were not used to diagnose respiratory tract infections, which probably resembles the current situation in many ICUs. As a result, the prevalence of respiratory tract infections may have been overestimated, which may have led to unnecessary antibiotic use.

We implemented bronchoscopic techniques in the routine workup of a clinical suspicion of nosocomial pneumonia in ventilated patients. A clinical suspicion of pneumonia was based on a combination of clinical, radiographic, and microbiologic criteria, which are generally used in clinical practice. These criteria have a high sensitivity but a low specificity for VAP (5, 6). Many studies have demonstrated that addition of quantitative culture results of PSB or BAL increases the specificity for VAP in selected patient groups (2, 6, 7). However, few studies prospectively analyzed the effects of implementation of these new diagnostic techniques on clinical practice. The results of the present study show that addition of bronchoscopic techniques to the diagnostic criteria for pneumonia resulted in adjustment of empirical antimicrobial therapy in 19% of patients with bronchoscopically proven episodes of VAP and in withholding of antibiotics in 48 (35%) of 138 patients. However, the study design employed does not allow us to conclude that bronchoscopy alone was responsible for this reduction. For example, in some patients fulfilling the criteria of pneumonia, the physicians' clinical judgment allowed them to withhold antibiotics, and bronchoscopy confirmed this clinical judgment. Importantly, withholding of antibiotics in these cases did not result in a higher recurrence rate of a clinical suspicion of pneumonia during the same period of admission and was not associated with a higher mortality rate. The absence of differences in mortality was compatible with similar APACHE II scores on admission in these patient groups. On the other hand, there were 17 patients who were continued on antibiotics because of the physicians' clinical judgment, despite negative results from bronchoscopy. In these cases, the responsible physicians did feel reluctant to discontinue antibiotics. The decision to institute and continue antibiotic therapy was in the hands of the responsible attending physicians and was in no way prescribed by protocol. Indeed, false-negative results of bronchoscopic techniques have been reported and may occur because of prior antibiotic therapy or because of "borderline" results of quantitative analysis, possibly indicating that the infection is in an early stage (8).

Our data confirm data reported by Croce and coworkers (9). They calculated that addition of bronchoscopic techniques for diagnosing VAP could significantly reduce associated charges. In a pilot study of 10 trauma patients, the diagnosis of VAP was rejected in seven patients. Discontinuation of empiric therapy had no negative influence on the clinical condition of these patients and, in that medical center, withholding of complete therapy resulted in net savings of $17,060.12 (9). In contrast, other investigators reported that addition of BAL for diagnosing VAP did not improve patient outcome (10). In that study, the adequacy of antibiotic therapy, administered before VAP was diagnosed, seemed to be the most important variable influencing outcome: 50 (77%) of 65 patients in which the clinical suspicion of VAP was confirmed by BAL received antibiotics before bronchoscopy was performed. If these antibiotics were adequate for the microorganisms recovered from BAL, mortality was 38%, whereas mortality rates were 91% and 60% when these antibiotics were inadequate or when no antibiotics had been administered, respectively (10). However, since in that study, 77% of patients with VAP received antibiotics before bronchoscopy, and overall mortality was 67% (similar for patients with and without confirmed pneumonia), with 50% of the nonsurvivors dying within 48 h after diagnosis of VAP, it seems difficult to compare the findings in that population to our situation.

Cost-effectiveness is an important criterion for introduction of a new diagnostic test. The present study shows that, in our ICU, addition of bronchoscopic techniques to the generally used criteria for VAP may have helped to reduce antibiotic therapy in 35% of cases. In 17 patients, antibiotics were discontinued because of bronchoscopy results, but, as discussed above, in those cases that empirical antibiotic therapy was withheld (in 31 patients); bronchoscopy just helped to confirm the clinical judgment of the physician. The superficial analysis of costs as performed should not be regarded as an appropriate cost-benefit analysis. However, the results show that a formal cost-benefit analysis will be influenced by variables that will not be the same among different settings. For instance, antibiotic therapy depends on local susceptibilities of microorganisms, with a general rule that more resistant bacteria must be treated with more expensive antibiotics. In our ICU, all strains of S. aureus were susceptible to methicillin, extended-spectrum beta-lactamases had not been isolated, and all strains of P. aeruginosa were susceptible to piperacillin and gentamicin. Therefore, the calculated costs for therapy in the present study ($39.00 per day) probably are low when compared with other centers with more resistant pathogens. For example, in the study of Croce and coworkers performed in the United States, standard empiric therapy (ceftazidime and vancomycin) cost $221.56 per day (9). Another important variable is the proportion of "negative" culture results of bronchoscopic techniques. The incidence of suspected pneumonia, based on the generally used combination of radiographic, clinical, and microbiologic criteria, may seem high in our study. However, our study population consisted of patients with a mean duration of ventilation of 13 d and the proportion in which the clinical suspicion was not confirmed (48%) is similar to or lower than those reported by others. Timsit and associates studied 112 patients with a clinical suspicion of VAP and this suspicion was confirmed by quantitative cultures from samples obtained by bronchoscopic techniques in 50% (11), and Luna and colleagues reported that BAL did not confirm a clinical suspicion of pneumonia in 49% (10). Others, however, excluded VAP after performance of PBS in 71% of the cases (7), thereby increasing the "benefit" of bronchoscopy. In addition, the costs for bronchoscopy, although regarded as expensive, may differ between centers. Interestingly, several recent reports suggested good correlations between quantitative culture results of PSB or BAL with quantitative culture results of samples obtained by deep endotracheal aspiration (12, 13) or blind PSB (14). If these cheaper, and less invasive, techniques prove to be equivalent to quantitative cultures of bronchoscopic techniques, they are much more likely to be cost- effective for diagnosing VAP. Besides the financial aspects of diagnosing VAP, aspects like toxic effects of unnecessary prescribed antibiotics and their long-term influence on antibiotic susceptibilities should be considered. On the other hand, complications of bronchoscopy, although occurring infrequently, include a drop in Pa_O2, arrhythmias, pneumothorax, transient progression of pulmonary infiltrates, and fever (15, 16). No serious adverse effects of bronchoscopy were encountered during the present study.

In conclusion, bronchoscopic techniques may be helpful and safe tools in the diagnosis of VAP. Under the circumstances tested, implementation of these techniques in the routine workup of a clinical suspicion of VAP influenced clinical practice, decreasing the number of patients receiving antibiotic therapy. Withholding antibiotics to these patients did not affect the recurrence rate of a clinical suspicion of pneumonia and was not associated with a higher mortality rate. The clinical impacts and a formal cost-benefit analysis of these new diagnostic tools should be determined in a prospective randomized fashion.